open-consul/agent/consul/util.go

440 lines
12 KiB
Go
Raw Normal View History

2013-12-12 19:07:14 +00:00
package consul
2013-12-19 22:18:55 +00:00
import (
2014-01-08 00:58:16 +00:00
"encoding/binary"
2013-12-31 23:44:17 +00:00
"fmt"
"net"
"runtime"
2013-12-19 22:37:54 +00:00
"strconv"
"github.com/hashicorp/consul/agent/metadata"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
"github.com/hashicorp/consul/agent/structs"
"github.com/hashicorp/go-version"
"github.com/hashicorp/serf/serf"
2013-12-19 22:18:55 +00:00
)
2013-12-31 23:44:17 +00:00
/*
* Contains an entry for each private block:
* 10.0.0.0/8
* 100.64.0.0/10
* 127.0.0.0/8
* 169.254.0.0/16
2013-12-31 23:44:17 +00:00
* 172.16.0.0/12
* 192.168.0.0/16
2013-12-31 23:44:17 +00:00
*/
var privateBlocks []*net.IPNet
func init() {
// Add each private block
privateBlocks = make([]*net.IPNet, 6)
2013-12-31 23:44:17 +00:00
_, block, err := net.ParseCIDR("10.0.0.0/8")
if err != nil {
panic(fmt.Sprintf("Bad cidr. Got %v", err))
}
privateBlocks[0] = block
_, block, err = net.ParseCIDR("100.64.0.0/10")
2013-12-31 23:44:17 +00:00
if err != nil {
panic(fmt.Sprintf("Bad cidr. Got %v", err))
}
privateBlocks[1] = block
_, block, err = net.ParseCIDR("127.0.0.0/8")
2013-12-31 23:44:17 +00:00
if err != nil {
panic(fmt.Sprintf("Bad cidr. Got %v", err))
}
privateBlocks[2] = block
_, block, err = net.ParseCIDR("169.254.0.0/16")
if err != nil {
panic(fmt.Sprintf("Bad cidr. Got %v", err))
}
privateBlocks[3] = block
_, block, err = net.ParseCIDR("172.16.0.0/12")
if err != nil {
panic(fmt.Sprintf("Bad cidr. Got %v", err))
}
privateBlocks[4] = block
_, block, err = net.ParseCIDR("192.168.0.0/16")
if err != nil {
panic(fmt.Sprintf("Bad cidr. Got %v", err))
}
privateBlocks[5] = block
2013-12-31 23:44:17 +00:00
}
// CanServersUnderstandProtocol checks to see if all the servers in the given
// list understand the given protocol version. If there are no servers in the
// list then this will return false.
func CanServersUnderstandProtocol(members []serf.Member, version uint8) (bool, error) {
numServers, numWhoGrok := 0, 0
for _, m := range members {
if m.Tags["role"] != "consul" {
continue
}
numServers++
2018-01-28 18:40:13 +00:00
vsnMin, err := strconv.Atoi(m.Tags["vsn_min"])
if err != nil {
return false, err
}
2018-01-28 18:40:13 +00:00
vsnMax, err := strconv.Atoi(m.Tags["vsn_max"])
if err != nil {
return false, err
}
v := int(version)
2018-01-28 18:40:13 +00:00
if (v >= vsnMin) && (v <= vsnMax) {
numWhoGrok++
}
}
return (numServers > 0) && (numWhoGrok == numServers), nil
}
// Returns if a member is a consul node. Returns a bool,
2015-02-19 22:45:47 +00:00
// and the datacenter.
func isConsulNode(m serf.Member) (bool, string) {
2014-01-30 21:13:29 +00:00
if m.Tags["role"] != "node" {
return false, ""
}
2014-01-30 21:13:29 +00:00
return true, m.Tags["dc"]
}
2013-12-31 23:44:17 +00:00
// Returns if the given IP is in a private block
2018-01-28 18:40:13 +00:00
func isPrivateIP(ipStr string) bool {
ip := net.ParseIP(ipStr)
2013-12-31 23:44:17 +00:00
for _, priv := range privateBlocks {
if priv.Contains(ip) {
return true
}
}
return false
}
// Returns addresses from interfaces that is up
func activeInterfaceAddresses() ([]net.Addr, error) {
var upAddrs []net.Addr
2015-09-02 10:24:44 +00:00
var loAddrs []net.Addr
interfaces, err := net.Interfaces()
if err != nil {
return nil, fmt.Errorf("Failed to get interfaces: %v", err)
}
for _, iface := range interfaces {
// Require interface to be up
if iface.Flags&net.FlagUp == 0 {
continue
}
addresses, err := iface.Addrs()
if err != nil {
return nil, fmt.Errorf("Failed to get interface addresses: %v", err)
}
2015-09-02 10:24:44 +00:00
if iface.Flags&net.FlagLoopback != 0 {
loAddrs = append(loAddrs, addresses...)
continue
}
upAddrs = append(upAddrs, addresses...)
