open-nomad/command/agent/config.go

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

2408 lines
73 KiB
Go
Raw Normal View History

2015-08-16 01:54:41 +00:00
package agent
import (
"encoding/base64"
"errors"
2015-08-16 01:54:41 +00:00
"fmt"
"io"
"net"
2015-08-16 01:54:41 +00:00
"os"
"os/exec"
2015-08-16 01:54:41 +00:00
"path/filepath"
2015-10-29 12:30:26 +00:00
"runtime"
2015-12-15 07:05:25 +00:00
"sort"
"strconv"
2015-08-16 01:54:41 +00:00
"strings"
"time"
2015-08-16 01:54:41 +00:00
"github.com/hashicorp/go-secure-stdlib/listenerutil"
sockaddr "github.com/hashicorp/go-sockaddr"
"github.com/hashicorp/go-sockaddr/template"
2015-09-06 01:41:00 +00:00
client "github.com/hashicorp/nomad/client/config"
"github.com/hashicorp/nomad/client/fingerprint"
2017-01-18 23:55:14 +00:00
"github.com/hashicorp/nomad/helper"
"github.com/hashicorp/nomad/helper/pointer"
"github.com/hashicorp/nomad/helper/users"
2015-09-06 01:41:00 +00:00
"github.com/hashicorp/nomad/nomad"
2018-10-02 20:36:04 +00:00
"github.com/hashicorp/nomad/nomad/structs"
"github.com/hashicorp/nomad/nomad/structs/config"
"github.com/hashicorp/nomad/version"
"golang.org/x/exp/maps"
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
"golang.org/x/exp/slices"
2015-08-16 01:54:41 +00:00
)
// Config is the configuration for the Nomad agent.
2019-04-25 20:47:03 +00:00
//
// time.Duration values have two parts:
// - a string field tagged with an hcl:"foo" and json:"-"
// - a time.Duration field in the same struct and a call to duration
// in config_parse.go ParseConfigFile
2019-04-25 20:47:03 +00:00
//
// All config structs should have an ExtraKeysHCL field to check for
// unexpected keys
2015-08-16 01:54:41 +00:00
type Config struct {
// Region is the region this agent is in. Defaults to global.
Region string `hcl:"region"`
2015-08-16 01:54:41 +00:00
2015-08-31 01:10:23 +00:00
// Datacenter is the datacenter this agent is in. Defaults to dc1
Datacenter string `hcl:"datacenter"`
2015-08-31 01:10:23 +00:00
// NodeName is the name we register as. Defaults to hostname.
NodeName string `hcl:"name"`
2015-08-31 01:10:23 +00:00
2015-08-16 01:54:41 +00:00
// DataDir is the directory to store our state in
DataDir string `hcl:"data_dir"`
2015-08-16 01:54:41 +00:00
2018-08-30 20:43:09 +00:00
// PluginDir is the directory to lookup plugins.
PluginDir string `hcl:"plugin_dir"`
2018-08-30 20:43:09 +00:00
2019-01-11 19:36:26 +00:00
// LogLevel is the level of the logs to put out
LogLevel string `hcl:"log_level"`
2015-08-16 01:54:41 +00:00
2019-01-11 19:36:26 +00:00
// LogJson enables log output in a JSON format
LogJson bool `hcl:"log_json"`
2019-01-11 19:36:26 +00:00
// LogFile enables logging to a file
LogFile string `hcl:"log_file"`
// LogRotateDuration is the time period that logs should be rotated in
LogRotateDuration string `hcl:"log_rotate_duration"`
// LogRotateBytes is the max number of bytes that should be written to a file
LogRotateBytes int `hcl:"log_rotate_bytes"`
// LogRotateMaxFiles is the max number of log files to keep
LogRotateMaxFiles int `hcl:"log_rotate_max_files"`
// BindAddr is the address on which all of nomad's services will
// be bound. If not specified, this defaults to 127.0.0.1.
BindAddr string `hcl:"bind_addr"`
2015-08-24 00:50:33 +00:00
// EnableDebug is used to enable debugging HTTP endpoints
EnableDebug bool `hcl:"enable_debug"`
2015-08-24 00:50:33 +00:00
// Ports is used to control the network ports we bind to.
Ports *Ports `hcl:"ports"`
// Addresses is used to override the network addresses we bind to.
//
// Use normalizedAddrs if you need the host+port to bind to.
Addresses *Addresses `hcl:"addresses"`
// normalizedAddr is set to the Address+Port by normalizeAddrs()
normalizedAddrs *NormalizedAddrs
// AdvertiseAddrs is used to control the addresses we advertise.
AdvertiseAddrs *AdvertiseAddrs `hcl:"advertise"`
2015-08-16 01:54:41 +00:00
// Client has our client related settings
Client *ClientConfig `hcl:"client"`
2015-08-16 01:54:41 +00:00
// Server has our server related settings
Server *ServerConfig `hcl:"server"`
2015-08-16 01:54:41 +00:00
2017-08-13 20:46:47 +00:00
// ACL has our acl related settings
ACL *ACLConfig `hcl:"acl"`
2017-08-13 20:46:47 +00:00
// Telemetry is used to configure sending telemetry
Telemetry *Telemetry `hcl:"telemetry"`
2015-08-16 20:54:49 +00:00
// LeaveOnInt is used to gracefully leave on the interrupt signal
LeaveOnInt bool `hcl:"leave_on_interrupt"`
2015-08-16 21:34:38 +00:00
// LeaveOnTerm is used to gracefully leave on the terminate signal
LeaveOnTerm bool `hcl:"leave_on_terminate"`
2015-08-16 21:34:38 +00:00
// EnableSyslog is used to enable sending logs to syslog
EnableSyslog bool `hcl:"enable_syslog"`
2015-08-24 00:40:27 +00:00
// SyslogFacility is used to control the syslog facility used.
SyslogFacility string `hcl:"syslog_facility"`
// DisableUpdateCheck is used to disable the periodic update
// and security bulletin checking.
DisableUpdateCheck *bool `hcl:"disable_update_check"`
// DisableAnonymousSignature is used to disable setting the
// anonymous signature when doing the update check and looking
// for security bulletins
DisableAnonymousSignature bool `hcl:"disable_anonymous_signature"`
2015-09-06 01:41:00 +00:00
// Consul contains the configuration for the Consul Agent and
// parameters necessary to register services, their checks, and
// discover the current Nomad servers.
Consul *config.ConsulConfig `hcl:"consul"`
2016-03-31 06:20:24 +00:00
2016-08-06 01:13:06 +00:00
// Vault contains the configuration for the Vault Agent and
// parameters necessary to derive tokens.
Vault *config.VaultConfig `hcl:"vault"`
2016-08-06 01:13:06 +00:00
// UI is used to configure the web UI
UI *config.UIConfig `hcl:"ui"`
2015-09-06 01:41:00 +00:00
// NomadConfig is used to override the default config.
2018-03-11 18:21:52 +00:00
// This is largely used for testing purposes.
NomadConfig *nomad.Config `hcl:"-" json:"-"`
2015-09-06 01:41:00 +00:00
// ClientConfig is used to override the default config.
2018-03-11 18:21:52 +00:00
// This is largely used for testing purposes.
ClientConfig *client.Config `hcl:"-" json:"-"`
// DevMode is set by the -dev CLI flag.
DevMode bool `hcl:"-"`
// Version information is set at compilation time
Version *version.VersionInfo
// List of config files that have been loaded (in order)
Files []string `hcl:"-"`
2016-10-24 20:46:22 +00:00
// TLSConfig provides TLS related configuration for the Nomad server and
// client
TLSConfig *config.TLSConfig `hcl:"tls"`
// HTTPAPIResponseHeaders allows users to configure the Nomad http agent to
2018-03-11 17:40:03 +00:00
// set arbitrary headers on API responses
HTTPAPIResponseHeaders map[string]string `hcl:"http_api_response_headers"`
2017-09-19 14:47:10 +00:00
// Sentinel holds sentinel related settings
Sentinel *config.SentinelConfig `hcl:"sentinel"`
// Autopilot contains the configuration for Autopilot behavior.
Autopilot *config.AutopilotConfig `hcl:"autopilot"`
2018-08-29 23:26:44 +00:00
// Plugins is the set of configured plugins
Plugins []*config.PluginConfig `hcl:"plugin"`
// Limits contains the configuration for timeouts.
Limits config.Limits `hcl:"limits"`
// Audit contains the configuration for audit logging.
Audit *config.AuditConfig `hcl:"audit"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
2015-08-16 01:54:41 +00:00
}
// ClientConfig is configuration specific to the client mode
2015-08-16 01:54:41 +00:00
type ClientConfig struct {
// Enabled controls if we are a client
Enabled bool `hcl:"enabled"`
2015-08-16 01:54:41 +00:00
2015-08-31 01:10:23 +00:00
// StateDir is the state directory
StateDir string `hcl:"state_dir"`
2015-08-16 01:54:41 +00:00
2015-08-31 01:10:23 +00:00
// AllocDir is the directory for storing allocation data
AllocDir string `hcl:"alloc_dir"`
2015-08-31 01:10:23 +00:00
// Servers is a list of known server addresses. These are as "host:port"
Servers []string `hcl:"servers"`
2015-08-16 01:54:41 +00:00
2015-08-31 01:10:23 +00:00
// NodeClass is used to group the node by class
NodeClass string `hcl:"node_class"`
2015-08-31 01:10:23 +00:00
// Options is used for configuration of nomad internals,
// like fingerprinters and drivers. The format is:
//
// namespace.option = value
Options map[string]string `hcl:"options"`
2015-08-31 01:10:23 +00:00
// Metadata associated with the node
Meta map[string]string `hcl:"meta"`
// A mapping of directories on the host OS to attempt to embed inside each
// task's chroot.
ChrootEnv map[string]string `hcl:"chroot_env"`
// Interface to use for network fingerprinting
NetworkInterface string `hcl:"network_interface"`
2015-10-03 00:32:11 +00:00
// NetworkSpeed is used to override any detected or default network link
// speed.
NetworkSpeed int `hcl:"network_speed"`
// CpuCompute is used to override any detected or default total CPU compute.
CpuCompute int `hcl:"cpu_total_compute"`
// MemoryMB is used to override any detected or default total memory.
MemoryMB int `hcl:"memory_total_mb"`
// ReservableCores is used to override detected reservable cpu cores.
ReserveableCores string `hcl:"reservable_cores"`
// MaxKillTimeout allows capping the user-specifiable KillTimeout.
MaxKillTimeout string `hcl:"max_kill_timeout"`
// ClientMaxPort is the upper range of the ports that the client uses for
// communicating with plugin subsystems
ClientMaxPort int `hcl:"client_max_port"`
// ClientMinPort is the lower range of the ports that the client uses for
// communicating with plugin subsystems
ClientMinPort int `hcl:"client_min_port"`
// MaxDynamicPort is the upper range of the dynamic ports that the client
// uses for allocations
MaxDynamicPort int `hcl:"max_dynamic_port"`
// MinDynamicPort is the lower range of the dynamic ports that the client
// uses for allocations
MinDynamicPort int `hcl:"min_dynamic_port"`
// Reserved is used to reserve resources from being used by Nomad. This can
// be used to target a certain utilization or to prevent Nomad from using a
// particular set of ports.
Reserved *Resources `hcl:"reserved"`
2017-01-31 23:32:20 +00:00
// GCInterval is the time interval at which the client triggers garbage
// collection
GCInterval time.Duration
GCIntervalHCL string `hcl:"gc_interval" json:"-"`
2017-01-31 23:32:20 +00:00
// GCParallelDestroys is the number of parallel destroys the garbage
// collector will allow.