}
2015-09-02 10:24:44 +00:00
if len(upAddrs) == 0 {
return loAddrs, nil
}
return upAddrs, nil
}
2013-12-31 23:44:17 +00:00
// GetPrivateIP is used to return the first private IP address
// associated with an interface on the machine
func GetPrivateIP() (net.IP, error) {
addresses, err := activeInterfaceAddresses()
2013-12-31 23:44:17 +00:00
if err != nil {
return nil, fmt.Errorf("Failed to get interface addresses: %v", err)
}
return getPrivateIP(addresses)
}
func getPrivateIP(addresses []net.Addr) (net.IP, error) {
var candidates []net.IP
2013-12-31 23:44:17 +00:00
// Find private IPv4 address
for _, rawAddr := range addresses {
var ip net.IP
switch addr := rawAddr.(type) {
case *net.IPAddr:
ip = addr.IP
case *net.IPNet:
ip = addr.IP
default:
2013-12-31 23:44:17 +00:00
continue
}
if ip.To4() == nil {
2013-12-31 23:44:17 +00:00
continue
}
if !isPrivateIP(ip.String()) {
2013-12-31 23:44:17 +00:00
continue
}
candidates = append(candidates, ip)
}
numIps := len(candidates)
switch numIps {
case 0:
return nil, fmt.Errorf("No private IP address found")
case 1:
return candidates[0], nil
default:
return nil, fmt.Errorf("Multiple private IPs found. Please configure one.")
2013-12-31 23:44:17 +00:00
}
2013-12-31 23:44:17 +00:00
}
2014-01-08 00:58:16 +00:00
// GetPublicIPv6 is used to return the first public IP address
// associated with an interface on the machine
func GetPublicIPv6() (net.IP, error) {
addresses, err := net.InterfaceAddrs()
if err != nil {
return nil, fmt.Errorf("Failed to get interface addresses: %v", err)
}
return getPublicIPv6(addresses)
}
func isUniqueLocalAddress(ip net.IP) bool {
return len(ip) == net.IPv6len && ip[0] == 0xfc && ip[1] == 0x00
}
func getPublicIPv6(addresses []net.Addr) (net.IP, error) {
var candidates []net.IP
// Find public IPv6 address
for _, rawAddr := range addresses {
var ip net.IP
switch addr := rawAddr.(type) {
case *net.IPAddr:
ip = addr.IP
case *net.IPNet:
ip = addr.IP
default:
continue
}
if ip.To4() != nil {
continue
}
if ip.IsLinkLocalUnicast() || isUniqueLocalAddress(ip) || ip.IsLoopback() {
continue
}
candidates = append(candidates, ip)
}
numIps := len(candidates)
switch numIps {
case 0:
return nil, fmt.Errorf("No public IPv6 address found")
case 1:
return candidates[0], nil
default:
return nil, fmt.Errorf("Multiple public IPv6 addresses found. Please configure one.")
}
}
2014-01-08 00:58:16 +00:00
// Converts bytes to an integer
func bytesToUint64(b []byte) uint64 {
return binary.BigEndian.Uint64(b)
}
// Converts a uint to a byte slice
func uint64ToBytes(u uint64) []byte {
buf := make([]byte, 8)
binary.BigEndian.PutUint64(buf, u)
return buf
}
// runtimeStats is used to return various runtime information
func runtimeStats() map[string]string {
return map[string]string{
"os": runtime.GOOS,
"arch": runtime.GOARCH,
"version": runtime.Version(),
"max_procs": strconv.FormatInt(int64(runtime.GOMAXPROCS(0)), 10),
"goroutines": strconv.FormatInt(int64(runtime.NumGoroutine()), 10),
"cpu_count": strconv.FormatInt(int64(runtime.NumCPU()), 10),
}
}
// checkServersProvider exists so that we can unit tests the requirements checking functions
// without having to spin up a whole agent/server.
type checkServersProvider interface {
CheckServers(datacenter string, fn func(*metadata.Server) bool)
}
// serverRequirementsFn should inspect the given metadata.Server struct
// and return two booleans. The first indicates whether the given requirements
// are met. The second indicates whether this server should be considered filtered.