GCParallelDestroys int `hcl:"gc_parallel_destroys"`
// GCDiskUsageThreshold is the disk usage threshold given as a percent
// beyond which the Nomad client triggers GC of terminal allocations
GCDiskUsageThreshold float64 `hcl:"gc_disk_usage_threshold"`
2017-01-31 23:32:20 +00:00
// GCInodeUsageThreshold is the inode usage threshold beyond which the Nomad
// client triggers GC of the terminal allocations
GCInodeUsageThreshold float64 `hcl:"gc_inode_usage_threshold"`
2017-02-27 21:42:37 +00:00
// GCMaxAllocs is the maximum number of allocations a node can have
// before garbage collection is triggered.
GCMaxAllocs int `hcl:"gc_max_allocs"`
2017-02-27 21:42:37 +00:00
// NoHostUUID disables using the host's UUID and will force generation of a
// random UUID.
NoHostUUID *bool `hcl:"no_host_uuid"`
// DisableRemoteExec disables remote exec targeting tasks on this client
DisableRemoteExec bool `hcl:"disable_remote_exec"`
// TemplateConfig includes configuration for template rendering
TemplateConfig *client.ClientTemplateConfig `hcl:"template"`
// ServerJoin contains information that is used to attempt to join servers
ServerJoin *ServerJoin `hcl:"server_join"`
// HostVolumes contains information about the volumes an operator has made
// available to jobs running on this node.
HostVolumes []*structs.ClientHostVolumeConfig `hcl:"host_volume"`
// CNIPath is the path to search for CNI plugins, multiple paths can be
// specified colon delimited
CNIPath string `hcl:"cni_path"`
// CNIConfigDir is the directory where CNI network configuration is located. The
// client will use this path when fingerprinting CNI networks.
CNIConfigDir string `hcl:"cni_config_dir"`
// BridgeNetworkName is the name of the bridge to create when using the
// bridge network mode
BridgeNetworkName string `hcl:"bridge_network_name"`
// BridgeNetworkSubnet is the subnet to allocate IP addresses from when
// creating allocations with bridge networking mode. This range is local to
// the host
BridgeNetworkSubnet string `hcl:"bridge_network_subnet"`
// HostNetworks describes the different host networks available to the host
// if the host uses multiple interfaces
2020-06-16 15:53:10 +00:00
HostNetworks []*structs.ClientHostNetworkConfig `hcl:"host_network"`
// BindWildcardDefaultHostNetwork toggles if when there are no host networks,
// should the port mapping rules match the default network address (false) or
// matching any destination address (true). Defaults to true
BindWildcardDefaultHostNetwork bool `hcl:"bind_wildcard_default_host_network"`
// CgroupParent sets the parent cgroup for subsystems managed by Nomad. If the cgroup
// doest not exist Nomad will attempt to create it during startup. Defaults to '/nomad'
CgroupParent string `hcl:"cgroup_parent"`
// NomadServiceDiscovery is a boolean parameter which allows operators to
// enable/disable to Nomad native service discovery feature on the client.
// This parameter is exposed via the Nomad fingerprinter and used to ensure
// correct scheduling decisions on allocations which require this.
NomadServiceDiscovery *bool `hcl:"nomad_service_discovery"`
// Artifact contains the configuration for artifacts.
Artifact *config.ArtifactConfig `hcl:"artifact"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
2015-08-16 01:54:41 +00:00
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (c *ClientConfig) Copy() *ClientConfig {
if c == nil {
return c
}
nc := *c
nc.Servers = slices.Clone(c.Servers)
nc.Options = maps.Clone(c.Options)
nc.Meta = maps.Clone(c.Meta)
nc.ChrootEnv = maps.Clone(c.ChrootEnv)
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
nc.Reserved = c.Reserved.Copy()
nc.NoHostUUID = pointer.Copy(c.NoHostUUID)
nc.TemplateConfig = c.TemplateConfig.Copy()
nc.ServerJoin = c.ServerJoin.Copy()
nc.HostVolumes = helper.CopySlice(c.HostVolumes)
nc.HostNetworks = helper.CopySlice(c.HostNetworks)
nc.NomadServiceDiscovery = pointer.Copy(c.NomadServiceDiscovery)
nc.Artifact = c.Artifact.Copy()
nc.ExtraKeysHCL = slices.Clone(c.ExtraKeysHCL)
return &nc
}
2017-08-13 20:46:47 +00:00
// ACLConfig is configuration specific to the ACL system
type ACLConfig struct {
// Enabled controls if we are enforce and manage ACLs
Enabled bool `hcl:"enabled"`
2017-08-13 20:46:47 +00:00
// TokenTTL controls how long we cache ACL tokens. This controls
// how stale they can be when we are enforcing policies. Defaults
// to "30s". Reducing this impacts performance by forcing more
// frequent resolution.
TokenTTL time.Duration
TokenTTLHCL string `hcl:"token_ttl" json:"-"`
2017-08-13 20:46:47 +00:00
// PolicyTTL controls how long we cache ACL policies. This controls
// how stale they can be when we are enforcing policies. Defaults
// to "30s". Reducing this impacts performance by forcing more
// frequent resolution.
PolicyTTL time.Duration
PolicyTTLHCL string `hcl:"policy_ttl" json:"-"`
// RoleTTL controls how long we cache ACL roles. This controls how stale
// they can be when we are enforcing policies. Defaults to "30s".
// Reducing this impacts performance by forcing more frequent resolution.
RoleTTL time.Duration
RoleTTLHCL string `hcl:"role_ttl" json:"-"`
// ReplicationToken is used by servers to replicate tokens and policies
// from the authoritative region. This must be a valid management token
// within the authoritative region.
ReplicationToken string `hcl:"replication_token"`
// TokenMinExpirationTTL is used to enforce the lowest acceptable value for
// ACL token expiration. This is used by the Nomad servers to validate ACL
// tokens with an expiration value set upon creation.
TokenMinExpirationTTL time.Duration
TokenMinExpirationTTLHCL string `hcl:"token_min_expiration_ttl" json:"-"`
// TokenMaxExpirationTTL is used to enforce the highest acceptable value
// for ACL token expiration. This is used by the Nomad servers to validate
// ACL tokens with an expiration value set upon creation.
TokenMaxExpirationTTL time.Duration
TokenMaxExpirationTTLHCL string `hcl:"token_max_expiration_ttl" json:"-"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
2017-08-13 20:46:47 +00:00
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (a *ACLConfig) Copy() *ACLConfig {
if a == nil {
return nil
}
na := *a
na.ExtraKeysHCL = slices.Clone(a.ExtraKeysHCL)
return &na
}
// ServerConfig is configuration specific to the server mode
2015-08-16 01:54:41 +00:00
type ServerConfig struct {
// Enabled controls if we are a server
Enabled bool `hcl:"enabled"`
2015-08-16 01:54:41 +00:00
2017-08-13 20:46:47 +00:00
// AuthoritativeRegion is used to control which region is treated as
// the source of truth for global tokens and ACL policies.
AuthoritativeRegion string `hcl:"authoritative_region"`
2017-08-13 20:46:47 +00:00
2015-08-16 01:54:41 +00:00
// BootstrapExpect tries to automatically bootstrap the Consul cluster,
2016-05-15 16:41:34 +00:00
// by withholding peers until enough servers join.
BootstrapExpect int `hcl:"bootstrap_expect"`
2015-08-16 01:54:41 +00:00
2015-08-31 01:10:23 +00:00
// DataDir is the directory to store our state in
DataDir string `hcl:"data_dir"`
2015-08-31 01:10:23 +00:00
// ProtocolVersion is the protocol version to speak. This must be between
// ProtocolVersionMin and ProtocolVersionMax.
//
// Deprecated: This has never been used and will emit a warning if nonzero.
ProtocolVersion int `hcl:"protocol_version" json:"-"`
2015-08-31 01:10:23 +00:00
// RaftProtocol is the Raft protocol version to speak. This must be from [1-3].
RaftProtocol int `hcl:"raft_protocol"`
2020-05-30 13:27:33 +00:00
// RaftMultiplier scales the Raft timing parameters
RaftMultiplier *int `hcl:"raft_multiplier"`
2015-08-31 01:10:23 +00:00
// NumSchedulers is the number of scheduler thread that are run.
// This can be as many as one per core, or zero to disable this server
// from doing any scheduling work.
NumSchedulers *int `hcl:"num_schedulers"`
2015-08-31 01:10:23 +00:00
// EnabledSchedulers controls the set of sub-schedulers that are
// enabled for this server to handle. This will restrict the evaluations
// that the workers dequeue for processing.
EnabledSchedulers []string `hcl:"enabled_schedulers"`
2016-03-04 23:44:12 +00:00
// NodeGCThreshold controls how "old" a node must be to be collected by GC.
// Age is not the only requirement for a node to be GCed but the threshold
// can be used to filter by age.
NodeGCThreshold string `hcl:"node_gc_threshold"`
// JobGCInterval controls how often we dispatch a job to GC jobs that are
// available for garbage collection.
JobGCInterval string `hcl:"job_gc_interval"`
// JobGCThreshold controls how "old" a job must be to be collected by GC.
// Age is not the only requirement for a Job to be GCed but the threshold
// can be used to filter by age.
JobGCThreshold string `hcl:"job_gc_threshold"`
// EvalGCThreshold controls how "old" an eval must be to be collected by GC.
// Age is not the only requirement for a eval to be GCed but the threshold
// can be used to filter by age.
EvalGCThreshold string `hcl:"eval_gc_threshold"`
2017-06-29 18:29:44 +00:00
// DeploymentGCThreshold controls how "old" a deployment must be to be
// collected by GC. Age is not the only requirement for a deployment to be
2017-06-29 18:29:44 +00:00
// GCed but the threshold can be used to filter by age.
DeploymentGCThreshold string `hcl:"deployment_gc_threshold"`
2017-06-29 18:29:44 +00:00
// CSIVolumeClaimGCThreshold controls how "old" a CSI volume must be to
// have its claims collected by GC. Age is not the only requirement for
// a volume to be GCed but the threshold can be used to filter by age.
CSIVolumeClaimGCThreshold string `hcl:"csi_volume_claim_gc_threshold"`
// CSIPluginGCThreshold controls how "old" a CSI plugin must be to be
// collected by GC. Age is not the only requirement for a plugin to be
// GCed but the threshold can be used to filter by age.
CSIPluginGCThreshold string `hcl:"csi_plugin_gc_threshold"`
// ACLTokenGCThreshold controls how "old" an expired ACL token must be to
// be collected by GC.
ACLTokenGCThreshold string `hcl:"acl_token_gc_threshold"`
// RootKeyGCInterval is how often we dispatch a job to GC
// encryption key metadata
RootKeyGCInterval string `hcl:"root_key_gc_interval"`
// RootKeyGCThreshold is how "old" encryption key metadata must be
// to be eligible for GC.
RootKeyGCThreshold string `hcl:"root_key_gc_threshold"`
// RootKeyRotationThreshold is how "old" an encryption key must be
// before it is automatically rotated on the next garbage
// collection interval.
RootKeyRotationThreshold string `hcl:"root_key_rotation_threshold"`
2016-03-04 23:44:12 +00:00
// HeartbeatGrace is the grace period beyond the TTL to account for network,
// processing delays and clock skew before marking a node as "down".
HeartbeatGrace time.Duration
HeartbeatGraceHCL string `hcl:"heartbeat_grace" json:"-"`
// MinHeartbeatTTL is the minimum time between heartbeats. This is used as
// a floor to prevent excessive updates.
MinHeartbeatTTL time.Duration
MinHeartbeatTTLHCL string `hcl:"min_heartbeat_ttl" json:"-"`
// MaxHeartbeatsPerSecond is the maximum target rate of heartbeats
// being processed per second. This allows the TTL to be increased
// to meet the target rate.
MaxHeartbeatsPerSecond float64 `hcl:"max_heartbeats_per_second"`
2016-03-04 23:44:12 +00:00
// FailoverHeartbeatTTL is the TTL applied to heartbeats after
// a new leader is elected, since we no longer know the status
// of all the heartbeats.
FailoverHeartbeatTTL time.Duration
FailoverHeartbeatTTLHCL string `hcl:"failover_heartbeat_ttl" json:"-"`
// StartJoin is a list of addresses to attempt to join when the
// agent starts. If Serf is unable to communicate with any of these
// addresses, then the agent will error and exit.
// Deprecated in Nomad 0.10
StartJoin []string `hcl:"start_join"`
// RetryJoin is a list of addresses to join with retry enabled.
// Deprecated in Nomad 0.10
RetryJoin []string `hcl:"retry_join"`
// RetryMaxAttempts specifies the maximum number of times to retry joining a
// host on startup. This is useful for cases where we know the node will be
// online eventually.
// Deprecated in Nomad 0.10
RetryMaxAttempts int `hcl:"retry_max"`
// RetryInterval specifies the amount of time to wait in between join
// attempts on agent start. The minimum allowed value is 1 second and
// the default is 30s.
// Deprecated in Nomad 0.10
RetryInterval time.Duration
RetryIntervalHCL string `hcl:"retry_interval" json:"-"`
// RejoinAfterLeave controls our interaction with the cluster after leave.
// When set to false (default), a leave causes Consul to not rejoin
// the cluster until an explicit join is received. If this is set to
// true, we ignore the leave, and rejoin the cluster on start.
RejoinAfterLeave bool `hcl:"rejoin_after_leave"`
// (Enterprise-only) NonVotingServer is whether this server will act as a
// non-voting member of the cluster to help provide read scalability.
NonVotingServer bool `hcl:"non_voting_server"`
// (Enterprise-only) RedundancyZone is the redundancy zone to use for this server.
RedundancyZone string `hcl:"redundancy_zone"`
// (Enterprise-only) UpgradeVersion is the custom upgrade version to use when
// performing upgrade migrations.
UpgradeVersion string `hcl:"upgrade_version"`
// Encryption key to use for the Serf communication
EncryptKey string `hcl:"encrypt" json:"-"`
// ServerJoin contains information that is used to attempt to join servers
ServerJoin *ServerJoin `hcl:"server_join"`
// DefaultSchedulerConfig configures the initial scheduler config to be persisted in Raft.
// Once the cluster is bootstrapped, and Raft persists the config (from here or through API),
// This value is ignored.
DefaultSchedulerConfig *structs.SchedulerConfiguration `hcl:"default_scheduler_config"`
// PlanRejectionTracker configures the node plan rejection tracker that
// detects potentially bad nodes.
PlanRejectionTracker *PlanRejectionTracker `hcl:"plan_rejection_tracker"`
// EnableEventBroker configures whether this server's state store
2020-10-02 19:04:05 +00:00
// will generate events for its event stream.
EnableEventBroker *bool `hcl:"enable_event_broker"`
// EventBufferSize configure the amount of events to be held in memory.
// If EnableEventBroker is set to true, the minimum allowable value
// for the EventBufferSize is 1.
EventBufferSize *int `hcl:"event_buffer_size"`
2020-10-02 19:04:05 +00:00
// LicensePath is the path to search for an enterprise license.
LicensePath string `hcl:"license_path"`
// LicenseEnv is the full enterprise license. If NOMAD_LICENSE
// is set, LicenseEnv will be set to the value at startup.
LicenseEnv string
// licenseAdditionalPublicKeys is an internal-only field used to
// setup test licenses.
licenseAdditionalPublicKeys []string
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
// Search configures UI search features.
Search *Search `hcl:"search"`
2021-06-04 19:38:46 +00:00
// DeploymentQueryRateLimit is in queries per second and is used by the
// DeploymentWatcher to throttle the amount of simultaneously deployments
DeploymentQueryRateLimit float64 `hcl:"deploy_query_rate_limit"`
// RaftBoltConfig configures boltdb as used by raft.
RaftBoltConfig *RaftBoltConfig `hcl:"raft_boltdb"`
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (s *ServerConfig) Copy() *ServerConfig {
if s == nil {
return nil
}
ns := *s
ns.RaftMultiplier = pointer.Copy(s.RaftMultiplier)
ns.NumSchedulers = pointer.Copy(s.NumSchedulers)
ns.EnabledSchedulers = slices.Clone(s.EnabledSchedulers)
ns.StartJoin = slices.Clone(s.StartJoin)
ns.RetryJoin = slices.Clone(s.RetryJoin)
ns.ServerJoin = s.ServerJoin.Copy()
ns.DefaultSchedulerConfig = s.DefaultSchedulerConfig.Copy()
ns.PlanRejectionTracker = s.PlanRejectionTracker.Copy()
ns.EnableEventBroker = pointer.Copy(s.EnableEventBroker)
ns.EventBufferSize = pointer.Copy(s.EventBufferSize)
ns.licenseAdditionalPublicKeys = slices.Clone(s.licenseAdditionalPublicKeys)
ns.ExtraKeysHCL = slices.Clone(s.ExtraKeysHCL)
ns.Search = s.Search.Copy()
ns.RaftBoltConfig = s.RaftBoltConfig.Copy()
return &ns
}
// RaftBoltConfig is used in servers to configure parameters of the boltdb
// used for raft consensus.
type RaftBoltConfig struct {
// NoFreelistSync toggles whether the underlying raft storage should sync its
// freelist to disk within the bolt .db file. When disabled, IO performance
// will be improved but at the expense of longer startup times.
//
// Default: false.
NoFreelistSync bool `hcl:"no_freelist_sync"`
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (r *RaftBoltConfig) Copy() *RaftBoltConfig {
if r == nil {
return nil
}
nr := *r
return &nr
}
// PlanRejectionTracker is used in servers to configure the plan rejection
// tracker.
type PlanRejectionTracker struct {
// Enabled controls if the plan rejection tracker is active or not.
Enabled *bool `hcl:"enabled"`
// NodeThreshold is the number of times a node can have plan rejections
// before it is marked as ineligible.
NodeThreshold int `hcl:"node_threshold"`
// NodeWindow is the time window used to track active plan rejections for
// nodes.
NodeWindow time.Duration
NodeWindowHCL string `hcl:"node_window" json:"-"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (p *PlanRejectionTracker) Copy() *PlanRejectionTracker {
if p == nil {
return nil
}
np := *p
np.Enabled = pointer.Copy(p.Enabled)
np.ExtraKeysHCL = slices.Clone(p.ExtraKeysHCL)
return &np
}
func (p *PlanRejectionTracker) Merge(b *PlanRejectionTracker) *PlanRejectionTracker {
if p == nil {
return b
}
result := *p
if b == nil {
return &result
}
if b.Enabled != nil {
result.Enabled = b.Enabled
}
if b.NodeThreshold != 0 {
result.NodeThreshold = b.NodeThreshold
}
if b.NodeWindow != 0 {
result.NodeWindow = b.NodeWindow
}
if b.NodeWindowHCL != "" {
result.NodeWindowHCL = b.NodeWindowHCL
}
return &result
}
// Search is used in servers to configure search API options.
type Search struct {
// FuzzyEnabled toggles whether the FuzzySearch API is enabled. If not
// enabled, requests to /v1/search/fuzzy will reply with a 404 response code.
//
// Default: enabled.
FuzzyEnabled bool `hcl:"fuzzy_enabled"`
// LimitQuery limits the number of objects searched in the FuzzySearch API.
// The results are indicated as truncated if the limit is reached.
//
// Lowering this value can reduce resource consumption of Nomad server when
// the FuzzySearch API is enabled.
//
// Default value: 20.
LimitQuery int `hcl:"limit_query"`
// LimitResults limits the number of results provided by the FuzzySearch API.
// The results are indicated as truncate if the limit is reached.
//
// Lowering this value can reduce resource consumption of Nomad server per
// fuzzy search request when the FuzzySearch API is enabled.
//
// Default value: 100.
LimitResults int `hcl:"limit_results"`
// MinTermLength is the minimum length of Text required before the FuzzySearch
// API will return results.
//
// Increasing this value can avoid resource consumption on Nomad server by
// reducing searches with less meaningful results.
//
// Default value: 2.
MinTermLength int `hcl:"min_term_length"`
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (s *Search) Copy() *Search {
if s == nil {
return nil
}
ns := *s
return &ns
}
// ServerJoin is used in both clients and servers to bootstrap connections to
// servers
type ServerJoin struct {
// StartJoin is a list of addresses to attempt to join when the
// agent starts. If Serf is unable to communicate with any of these
// addresses, then the agent will error and exit.
StartJoin []string `hcl:"start_join"`
// RetryJoin is a list of addresses to join with retry enabled, or a single
// value to find multiple servers using go-discover syntax.
RetryJoin []string `hcl:"retry_join"`
// RetryMaxAttempts specifies the maximum number of times to retry joining a
// host on startup. This is useful for cases where we know the node will be
// online eventually.
RetryMaxAttempts int `hcl:"retry_max"`
// RetryInterval specifies the amount of time to wait in between join
// attempts on agent start. The minimum allowed value is 1 second and
// the default is 30s.
RetryInterval time.Duration
RetryIntervalHCL string `hcl:"retry_interval" json:"-"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (s *ServerJoin) Copy() *ServerJoin {
if s == nil {
return nil
}
ns := *s
ns.StartJoin = slices.Clone(s.StartJoin)
ns.RetryJoin = slices.Clone(s.RetryJoin)
ns.ExtraKeysHCL = slices.Clone(s.ExtraKeysHCL)
return &ns
}
func (s *ServerJoin) Merge(b *ServerJoin) *ServerJoin {
if s == nil {
return b
}
result := *s
if b == nil {
return &result
}
if len(b.StartJoin) != 0 {
result.StartJoin = b.StartJoin
}
if len(b.RetryJoin) != 0 {
result.RetryJoin = b.RetryJoin
}
if b.RetryMaxAttempts != 0 {
result.RetryMaxAttempts = b.RetryMaxAttempts
}
if b.RetryInterval != 0 {
result.RetryInterval = b.RetryInterval
}
return &result
}
// EncryptBytes returns the encryption key configured.
func (s *ServerConfig) EncryptBytes() ([]byte, error) {
return base64.StdEncoding.DecodeString(s.EncryptKey)
2015-08-16 01:54:41 +00:00
}
// Telemetry is the telemetry configuration for the server
type Telemetry struct {
StatsiteAddr string `hcl:"statsite_address"`
StatsdAddr string `hcl:"statsd_address"`
DataDogAddr string `hcl:"datadog_address"`
DataDogTags []string `hcl:"datadog_tags"`
PrometheusMetrics bool `hcl:"prometheus_metrics"`
DisableHostname bool `hcl:"disable_hostname"`
UseNodeName bool `hcl:"use_node_name"`
CollectionInterval string `hcl:"collection_interval"`
collectionInterval time.Duration `hcl:"-"`
PublishAllocationMetrics bool `hcl:"publish_allocation_metrics"`
PublishNodeMetrics bool `hcl:"publish_node_metrics"`
// PrefixFilter allows for filtering out metrics from being collected
PrefixFilter []string `hcl:"prefix_filter"`
// FilterDefault controls whether to allow metrics that have not been specified
// by the filter
FilterDefault *bool `hcl:"filter_default"`
2018-11-30 03:27:39 +00:00
// DisableDispatchedJobSummaryMetrics allows ignoring dispatched jobs when
// publishing Job summary metrics. This is useful in environments that produce
// high numbers of single count dispatch jobs as the metrics for each take up
// a small memory overhead.