//
// The reason for the two booleans is so that a requirement function could "filter"
// out the left server members if we only want to consider things which are still
// around or likely to come back (failed state).
type serverRequirementFn func(*metadata.Server) (ok bool, filtered bool)
type serversMeetRequirementsState struct {
// meetsRequirements is the callback to actual check for some specific requirement
meetsRequirements serverRequirementFn
// ok indicates whether all unfiltered servers meet the desired requirements
ok bool
// found is a boolean indicating that the meetsRequirement function accepted at
// least one unfiltered server.
found bool
}
func (s *serversMeetRequirementsState) update(srv *metadata.Server) bool {
ok, filtered := s.meetsRequirements(srv)
if filtered {
// keep going but don't update any of the internal state as this server
// was filtered by the requirements function
return true
}
// mark that at least one server processed was not filtered
s.found = true
if !ok {
// mark that at least one server does not meet the requirements
s.ok = false
// prevent continuing server evaluation
return false
}
// this should already be set but this will prevent accidentally reusing
// the state object from causing false-negatives.
s.ok = true
// continue evaluating servers
return true
}
// ServersInDCMeetRequirements returns whether the given server members meet the requirements as defined by the
// callback function and whether at least one server remains unfiltered by the requirements function.
func ServersInDCMeetRequirements(provider checkServersProvider, datacenter string, meetsRequirements serverRequirementFn) (ok bool, found bool) {
state := serversMeetRequirementsState{meetsRequirements: meetsRequirements, found: false, ok: true}
provider.CheckServers(datacenter, state.update)
return state.ok, state.found
}
// ServersInDCMeetMinimumVersion returns whether the given alive servers from a particular
// datacenter are at least on the given Consul version. This also returns whether any
// alive or failed servers are known in that datacenter (ignoring left and leaving ones)
func ServersInDCMeetMinimumVersion(provider checkServersProvider, datacenter string, minVersion *version.Version) (ok bool, found bool) {
return ServersInDCMeetRequirements(provider, datacenter, func(srv *metadata.Server) (bool, bool) {
if srv.Status != serf.StatusAlive && srv.Status != serf.StatusFailed {
// filter out the left servers as those should not be factored into our requirements
return true, true
}
return !srv.Build.LessThan(minVersion), false
})
}
// CheckServers implements the checkServersProvider interface for the Server
func (s *Server) CheckServers(datacenter string, fn func(*metadata.Server) bool) {
if datacenter == s.config.Datacenter {
// use the ServerLookup type for the local DC
s.serverLookup.CheckServers(fn)
} else {
// use the router for all non-local DCs
s.router.CheckServers(datacenter, fn)
}
}
// CheckServers implements the checkServersProvider interface for the Client
func (c *Client) CheckServers(datacenter string, fn func(*metadata.Server) bool) {
if datacenter != c.config.Datacenter {
return
}
c.routers.CheckServers(fn)
}
type serversACLMode struct {
// leader is the address of the leader
leader string
// mode indicates the overall ACL mode of the servers
mode structs.ACLMode
// leaderMode is the ACL mode of the leader server
leaderMode structs.ACLMode
// indicates that at least one server was processed
found bool
}
func (s *serversACLMode) init(leader string) {
s.leader = leader
s.mode = structs.ACLModeEnabled
s.leaderMode = structs.ACLModeUnknown
s.found = false
}
func (s *serversACLMode) update(srv *metadata.Server) bool {
if srv.Status != serf.StatusAlive && srv.Status != serf.StatusFailed {
// they are left or something so regardless we treat these servers as meeting
// the version requirement
return true
}
// mark that we processed at least one server
s.found = true
if srvAddr := srv.Addr.String(); srvAddr == s.leader {
s.leaderMode = srv.ACLs
}
switch srv.ACLs {
case structs.ACLModeDisabled:
// anything disabled means we cant enable ACLs
s.mode = structs.ACLModeDisabled
case structs.ACLModeEnabled:
// do nothing
case structs.ACLModeLegacy:
// This covers legacy mode and older server versions that don't advertise ACL support
if s.mode != structs.ACLModeDisabled && s.mode != structs.ACLModeUnknown {
s.mode = structs.ACLModeLegacy
}
default:
if s.mode != structs.ACLModeDisabled {
s.mode = structs.ACLModeUnknown
}
}
return true
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// ServersGetACLMode checks all the servers in a particular datacenter and determines
// what the minimum ACL mode amongst them is and what the leaders ACL mode is.
// The "found" return value indicates whether there were any servers considered in
// this datacenter. If that is false then the other mode return values are meaningless
// as they will be ACLModeEnabled and ACLModeUnkown respectively.
func ServersGetACLMode(provider checkServersProvider, leaderAddr string, datacenter string) (found bool, mode structs.ACLMode, leaderMode structs.ACLMode) {
var state serversACLMode
state.init(leaderAddr)
provider.CheckServers(datacenter, state.update)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
return state.found, state.mode, state.leaderMode
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}