DisableDispatchedJobSummaryMetrics bool `hcl:"disable_dispatched_job_summary_metrics"`
// Circonus: see https://github.com/circonus-labs/circonus-gometrics
// for more details on the various configuration options.
// Valid configuration combinations:
// - CirconusAPIToken
// metric management enabled (search for existing check or create a new one)
// - CirconusSubmissionUrl
// metric management disabled (use check with specified submission_url,
// broker must be using a public SSL certificate)
// - CirconusAPIToken + CirconusCheckSubmissionURL
// metric management enabled (use check with specified submission_url)
// - CirconusAPIToken + CirconusCheckID
// metric management enabled (use check with specified id)
// CirconusAPIToken is a valid API Token used to create/manage check. If provided,
// metric management is enabled.
// Default: none
CirconusAPIToken string `hcl:"circonus_api_token"`
// CirconusAPIApp is an app name associated with API token.
// Default: "nomad"
CirconusAPIApp string `hcl:"circonus_api_app"`
// CirconusAPIURL is the base URL to use for contacting the Circonus API.
// Default: "https://api.circonus.com/v2"
CirconusAPIURL string `hcl:"circonus_api_url"`
// CirconusSubmissionInterval is the interval at which metrics are submitted to Circonus.
// Default: 10s
CirconusSubmissionInterval string `hcl:"circonus_submission_interval"`
// CirconusCheckSubmissionURL is the check.config.submission_url field from a
// previously created HTTPTRAP check.
// Default: none
CirconusCheckSubmissionURL string `hcl:"circonus_submission_url"`
// CirconusCheckID is the check id (not check bundle id) from a previously created
// HTTPTRAP check. The numeric portion of the check._cid field.
// Default: none
CirconusCheckID string `hcl:"circonus_check_id"`
// CirconusCheckForceMetricActivation will force enabling metrics, as they are encountered,
// if the metric already exists and is NOT active. If check management is enabled, the default
2018-03-11 17:58:59 +00:00
// behavior is to add new metrics as they are encountered. If the metric already exists in the
// check, it will *NOT* be activated. This setting overrides that behavior.
// Default: "false"
CirconusCheckForceMetricActivation string `hcl:"circonus_check_force_metric_activation"`
// CirconusCheckInstanceID serves to uniquely identify the metrics coming from this "instance".
// It can be used to maintain metric continuity with transient or ephemeral instances as
// they move around within an infrastructure.
// Default: hostname:app
CirconusCheckInstanceID string `hcl:"circonus_check_instance_id"`
// CirconusCheckSearchTag is a special tag which, when coupled with the instance id, helps to
// narrow down the search results when neither a Submission URL or Check ID is provided.
// Default: service:app (e.g. service:nomad)
CirconusCheckSearchTag string `hcl:"circonus_check_search_tag"`
// CirconusCheckTags is a comma separated list of tags to apply to the check. Note that
// the value of CirconusCheckSearchTag will always be added to the check.
// Default: none
CirconusCheckTags string `hcl:"circonus_check_tags"`
// CirconusCheckDisplayName is the name for the check which will be displayed in the Circonus UI.
// Default: value of CirconusCheckInstanceID
CirconusCheckDisplayName string `hcl:"circonus_check_display_name"`
// CirconusBrokerID is an explicit broker to use when creating a new check. The numeric portion
// of broker._cid. If metric management is enabled and neither a Submission URL nor Check ID
// is provided, an attempt will be made to search for an existing check using Instance ID and
// Search Tag. If one is not found, a new HTTPTRAP check will be created.
// Default: use Select Tag if provided, otherwise, a random Enterprise Broker associated
// with the specified API token or the default Circonus Broker.
// Default: none
CirconusBrokerID string `hcl:"circonus_broker_id"`
// CirconusBrokerSelectTag is a special tag which will be used to select a broker when
// a Broker ID is not provided. The best use of this is to as a hint for which broker
// should be used based on *where* this particular instance is running.
// (e.g. a specific geo location or datacenter, dc:sfo)
// Default: none
CirconusBrokerSelectTag string `hcl:"circonus_broker_select_tag"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
2015-08-16 01:54:41 +00:00
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (t *Telemetry) Copy() *Telemetry {
if t == nil {
return nil
}
nt := *t
nt.DataDogTags = slices.Clone(t.DataDogTags)
nt.PrefixFilter = slices.Clone(t.PrefixFilter)
nt.FilterDefault = pointer.Copy(t.FilterDefault)
nt.ExtraKeysHCL = slices.Clone(t.ExtraKeysHCL)
return &nt
}
// PrefixFilters parses the PrefixFilter field and returns a list of allowed and blocked filters
func (a *Telemetry) PrefixFilters() (allowed, blocked []string, err error) {
for _, rule := range a.PrefixFilter {
if rule == "" {
continue
}
switch rule[0] {
case '+':
allowed = append(allowed, rule[1:])
case '-':
blocked = append(blocked, rule[1:])
default:
return nil, nil, fmt.Errorf("Filter rule must begin with either '+' or '-': %q", rule)
}
}
return allowed, blocked, nil
}
// Ports encapsulates the various ports we bind to for network services. If any
// are not specified then the defaults are used instead.
type Ports struct {
HTTP int `hcl:"http"`
RPC int `hcl:"rpc"`
Serf int `hcl:"serf"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (p *Ports) Copy() *Ports {
if p == nil {
return nil
}
np := *p
np.ExtraKeysHCL = slices.Clone(p.ExtraKeysHCL)
return &np
}
// Addresses encapsulates all of the addresses we bind to for various
// network services. Everything is optional and defaults to BindAddr.
type Addresses struct {
HTTP string `hcl:"http"`
RPC string `hcl:"rpc"`
Serf string `hcl:"serf"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (a *Addresses) Copy() *Addresses {
if a == nil {
return nil
}
na := *a
na.ExtraKeysHCL = slices.Clone(a.ExtraKeysHCL)
return &na
}
// AdvertiseAddrs is used to control the addresses we advertise out for
// different network services. All are optional and default to BindAddr and
// their default Port.
type NormalizedAddrs struct {
HTTP []string
RPC string
Serf string
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (n *NormalizedAddrs) Copy() *NormalizedAddrs {
if n == nil {
return nil
}
nn := *n
nn.HTTP = slices.Clone(n.HTTP)
return &nn
}
// AdvertiseAddrs is used to control the addresses we advertise out for
// different network services. All are optional and default to BindAddr and
// their default Port.
type AdvertiseAddrs struct {
HTTP string `hcl:"http"`
RPC string `hcl:"rpc"`
Serf string `hcl:"serf"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (a *AdvertiseAddrs) Copy() *AdvertiseAddrs {
if a == nil {
return nil
}
na := *a
na.ExtraKeysHCL = slices.Clone(a.ExtraKeysHCL)
return &na
}
type Resources struct {
CPU int `hcl:"cpu"`
MemoryMB int `hcl:"memory"`
DiskMB int `hcl:"disk"`
ReservedPorts string `hcl:"reserved_ports"`
Cores string `hcl:"cores"`
// ExtraKeysHCL is used by hcl to surface unexpected keys
ExtraKeysHCL []string `hcl:",unusedKeys" json:"-"`
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
func (r *Resources) Copy() *Resources {
if r == nil {
return nil
}
nr := *r
nr.ExtraKeysHCL = slices.Clone(r.ExtraKeysHCL)
return &nr
}
2019-08-30 12:15:50 +00:00
// devModeConfig holds the config for the -dev and -dev-connect flags
type devModeConfig struct {
2019-08-30 12:15:50 +00:00
// mode flags are set at the command line via -dev and -dev-connect
defaultMode bool
connectMode bool
bindAddr string
iface string
}
// newDevModeConfig parses the optional string value of the -dev flag
2019-08-30 12:15:50 +00:00
func newDevModeConfig(devMode, connectMode bool) (*devModeConfig, error) {
if !devMode && !connectMode {
return nil, nil
}
mode := &devModeConfig{}
2019-08-30 12:15:50 +00:00
mode.defaultMode = devMode
if connectMode {
if runtime.GOOS != "linux" {
// strictly speaking -dev-connect only binds to the
// non-localhost interface, but given its purpose
// is to support a feature with network namespaces
// we'll return an error here rather than let the agent
// come up and fail unexpectedly to run jobs
return nil, fmt.Errorf("-dev-connect is only supported on linux.")
}
u, err := users.Current()
2019-08-30 12:15:50 +00:00
if err != nil {
return nil, fmt.Errorf(
"-dev-connect uses network namespaces and is only supported for root: %v", err)
}
if u.Uid != "0" {
return nil, fmt.Errorf(
"-dev-connect uses network namespaces and is only supported for root.")
}
// Ensure Consul is on PATH
if _, err := exec.LookPath("consul"); err != nil {
return nil, fmt.Errorf("-dev-connect requires a 'consul' binary in Nomad's $PATH")
}
2019-08-30 12:15:50 +00:00
mode.connectMode = true
}
err := mode.networkConfig()
if err != nil {
return nil, err
}
return mode, nil
}
func (mode *devModeConfig) networkConfig() error {
if runtime.GOOS == "windows" {
mode.bindAddr = "127.0.0.1"
mode.iface = "Loopback Pseudo-Interface 1"
return nil
}
if runtime.GOOS == "darwin" {
mode.bindAddr = "127.0.0.1"
mode.iface = "lo0"
return nil
}
if mode != nil && mode.connectMode {
// if we hit either of the errors here we're in a weird situation
// where syscalls to get the list of network interfaces are failing.
// rather than throwing errors, we'll fall back to the default.
ifAddrs, err := sockaddr.GetDefaultInterfaces()
errMsg := "-dev=connect uses network namespaces: %v"
if err != nil {
return fmt.Errorf(errMsg, err)
}
if len(ifAddrs) < 1 {
return fmt.Errorf(errMsg, "could not find public network interface")
}
iface := ifAddrs[0].Name
mode.iface = iface
mode.bindAddr = "0.0.0.0" // allows CLI to "just work"
return nil
}
mode.bindAddr = "127.0.0.1"
mode.iface = "lo"
return nil
}
2015-08-16 01:54:41 +00:00
// DevConfig is a Config that is used for dev mode of Nomad.
func DevConfig(mode *devModeConfig) *Config {
if mode == nil {
mode = &devModeConfig{defaultMode: true}
mode.networkConfig()
}
2015-09-06 00:06:05 +00:00
conf := DefaultConfig()
conf.BindAddr = mode.bindAddr
2015-09-06 00:06:05 +00:00
conf.LogLevel = "DEBUG"
conf.Client.Enabled = true
conf.Server.Enabled = true
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
conf.DevMode = true
conf.Server.BootstrapExpect = 1
2015-09-06 00:06:05 +00:00
conf.EnableDebug = true
conf.DisableAnonymousSignature = true
conf.Consul.AutoAdvertise = pointer.Of(true)
conf.Client.NetworkInterface = mode.iface
conf.Client.Options = map[string]string{
"driver.raw_exec.enable": "true",
"driver.docker.volumes": "true",
}
2017-01-31 23:32:20 +00:00
conf.Client.GCInterval = 10 * time.Minute
conf.Client.GCDiskUsageThreshold = 99
conf.Client.GCInodeUsageThreshold = 99
2017-05-12 22:57:27 +00:00
conf.Client.GCMaxAllocs = 50
conf.Client.TemplateConfig = &client.ClientTemplateConfig{
FunctionDenylist: client.DefaultTemplateFunctionDenylist,
DisableSandbox: false,
}
conf.Client.Options[fingerprint.TightenNetworkTimeoutsConfig] = "true"
conf.Client.BindWildcardDefaultHostNetwork = true
conf.Client.NomadServiceDiscovery = pointer.Of(true)
conf.Telemetry.PrometheusMetrics = true
conf.Telemetry.PublishAllocationMetrics = true
conf.Telemetry.PublishNodeMetrics = true
2015-10-29 12:30:26 +00:00
2015-09-06 00:06:05 +00:00
return conf
2015-08-16 20:54:49 +00:00
}
// DefaultConfig is the baseline configuration for Nomad.
2015-08-16 20:54:49 +00:00
func DefaultConfig() *Config {
return &Config{
LogLevel: "INFO",
Region: "global",
Datacenter: "dc1",
BindAddr: "0.0.0.0",
Ports: &Ports{
HTTP: 4646,
RPC: 4647,
Serf: 4648,
},
Addresses: &Addresses{},
AdvertiseAddrs: &AdvertiseAddrs{},
2016-06-17 03:41:05 +00:00
Consul: config.DefaultConsulConfig(),
2016-08-06 01:13:06 +00:00
Vault: config.DefaultVaultConfig(),
UI: config.DefaultUIConfig(),
2015-09-06 00:06:05 +00:00
Client: &ClientConfig{
2017-09-05 17:30:51 +00:00
Enabled: false,
MaxKillTimeout: "30s",
ClientMinPort: 14000,
ClientMaxPort: 14512,
MinDynamicPort: 20000,
MaxDynamicPort: 32000,
2017-09-05 17:30:51 +00:00
Reserved: &Resources{},
GCInterval: 1 * time.Minute,
GCParallelDestroys: 2,
GCDiskUsageThreshold: 80,
GCInodeUsageThreshold: 70,
GCMaxAllocs: 50,
NoHostUUID: pointer.Of(true),
DisableRemoteExec: false,
ServerJoin: &ServerJoin{
RetryJoin: []string{},
RetryInterval: 30 * time.Second,
RetryMaxAttempts: 0,
},
TemplateConfig: &client.ClientTemplateConfig{
FunctionDenylist: client.DefaultTemplateFunctionDenylist,
DisableSandbox: false,
},
BindWildcardDefaultHostNetwork: true,
CNIPath: "/opt/cni/bin",
CNIConfigDir: "/opt/cni/config",
NomadServiceDiscovery: pointer.Of(true),
Artifact: config.DefaultArtifactConfig(),
2015-09-06 00:06:05 +00:00
},
Server: &ServerConfig{
Enabled: false,
EnableEventBroker: pointer.Of(true),
EventBufferSize: pointer.Of(100),
RaftProtocol: 3,
StartJoin: []string{},
PlanRejectionTracker: &PlanRejectionTracker{
Enabled: pointer.Of(false),
NodeThreshold: 100,
NodeWindow: 5 * time.Minute,
},
ServerJoin: &ServerJoin{
RetryJoin: []string{},
RetryInterval: 30 * time.Second,
RetryMaxAttempts: 0,
},
Search: &Search{
FuzzyEnabled: true,
LimitQuery: 20,
LimitResults: 100,
MinTermLength: 2,
},
2015-09-06 00:06:05 +00:00
},
2017-08-13 20:46:47 +00:00
ACL: &ACLConfig{
Enabled: false,
2017-08-19 22:30:01 +00:00
TokenTTL: 30 * time.Second,
PolicyTTL: 30 * time.Second,
RoleTTL: 30 * time.Second,
2017-08-13 20:46:47 +00:00
},
SyslogFacility: "LOCAL0",
Telemetry: &Telemetry{
CollectionInterval: "1s",
collectionInterval: 1 * time.Second,
},
TLSConfig: &config.TLSConfig{},
Sentinel: &config.SentinelConfig{},
Version: version.GetVersion(),
Autopilot: config.DefaultAutopilotConfig(),
Audit: &config.AuditConfig{},
DisableUpdateCheck: pointer.Of(false),
Limits: config.DefaultLimits(),
2015-08-16 20:54:49 +00:00
}
2015-08-16 01:54:41 +00:00
}
2015-12-09 00:05:15 +00:00
// Listener can be used to get a new listener using a custom bind address.
// If the bind provided address is empty, the BindAddr is used instead.
func (c *Config) Listener(proto, addr string, port int) (net.Listener, error) {
if addr == "" {
addr = c.BindAddr
}
// Do our own range check to avoid bugs in package net.
//
// golang.org/issue/11715
// golang.org/issue/13447
//
// Both of the above bugs were fixed by golang.org/cl/12447 which will be
// included in Go 1.6. The error returned below is the same as what Go 1.6
// will return.
if 0 > port || port > 65535 {
return nil, &net.OpError{
Op: "listen",
Net: proto,
Err: &net.AddrError{Err: "invalid port", Addr: fmt.Sprint(port)},
}
}
return net.Listen(proto, net.JoinHostPort(addr, strconv.Itoa(port)))
}
2015-08-16 01:54:41 +00:00
// Merge merges two configurations.
2015-12-09 00:05:15 +00:00
func (c *Config) Merge(b *Config) *Config {
result := *c
2015-08-16 01:54:41 +00:00
2015-09-10 04:42:50 +00:00
if b.Region != "" {
result.Region = b.Region
2015-08-16 01:54:41 +00:00
}
2015-09-10 04:42:50 +00:00
if b.Datacenter != "" {
result.Datacenter = b.Datacenter
}
if b.NodeName != "" {
result.NodeName = b.NodeName
}
if b.DataDir != "" {
result.DataDir = b.DataDir
}
2018-08-30 20:43:09 +00:00
if b.PluginDir != "" {
result.PluginDir = b.PluginDir
}
2015-09-10 04:42:50 +00:00
if b.LogLevel != "" {
result.LogLevel = b.LogLevel
}
2019-01-11 19:36:26 +00:00
if b.LogJson {
result.LogJson = true
}
if b.LogFile != "" {
result.LogFile = b.LogFile
}
if b.LogRotateDuration != "" {
result.LogRotateDuration = b.LogRotateDuration
}
if b.LogRotateBytes != 0 {
result.LogRotateBytes = b.LogRotateBytes
}
if b.LogRotateMaxFiles != 0 {
result.LogRotateMaxFiles = b.LogRotateMaxFiles
}
if b.BindAddr != "" {
result.BindAddr = b.BindAddr
}
2015-09-10 04:42:50 +00:00
if b.EnableDebug {
result.EnableDebug = true
}
if b.LeaveOnInt {
result.LeaveOnInt = true
}
if b.LeaveOnTerm {
result.LeaveOnTerm = true
}
if b.EnableSyslog {
result.EnableSyslog = true
}
if b.SyslogFacility != "" {
result.SyslogFacility = b.SyslogFacility
}
if b.DisableUpdateCheck != nil {
result.DisableUpdateCheck = pointer.Of(*b.DisableUpdateCheck)
2015-09-10 04:42:50 +00:00
}
if b.DisableAnonymousSignature {
result.DisableAnonymousSignature = true
}
// Apply the telemetry config
if result.Telemetry == nil && b.Telemetry != nil {
telemetry := *b.Telemetry
result.Telemetry = &telemetry
} else if b.Telemetry != nil {
result.Telemetry = result.Telemetry.Merge(b.Telemetry)
}
2016-10-24 20:46:22 +00:00
// Apply the TLS Config
if result.TLSConfig == nil && b.TLSConfig != nil {
result.TLSConfig = b.TLSConfig.Copy()
2016-10-24 20:46:22 +00:00
} else if b.TLSConfig != nil {
result.TLSConfig = result.TLSConfig.Merge(b.TLSConfig)
}
// Apply the client config
if result.Client == nil && b.Client != nil {
client := *b.Client
result.Client = &client
} else if b.Client != nil {
result.Client = result.Client.Merge(b.Client)
}
// Apply the server config
if result.Server == nil && b.Server != nil {
server := *b.Server
result.Server = &server
} else if b.Server != nil {
result.Server = result.Server.Merge(b.Server)
}
2017-08-13 20:46:47 +00:00
// Apply the acl config
if result.ACL == nil && b.ACL != nil {
server := *b.ACL
result.ACL = &server
} else if b.ACL != nil {
result.ACL = result.ACL.Merge(b.ACL)
}
// Apply the Audit config
if result.Audit == nil && b.Audit != nil {
audit := *b.Audit
result.Audit = &audit
} else if b.ACL != nil {
result.Audit = result.Audit.Merge(b.Audit)
}
// Apply the ports config
if result.Ports == nil && b.Ports != nil {
ports := *b.Ports
result.Ports = &ports
} else if b.Ports != nil {
result.Ports = result.Ports.Merge(b.Ports)
}
// Apply the address config
if result.Addresses == nil && b.Addresses != nil {
addrs := *b.Addresses
result.Addresses = &addrs
} else if b.Addresses != nil {
result.Addresses = result.Addresses.Merge(b.Addresses)
}
// Apply the advertise addrs config
if result.AdvertiseAddrs == nil && b.AdvertiseAddrs != nil {
advertise := *b.AdvertiseAddrs
result.AdvertiseAddrs = &advertise
} else if b.AdvertiseAddrs != nil {
result.AdvertiseAddrs = result.AdvertiseAddrs.Merge(b.AdvertiseAddrs)
}
// Apply the Consul Configuration
if result.Consul == nil && b.Consul != nil {
result.Consul = b.Consul.Copy()
} else if b.Consul != nil {
result.Consul = result.Consul.Merge(b.Consul)
}
2016-08-06 01:13:06 +00:00
// Apply the Vault Configuration
if result.Vault == nil && b.Vault != nil {
vaultConfig := *b.Vault
result.Vault = &vaultConfig
} else if b.Vault != nil {
result.Vault = result.Vault.Merge(b.Vault)
}
// Apply the UI Configuration
if result.UI == nil && b.UI != nil {
uiConfig := *b.UI
result.UI = &uiConfig
} else if b.UI != nil {
result.UI = result.UI.Merge(b.UI)
}
2017-09-19 14:47:10 +00:00
// Apply the sentinel config
if result.Sentinel == nil && b.Sentinel != nil {
server := *b.Sentinel
result.Sentinel = &server
} else if b.Sentinel != nil {
result.Sentinel = result.Sentinel.Merge(b.Sentinel)
}
if result.Autopilot == nil && b.Autopilot != nil {
autopilot := *b.Autopilot
result.Autopilot = &autopilot
} else if b.Autopilot != nil {
result.Autopilot = result.Autopilot.Merge(b.Autopilot)
}
2018-08-29 23:26:44 +00:00
if len(result.Plugins) == 0 && len(b.Plugins) != 0 {
copy := make([]*config.PluginConfig, len(b.Plugins))
for i, v := range b.Plugins {
copy[i] = v.Copy()
}
result.Plugins = copy
} else if len(b.Plugins) != 0 {
result.Plugins = config.PluginConfigSetMerge(result.Plugins, b.Plugins)
}
// Merge config files lists
result.Files = append(result.Files, b.Files...)
// Add the http API response header map values
if result.HTTPAPIResponseHeaders == nil {
result.HTTPAPIResponseHeaders = make(map[string]string)
}
for k, v := range b.HTTPAPIResponseHeaders {
result.HTTPAPIResponseHeaders[k] = v
}
result.Limits = c.Limits.Merge(b.Limits)
return &result
}
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
// Copy returns a deep copy safe for mutation.
func (c *Config) Copy() *Config {
if c == nil {
return nil
}
nc := *c
nc.Ports = c.Ports.Copy()
nc.Addresses = c.Addresses.Copy()
nc.normalizedAddrs = c.normalizedAddrs.Copy()
nc.AdvertiseAddrs = c.AdvertiseAddrs.Copy()
nc.Client = c.Client.Copy()
nc.Server = c.Server.Copy()
nc.ACL = c.ACL.Copy()
nc.Telemetry = c.Telemetry.Copy()
nc.DisableUpdateCheck = pointer.Copy(c.DisableUpdateCheck)
nc.Consul = c.Consul.Copy()
nc.Vault = c.Vault.Copy()
nc.UI = c.UI.Copy()
nc.NomadConfig = c.NomadConfig.Copy()
nc.ClientConfig = c.ClientConfig.Copy()
nc.Version = c.Version.Copy()
nc.Files = slices.Clone(c.Files)
nc.TLSConfig = c.TLSConfig.Copy()
nc.HTTPAPIResponseHeaders = maps.Clone(c.HTTPAPIResponseHeaders)
client: fix data races in config handling (#14139) Before this change, Client had 2 copies of the config object: config and configCopy. There was no guidance around which to use where (other than configCopy's comment to pass it to alloc runners), both are shared among goroutines and mutated in data racy ways. At least at one point I think the idea was to have `config` be mutable and then grab a lock to overwrite `configCopy`'s pointer atomically. This would have allowed alloc runners to read their config copies in data race safe ways, but this isn't how the current implementation worked. This change takes the following approach to safely handling configs in the client: 1. `Client.config` is the only copy of the config and all access must go through the `Client.configLock` mutex 2. Since the mutex *only protects the config pointer itself and not fields inside the Config struct:* all config mutation must be done on a *copy* of the config, and then Client's config pointer is overwritten while the mutex is acquired. Alloc runners and other goroutines with the old config pointer will not see config updates. 3. Deep copying is implemented on the Config struct to satisfy the previous approach. The TLS Keyloader is an exception because it has its own internal locking to support mutating in place. An unfortunate complication but one I couldn't find a way to untangle in a timely fashion. 4. To facilitate deep copying I made an *internally backward incompatible API change:* our `helper/funcs` used to turn containers (slices and maps) with 0 elements into nils. This probably saves a few memory allocations but makes it very easy to cause panics. Since my new config handling approach uses more copying, it became very difficult to ensure all code that used containers on configs could handle nils properly. Since this code has caused panics in the past, I fixed it: nil containers are copied as nil, but 0-element containers properly return a new 0-element container. No more "downgrading to nil!"
2022-08-18 23:32:04 +00:00
nc.Sentinel = c.Sentinel.Copy()
nc.Autopilot = c.Autopilot.Copy()
nc.Plugins = helper.CopySlice(c.Plugins)
nc.Limits = c.Limits.Copy()
nc.Audit = c.Audit.Copy()
nc.ExtraKeysHCL = slices.Clone(c.ExtraKeysHCL)
return &nc
}
2016-11-09 23:51:00 +00:00
// normalizeAddrs normalizes Addresses and AdvertiseAddrs to always be
// initialized and have reasonable defaults.
func (c *Config) normalizeAddrs() error {
if c.BindAddr != "" {
ipStr, err := listenerutil.ParseSingleIPTemplate(c.BindAddr)
if err != nil {
return fmt.Errorf("Bind address resolution failed: %v", err)
}
c.BindAddr = ipStr
}
httpAddrs, err := normalizeMultipleBind(c.Addresses.HTTP, c.BindAddr)
if err != nil {
return fmt.Errorf("Failed to parse HTTP address: %v", err)
}
c.Addresses.HTTP = strings.Join(httpAddrs, " ")
addr, err := normalizeBind(c.Addresses.RPC, c.BindAddr)
if err != nil {
return fmt.Errorf("Failed to parse RPC address: %v", err)
}
c.Addresses.RPC = addr
addr, err = normalizeBind(c.Addresses.Serf, c.BindAddr)
if err != nil {
return fmt.Errorf("Failed to parse Serf address: %v", err)
}
c.Addresses.Serf = addr
c.normalizedAddrs = &NormalizedAddrs{
HTTP: joinHostPorts(httpAddrs, strconv.Itoa(c.Ports.HTTP)),
RPC: net.JoinHostPort(c.Addresses.RPC, strconv.Itoa(c.Ports.RPC)),
Serf: net.JoinHostPort(c.Addresses.Serf, strconv.Itoa(c.Ports.Serf)),
}
addr, err = normalizeAdvertise(c.AdvertiseAddrs.HTTP, httpAddrs[0], c.Ports.HTTP, c.DevMode)
if err != nil {
2017-07-21 04:07:32 +00:00
return fmt.Errorf("Failed to parse HTTP advertise address (%v, %v, %v, %v): %v", c.AdvertiseAddrs.HTTP, c.Addresses.HTTP, c.Ports.HTTP, c.DevMode, err)
}
c.AdvertiseAddrs.HTTP = addr
addr, err = normalizeAdvertise(c.AdvertiseAddrs.RPC, c.Addresses.RPC, c.Ports.RPC, c.DevMode)
if err != nil {
return fmt.Errorf("Failed to parse RPC advertise address: %v", err)
}
c.AdvertiseAddrs.RPC = addr
// Skip serf if server is disabled
if c.Server != nil && c.Server.Enabled {
addr, err = normalizeAdvertise(c.AdvertiseAddrs.Serf, c.Addresses.Serf, c.Ports.Serf, c.DevMode)
if err != nil {
return fmt.Errorf("Failed to parse Serf advertise address: %v", err)
}
c.AdvertiseAddrs.Serf = addr
}
// Skip network_interface evaluation if not a client
if c.Client != nil && c.Client.Enabled && c.Client.NetworkInterface != "" {
parsed, err := parseSingleInterfaceTemplate(c.Client.NetworkInterface)
if err != nil {
return fmt.Errorf("Failed to parse network-interface: %v", err)
}
c.Client.NetworkInterface = parsed
}
return nil
}
// parseSingleInterfaceTemplate parses a go-sockaddr template and returns an
// error if it doesn't result in a single value.
func parseSingleInterfaceTemplate(tpl string) (string, error) {
out, err := template.Parse(tpl)
if err != nil {
// Typically something like:
// unable to parse template "{{printfl \"en50\"}}": template: sockaddr.Parse:1: function "printfl" not defined
return "", err
}
// Remove any extra empty space around the rendered result and check if the
// result is also not empty if the user provided a template.
out = strings.TrimSpace(out)
if tpl != "" && out == "" {
return "", fmt.Errorf("template %q evaluated to empty result", tpl)
}
// `template.Parse` returns a space-separated list of results, but on
// Windows network interfaces are allowed to have spaces, so there is no
// guaranteed separators that we can use to test if the template returned
// multiple interfaces.
// The test below checks if the template results to a single valid interface.
_, err = net.InterfaceByName(out)
if err != nil {
return "", fmt.Errorf("invalid interface name %q", out)
}
return out, nil
}
// parseMultipleIPTemplate is used as a helper function to parse out a multiple IP
// addresses from a config parameter.
func parseMultipleIPTemplate(ipTmpl string) ([]string, error) {
out, err := template.Parse(ipTmpl)
if err != nil {
return []string{}, fmt.Errorf("Unable to parse address template %q: %v", ipTmpl, err)
}
ips := strings.Split(out, " ")
if len(ips) == 0 {
return []string{}, errors.New("No addresses found, please configure one.")
}
return deduplicateAddrs(ips), nil
}
// normalizeBind returns a normalized bind address.
//
// If addr is set it is used, if not the default bind address is used.
func normalizeBind(addr, bind string) (string, error) {
if addr == "" {
return bind, nil
}
return listenerutil.ParseSingleIPTemplate(addr)
}
// normalizeMultipleBind returns normalized bind addresses.
//
// If addr is set it is used, if not the default bind address is used.
func normalizeMultipleBind(addr, bind string) ([]string, error) {
if addr == "" {
return []string{bind}, nil
}
return parseMultipleIPTemplate(addr)
}
// normalizeAdvertise returns a normalized advertise address.
//
// If addr is set, it is used and the default port is appended if no port is
// set.
//
// If addr is not set and bind is a valid address, the returned string is the
// bind+port.
//
// If addr is not set and bind is not a valid advertise address, the hostname
// is resolved and returned with the port.
//
// Loopback is only considered a valid advertise address in dev mode.
func normalizeAdvertise(addr string, bind string, defport int, dev bool) (string, error) {
addr, err := listenerutil.ParseSingleIPTemplate(addr)
if err != nil {
return "", fmt.Errorf("Error parsing advertise address template: %v", err)
}
if addr != "" {
// Default to using manually configured address
_, _, err = net.SplitHostPort(addr)
if err != nil {
if !isMissingPort(err) && !isTooManyColons(err) {
return "", fmt.Errorf("Error parsing advertise address %q: %v", addr, err)
}
// missing port, append the default
return net.JoinHostPort(addr, strconv.Itoa(defport)), nil
}
return addr, nil
}
// Fallback to bind address first, and then try resolving the local hostname
ips, err := net.LookupIP(bind)
if err != nil {
return "", fmt.Errorf("Error resolving bind address %q: %v", bind, err)
}
// Return the first non-localhost unicast address
for _, ip := range ips {
if ip.IsLinkLocalUnicast() || ip.IsGlobalUnicast() {
return net.JoinHostPort(ip.String(), strconv.Itoa(defport)), nil
}
if ip.IsLoopback() {
if dev {
// loopback is fine for dev mode
return net.JoinHostPort(ip.String(), strconv.Itoa(defport)), nil
}
return "", fmt.Errorf("Defaulting advertise to localhost is unsafe, please set advertise manually")
}
}
// Bind is not localhost but not a valid advertise IP, use first private IP
addr, err = listenerutil.ParseSingleIPTemplate("{{ GetPrivateIP }}")
if err != nil {
return "", fmt.Errorf("Unable to parse default advertise address: %v", err)
}
return net.JoinHostPort(addr, strconv.Itoa(defport)), nil
}
// isMissingPort returns true if an error is a "missing port" error from
// net.SplitHostPort.
func isMissingPort(err error) bool {
// matches error const in net/ipsock.go
const missingPort = "missing port in address"
2017-02-16 21:10:24 +00:00
return err != nil && strings.Contains(err.Error(), missingPort)
}
// isTooManyColons returns true if an error is a "too many colons" error from
// net.SplitHostPort.
func isTooManyColons(err error) bool {
// matches error const in net/ipsock.go
const tooManyColons = "too many colons in address"
return err != nil && strings.Contains(err.Error(), tooManyColons)
}
2017-08-22 00:09:09 +00:00
// Merge is used to merge two ACL configs together. The settings from the input always take precedence.
2017-08-13 20:46:47 +00:00
func (a *ACLConfig) Merge(b *ACLConfig) *ACLConfig {
result := *a
if b.Enabled {
result.Enabled = true
}
2017-08-19 22:30:01 +00:00
if b.TokenTTL != 0 {
2017-08-13 20:46:47 +00:00
result.TokenTTL = b.TokenTTL
}
if b.TokenTTLHCL != "" {
result.TokenTTLHCL = b.TokenTTLHCL
}
2017-08-19 22:30:01 +00:00
if b.PolicyTTL != 0 {
2017-08-13 20:46:47 +00:00
result.PolicyTTL = b.PolicyTTL
}
if b.PolicyTTLHCL != "" {
result.PolicyTTLHCL = b.PolicyTTLHCL
}
if b.RoleTTL != 0 {
result.RoleTTL = b.RoleTTL
}
if b.RoleTTLHCL != "" {
result.RoleTTLHCL = b.RoleTTLHCL
}
if b.TokenMinExpirationTTL != 0 {
result.TokenMinExpirationTTL = b.TokenMinExpirationTTL
}
if b.TokenMinExpirationTTLHCL != "" {
result.TokenMinExpirationTTLHCL = b.TokenMinExpirationTTLHCL
}
if b.TokenMaxExpirationTTL != 0 {
result.TokenMaxExpirationTTL = b.TokenMaxExpirationTTL
}
if b.TokenMaxExpirationTTLHCL != "" {
result.TokenMaxExpirationTTLHCL = b.TokenMaxExpirationTTLHCL
}
if b.ReplicationToken != "" {
result.ReplicationToken = b.ReplicationToken
}
2017-08-13 20:46:47 +00:00
return &result
}
// Merge is used to merge two server configs together
func (s *ServerConfig) Merge(b *ServerConfig) *ServerConfig {
result := *s
if b.Enabled {
result.Enabled = true
}
2017-08-13 20:46:47 +00:00
if b.AuthoritativeRegion != "" {
result.AuthoritativeRegion = b.AuthoritativeRegion
}
if b.BootstrapExpect > 0 {
result.BootstrapExpect = b.BootstrapExpect
}
if b.DataDir != "" {
result.DataDir = b.DataDir
}
if b.ProtocolVersion != 0 {
result.ProtocolVersion = b.ProtocolVersion
}
if b.RaftProtocol != 0 {
result.RaftProtocol = b.RaftProtocol
}
2020-05-30 13:27:33 +00:00
if b.RaftMultiplier != nil {
c := *b.RaftMultiplier
result.RaftMultiplier = &c
}
2018-05-31 20:11:11 +00:00
if b.NumSchedulers != nil {
result.NumSchedulers = pointer.Of(*b.NumSchedulers)
}
if b.NodeGCThreshold != "" {
result.NodeGCThreshold = b.NodeGCThreshold
}
if b.JobGCInterval != "" {
result.JobGCInterval = b.JobGCInterval
}
if b.JobGCThreshold != "" {
result.JobGCThreshold = b.JobGCThreshold
}
if b.EvalGCThreshold != "" {
result.EvalGCThreshold = b.EvalGCThreshold
}
2017-06-29 18:29:44 +00:00
if b.DeploymentGCThreshold != "" {
result.DeploymentGCThreshold = b.DeploymentGCThreshold
}
if b.CSIVolumeClaimGCThreshold != "" {
result.CSIVolumeClaimGCThreshold = b.CSIVolumeClaimGCThreshold
}
if b.CSIPluginGCThreshold != "" {
result.CSIPluginGCThreshold = b.CSIPluginGCThreshold
}
if b.ACLTokenGCThreshold != "" {
result.ACLTokenGCThreshold = b.ACLTokenGCThreshold
}
if b.RootKeyGCInterval != "" {
result.RootKeyGCInterval = b.RootKeyGCInterval
}
if b.RootKeyGCThreshold != "" {
result.RootKeyGCThreshold = b.RootKeyGCThreshold
}
if b.RootKeyRotationThreshold != "" {
result.RootKeyRotationThreshold = b.RootKeyRotationThreshold
}
if b.HeartbeatGrace != 0 {
2016-03-04 23:44:12 +00:00
result.HeartbeatGrace = b.HeartbeatGrace
}
if b.HeartbeatGraceHCL != "" {
result.HeartbeatGraceHCL = b.HeartbeatGraceHCL
}
if b.MinHeartbeatTTL != 0 {
result.MinHeartbeatTTL = b.MinHeartbeatTTL
}
if b.MinHeartbeatTTLHCL != "" {
result.MinHeartbeatTTLHCL = b.MinHeartbeatTTLHCL
}
if b.MaxHeartbeatsPerSecond != 0.0 {
result.MaxHeartbeatsPerSecond = b.MaxHeartbeatsPerSecond
}
if b.FailoverHeartbeatTTL != 0 {
result.FailoverHeartbeatTTL = b.FailoverHeartbeatTTL
}
if b.FailoverHeartbeatTTLHCL != "" {
result.FailoverHeartbeatTTLHCL = b.FailoverHeartbeatTTLHCL
}
if b.RetryMaxAttempts != 0 {
result.RetryMaxAttempts = b.RetryMaxAttempts
}
if b.RetryInterval != 0 {
result.RetryInterval = b.RetryInterval
}
if b.RetryIntervalHCL != "" {
result.RetryIntervalHCL = b.RetryIntervalHCL
}
if b.RejoinAfterLeave {
result.RejoinAfterLeave = true
}
if b.NonVotingServer {
result.NonVotingServer = true
}
if b.RedundancyZone != "" {
result.RedundancyZone = b.RedundancyZone
}
if b.UpgradeVersion != "" {
result.UpgradeVersion = b.UpgradeVersion
}
if b.EncryptKey != "" {
result.EncryptKey = b.EncryptKey
}
if b.ServerJoin != nil {
result.ServerJoin = result.ServerJoin.Merge(b.ServerJoin)
}
if b.LicensePath != "" {
result.LicensePath = b.LicensePath
}
if b.EnableEventBroker != nil {
result.EnableEventBroker = b.EnableEventBroker
}
if b.EventBufferSize != nil {
result.EventBufferSize = b.EventBufferSize
2020-10-02 19:04:05 +00:00
}
if b.PlanRejectionTracker != nil {
result.PlanRejectionTracker = result.PlanRejectionTracker.Merge(b.PlanRejectionTracker)
}
if b.DefaultSchedulerConfig != nil {
c := *b.DefaultSchedulerConfig
result.DefaultSchedulerConfig = &c
}
2021-06-04 19:38:46 +00:00
if b.DeploymentQueryRateLimit != 0 {
result.DeploymentQueryRateLimit = b.DeploymentQueryRateLimit
}
if b.Search != nil {
result.Search = &Search{FuzzyEnabled: b.Search.FuzzyEnabled}
if b.Search.LimitQuery > 0 {
result.Search.LimitQuery = b.Search.LimitQuery
}
if b.Search.LimitResults > 0 {
result.Search.LimitResults = b.Search.LimitResults
}
if b.Search.MinTermLength > 0 {
result.Search.MinTermLength = b.Search.MinTermLength
}
}
if b.RaftBoltConfig != nil {
result.RaftBoltConfig = &RaftBoltConfig{
NoFreelistSync: b.RaftBoltConfig.NoFreelistSync,
}
}
// Add the schedulers
result.EnabledSchedulers = append(result.EnabledSchedulers, b.EnabledSchedulers...)
// Copy the start join addresses
result.StartJoin = make([]string, 0, len(s.StartJoin)+len(b.StartJoin))
result.StartJoin = append(result.StartJoin, s.StartJoin...)
result.StartJoin = append(result.StartJoin, b.StartJoin...)
// Copy the retry join addresses
result.RetryJoin = make([]string, 0, len(s.RetryJoin)+len(b.RetryJoin))
result.RetryJoin = append(result.RetryJoin, s.RetryJoin...)
result.RetryJoin = append(result.RetryJoin, b.RetryJoin...)
return &result
}
// Merge is used to merge two client configs together
func (a *ClientConfig) Merge(b *ClientConfig) *ClientConfig {
2015-12-09 00:05:15 +00:00
result := *a
if b.Enabled {
result.Enabled = true
}
if b.StateDir != "" {
result.StateDir = b.StateDir
}
if b.AllocDir != "" {
result.AllocDir = b.AllocDir
}
if b.NodeClass != "" {
result.NodeClass = b.NodeClass
}
if b.NetworkInterface != "" {
result.NetworkInterface = b.NetworkInterface
}
2015-10-03 00:32:11 +00:00
if b.NetworkSpeed != 0 {
result.NetworkSpeed = b.NetworkSpeed
}
if b.CpuCompute != 0 {
result.CpuCompute = b.CpuCompute
}
if b.MemoryMB != 0 {
result.MemoryMB = b.MemoryMB
}
if b.MaxKillTimeout != "" {
result.MaxKillTimeout = b.MaxKillTimeout
}
if b.ClientMaxPort != 0 {
result.ClientMaxPort = b.ClientMaxPort
}
if b.ClientMinPort != 0 {
result.ClientMinPort = b.ClientMinPort
}
if b.MaxDynamicPort != 0 {
result.MaxDynamicPort = b.MaxDynamicPort
}
if b.MinDynamicPort != 0 {
result.MinDynamicPort = b.MinDynamicPort
}
2017-02-03 16:39:57 +00:00
if result.Reserved == nil && b.Reserved != nil {
reserved := *b.Reserved
result.Reserved = &reserved
} else if b.Reserved != nil {
result.Reserved = result.Reserved.Merge(b.Reserved)
}
if b.ReserveableCores != "" {
result.ReserveableCores = b.ReserveableCores
}
2017-02-01 23:58:21 +00:00
if b.GCInterval != 0 {
result.GCInterval = b.GCInterval
}
if b.GCIntervalHCL != "" {
result.GCIntervalHCL = b.GCIntervalHCL
}
if b.GCParallelDestroys != 0 {
result.GCParallelDestroys = b.GCParallelDestroys
}
2017-02-01 23:58:21 +00:00
if b.GCDiskUsageThreshold != 0 {
result.GCDiskUsageThreshold = b.GCDiskUsageThreshold
}
if b.GCInodeUsageThreshold != 0 {
result.GCInodeUsageThreshold = b.GCInodeUsageThreshold
}
if b.GCMaxAllocs != 0 {
result.GCMaxAllocs = b.GCMaxAllocs
}
// NoHostUUID defaults to true, merge if false
if b.NoHostUUID != nil {
2017-02-27 21:42:37 +00:00
result.NoHostUUID = b.NoHostUUID
}
if b.DisableRemoteExec {
result.DisableRemoteExec = b.DisableRemoteExec
}
if b.TemplateConfig != nil {
result.TemplateConfig = b.TemplateConfig
}
// Add the servers
result.Servers = append(result.Servers, b.Servers...)
// Add the options map values
if result.Options == nil {
result.Options = make(map[string]string)
}
for k, v := range b.Options {
result.Options[k] = v
}
// Add the meta map values
if result.Meta == nil {
result.Meta = make(map[string]string)
}
for k, v := range b.Meta {
result.Meta[k] = v
}
// Add the chroot_env map values
if result.ChrootEnv == nil {
result.ChrootEnv = make(map[string]string)
}
for k, v := range b.ChrootEnv {
result.ChrootEnv[k] = v
}
if b.ServerJoin != nil {
result.ServerJoin = result.ServerJoin.Merge(b.ServerJoin)
}
if len(a.HostVolumes) == 0 && len(b.HostVolumes) != 0 {
result.HostVolumes = structs.CopySliceClientHostVolumeConfig(b.HostVolumes)
} else if len(b.HostVolumes) != 0 {
result.HostVolumes = structs.HostVolumeSliceMerge(a.HostVolumes, b.HostVolumes)
}
if b.CNIPath != "" {
result.CNIPath = b.CNIPath
}
if b.CNIConfigDir != "" {
result.CNIConfigDir = b.CNIConfigDir
}
if b.BridgeNetworkName != "" {
result.BridgeNetworkName = b.BridgeNetworkName
}
if b.BridgeNetworkSubnet != "" {
result.BridgeNetworkSubnet = b.BridgeNetworkSubnet
}
result.HostNetworks = a.HostNetworks
2020-06-16 15:53:10 +00:00
if len(b.HostNetworks) != 0 {
result.HostNetworks = append(result.HostNetworks, b.HostNetworks...)
2020-06-16 15:53:10 +00:00
}
if b.BindWildcardDefaultHostNetwork {
result.BindWildcardDefaultHostNetwork = true
}
// This value is a pointer, therefore if it is not nil the user has
// supplied an override value.
if b.NomadServiceDiscovery != nil {
result.NomadServiceDiscovery = b.NomadServiceDiscovery
}
client: enable support for cgroups v2 This PR introduces support for using Nomad on systems with cgroups v2 [1] enabled as the cgroups controller mounted on /sys/fs/cgroups. Newer Linux distros like Ubuntu 21.10 are shipping with cgroups v2 only, causing problems for Nomad users. Nomad mostly "just works" with cgroups v2 due to the indirection via libcontainer, but not so for managing cpuset cgroups. Before, Nomad has been making use of a feature in v1 where a PID could be a member of more than one cgroup. In v2 this is no longer possible, and so the logic around computing cpuset values must be modified. When Nomad detects v2, it manages cpuset values in-process, rather than making use of cgroup heirarchy inheritence via shared/reserved parents. Nomad will only activate the v2 logic when it detects cgroups2 is mounted at /sys/fs/cgroups. This means on systems running in hybrid mode with cgroups2 mounted at /sys/fs/cgroups/unified (as is typical) Nomad will continue to use the v1 logic, and should operate as before. Systems that do not support cgroups v2 are also not affected. When v2 is activated, Nomad will create a parent called nomad.slice (unless otherwise configured in Client conifg), and create cgroups for tasks using naming convention <allocID>-<task>.scope. These follow the naming convention set by systemd and also used by Docker when cgroups v2 is detected. Client nodes now export a new fingerprint attribute, unique.cgroups.version which will be set to 'v1' or 'v2' to indicate the cgroups regime in use by Nomad. The new cpuset management strategy fixes #11705, where docker tasks that spawned processes on startup would "leak". In cgroups v2, the PIDs are started in the cgroup they will always live in, and thus the cause of the leak is eliminated. [1] https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html Closes #11289 Fixes #11705 #11773 #11933
2022-02-28 22:24:01 +00:00
if b.CgroupParent != "" {
result.CgroupParent = b.CgroupParent
}
result.Artifact = a.Artifact.Merge(b.Artifact)
return &result
}
// Merge is used to merge two telemetry configs together
func (a *Telemetry) Merge(b *Telemetry) *Telemetry {
2015-12-09 00:05:15 +00:00
result := *a
if b.StatsiteAddr != "" {
result.StatsiteAddr = b.StatsiteAddr
}
if b.StatsdAddr != "" {
result.StatsdAddr = b.StatsdAddr
}
2016-10-15 08:49:33 +00:00
if b.DataDogAddr != "" {
result.DataDogAddr = b.DataDogAddr
}
if b.DataDogTags != nil {
result.DataDogTags = b.DataDogTags
}
if b.PrometheusMetrics {
result.PrometheusMetrics = b.PrometheusMetrics
}
if b.DisableHostname {
result.DisableHostname = true
}
if b.UseNodeName {
result.UseNodeName = true
}
if b.CollectionInterval != "" {
result.CollectionInterval = b.CollectionInterval
}
if b.collectionInterval != 0 {
result.collectionInterval = b.collectionInterval
}
2016-08-31 18:19:00 +00:00
if b.PublishNodeMetrics {
result.PublishNodeMetrics = true
}
if b.PublishAllocationMetrics {
result.PublishAllocationMetrics = true
}
if b.CirconusAPIToken != "" {
result.CirconusAPIToken = b.CirconusAPIToken
}
if b.CirconusAPIApp != "" {
result.CirconusAPIApp = b.CirconusAPIApp
}
if b.CirconusAPIURL != "" {
result.CirconusAPIURL = b.CirconusAPIURL
}
if b.CirconusCheckSubmissionURL != "" {
result.CirconusCheckSubmissionURL = b.CirconusCheckSubmissionURL
}
if b.CirconusSubmissionInterval != "" {
result.CirconusSubmissionInterval = b.CirconusSubmissionInterval
}
if b.CirconusCheckID != "" {
result.CirconusCheckID = b.CirconusCheckID
}
if b.CirconusCheckForceMetricActivation != "" {
result.CirconusCheckForceMetricActivation = b.CirconusCheckForceMetricActivation
}
if b.CirconusCheckInstanceID != "" {
result.CirconusCheckInstanceID = b.CirconusCheckInstanceID
}
if b.CirconusCheckSearchTag != "" {
result.CirconusCheckSearchTag = b.CirconusCheckSearchTag
}
if b.CirconusCheckTags != "" {
result.CirconusCheckTags = b.CirconusCheckTags
}
if b.CirconusCheckDisplayName != "" {
result.CirconusCheckDisplayName = b.CirconusCheckDisplayName
}
if b.CirconusBrokerID != "" {
result.CirconusBrokerID = b.CirconusBrokerID
}
if b.CirconusBrokerSelectTag != "" {
result.CirconusBrokerSelectTag = b.CirconusBrokerSelectTag
}
if b.PrefixFilter != nil {
result.PrefixFilter = b.PrefixFilter
}
if b.FilterDefault != nil {
result.FilterDefault = b.FilterDefault
}
if b.DisableDispatchedJobSummaryMetrics {
result.DisableDispatchedJobSummaryMetrics = b.DisableDispatchedJobSummaryMetrics
}
2015-09-10 04:42:50 +00:00
return &result
2015-08-16 01:54:41 +00:00
}
// Merge is used to merge two port configurations.
func (a *Ports) Merge(b *Ports) *Ports {
2015-12-09 00:05:15 +00:00
result := *a
if b.HTTP != 0 {
result.HTTP = b.HTTP
}
if b.RPC != 0 {
result.RPC = b.RPC
}
if b.Serf != 0 {
result.Serf = b.Serf
}
return &result
}
// Merge is used to merge two address configs together.
func (a *Addresses) Merge(b *Addresses) *Addresses {
2015-12-09 00:05:15 +00:00
result := *a
if b.HTTP != "" {
result.HTTP = b.HTTP
}
if b.RPC != "" {
result.RPC = b.RPC
}
if b.Serf != "" {
result.Serf = b.Serf
}
return &result
}
// Merge merges two advertise addrs configs together.
func (a *AdvertiseAddrs) Merge(b *AdvertiseAddrs) *AdvertiseAddrs {
2015-12-09 00:05:15 +00:00
result := *a
if b.RPC != "" {
result.RPC = b.RPC
}
if b.Serf != "" {
result.Serf = b.Serf
}
if b.HTTP != "" {
result.HTTP = b.HTTP
}
return &result
}
func (r *Resources) Merge(b *Resources) *Resources {
result := *r
if b.CPU != 0 {
result.CPU = b.CPU
}
if b.MemoryMB != 0 {
result.MemoryMB = b.MemoryMB
}
if b.DiskMB != 0 {
result.DiskMB = b.DiskMB
}
if b.ReservedPorts != "" {
result.ReservedPorts = b.ReservedPorts
}
if b.Cores != "" {
result.Cores = b.Cores
}
return &result
}
// LoadConfig loads the configuration at the given path, regardless if its a file or
// directory. Called for each -config to build up the runtime config value. Do not apply any
// default values, defaults should be added once in DefaultConfig
2015-08-16 01:54:41 +00:00
func LoadConfig(path string) (*Config, error) {
fi, err := os.Stat(path)
if err != nil {
return nil, err
}
if fi.IsDir() {
return LoadConfigDir(path)
2015-08-16 01:54:41 +00:00
}
cleaned := filepath.Clean(path)
config, err := ParseConfigFile(cleaned)
2015-08-16 01:54:41 +00:00
if err != nil {
return nil, fmt.Errorf("Error loading %s: %s", cleaned, err)
}
config.Files = append(config.Files, cleaned)
return config, nil
2015-09-22 18:29:53 +00:00
}
2015-08-16 01:54:41 +00:00
// LoadConfigDir loads all the configurations in the given directory
// in alphabetical order.
func LoadConfigDir(dir string) (*Config, error) {
f, err := os.Open(dir)
if err != nil {
return nil, err
}
defer f.Close()
fi, err := f.Stat()
if err != nil {
return nil, err
}
if !fi.IsDir() {
return nil, fmt.Errorf(
"configuration path must be a directory: %s", dir)
2015-08-16 01:54:41 +00:00
}
var files []string
err = nil
for err != io.EOF {
var fis []os.FileInfo
fis, err = f.Readdir(128)
if err != nil && err != io.EOF {
return nil, err
}
for _, fi := range fis {
// Ignore directories
if fi.IsDir() {
continue
}
// Only care about files that are valid to load.
name := fi.Name()
skip := true
if strings.HasSuffix(name, ".hcl") {
skip = false
} else if strings.HasSuffix(name, ".json") {
skip = false
}
if skip || isTemporaryFile(name) {
continue
}
path := filepath.Join(dir, name)
files = append(files, path)
}
}
// Fast-path if we have no files
if len(files) == 0 {
return &Config{}, nil
}
2015-12-15 07:05:25 +00:00
sort.Strings(files)
2015-08-16 01:54:41 +00:00
var result *Config
for _, f := range files {
config, err := ParseConfigFile(f)
2015-08-16 01:54:41 +00:00
if err != nil {
return nil, fmt.Errorf("Error loading %s: %s", f, err)
}
config.Files = append(config.Files, f)
2015-08-16 01:54:41 +00:00
if result == nil {
result = config
} else {
result = result.Merge(config)
}
}
return result, nil
}
// joinHostPorts joins every addr in addrs with the specified port
func joinHostPorts(addrs []string, port string) []string {
localAddrs := make([]string, len(addrs))
for i, k := range addrs {
localAddrs[i] = net.JoinHostPort(k, port)
}
return localAddrs
}
2015-08-16 01:54:41 +00:00
// isTemporaryFile returns true or false depending on whether the
// provided file name is a temporary file for the following editors:
// emacs or vim.
func isTemporaryFile(name string) bool {
return strings.HasSuffix(name, "~") || // vim
strings.HasPrefix(name, ".#") || // emacs
(strings.HasPrefix(name, "#") && strings.HasSuffix(name, "#")) // emacs
}
func deduplicateAddrs(addrs []string) []string {
keys := make(map[string]bool)
list := []string{}
for _, entry := range addrs {
if _, value := keys[entry]; !value {
keys[entry] = true
list = append(list, entry)
}
}
return list
}