open-consul/agent/xds/clusters.go

537 lines
17 KiB
Go
Raw Normal View History

package xds
import (
"encoding/json"
"errors"
"fmt"
"time"
envoy "github.com/envoyproxy/go-control-plane/envoy/api/v2"
envoyauth "github.com/envoyproxy/go-control-plane/envoy/api/v2/auth"
envoycluster "github.com/envoyproxy/go-control-plane/envoy/api/v2/cluster"
envoycore "github.com/envoyproxy/go-control-plane/envoy/api/v2/core"
envoyendpoint "github.com/envoyproxy/go-control-plane/envoy/api/v2/endpoint"
envoytype "github.com/envoyproxy/go-control-plane/envoy/type"
"github.com/gogo/protobuf/jsonpb"
"github.com/gogo/protobuf/proto"
"github.com/gogo/protobuf/types"
"github.com/hashicorp/consul/agent/connect"
"github.com/hashicorp/consul/agent/proxycfg"
"github.com/hashicorp/consul/agent/structs"
)
// clustersFromSnapshot returns the xDS API representation of the "clusters" in the snapshot.
2020-03-27 21:57:16 +00:00
func (s *Server) clustersFromSnapshot(cfgSnap *proxycfg.ConfigSnapshot, _ string) ([]proto.Message, error) {
if cfgSnap == nil {
return nil, errors.New("nil config given")
}
switch cfgSnap.Kind {
case structs.ServiceKindConnectProxy:
2020-03-27 21:57:16 +00:00
return s.clustersFromSnapshotConnectProxy(cfgSnap)
case structs.ServiceKindMeshGateway:
2020-03-27 21:57:16 +00:00
return s.clustersFromSnapshotMeshGateway(cfgSnap)
default:
return nil, fmt.Errorf("Invalid service kind: %v", cfgSnap.Kind)
}
}
// clustersFromSnapshot returns the xDS API representation of the "clusters"
// (upstreams) in the snapshot.
2020-03-27 21:57:16 +00:00
func (s *Server) clustersFromSnapshotConnectProxy(cfgSnap *proxycfg.ConfigSnapshot) ([]proto.Message, error) {
// TODO(rb): this sizing is a low bound.
clusters := make([]proto.Message, 0, len(cfgSnap.Proxy.Upstreams)+1)
// Include the "app" cluster for the public listener
appCluster, err := s.makeAppCluster(cfgSnap, LocalAppClusterName, "", cfgSnap.Proxy.LocalServicePort)
if err != nil {
return nil, err
}
clusters = append(clusters, appCluster)
for _, u := range cfgSnap.Proxy.Upstreams {
id := u.Identifier()
if u.DestinationType == structs.UpstreamDestTypePreparedQuery {
upstreamCluster, err := s.makeUpstreamClusterForPreparedQuery(u, cfgSnap)
if err != nil {
return nil, err
}
clusters = append(clusters, upstreamCluster)
} else {
chain := cfgSnap.ConnectProxy.DiscoveryChain[id]
upstreamClusters, err := s.makeUpstreamClustersForDiscoveryChain(u, chain, cfgSnap)
if err != nil {
return nil, err
}
for _, cluster := range upstreamClusters {
clusters = append(clusters, cluster)
}
}
}
cfgSnap.Proxy.Expose.Finalize()
paths := cfgSnap.Proxy.Expose.Paths
// Add service health checks to the list of paths to create clusters for if needed
if cfgSnap.Proxy.Expose.Checks {
psid := structs.NewServiceID(cfgSnap.Proxy.DestinationServiceID, &cfgSnap.ProxyID.EnterpriseMeta)
for _, check := range s.CheckFetcher.ServiceHTTPBasedChecks(psid) {
p, err := parseCheckPath(check)
if err != nil {
s.Logger.Warn("failed to create cluster for", "check", check.CheckID, "error", err)
continue
}
paths = append(paths, p)
}
}
// Create a new cluster if we need to expose a port that is different from the service port
for _, path := range paths {
if path.LocalPathPort == cfgSnap.Proxy.LocalServicePort {
continue
}
c, err := s.makeAppCluster(cfgSnap, makeExposeClusterName(path.LocalPathPort), path.Protocol, path.LocalPathPort)
if err != nil {
s.Logger.Warn("failed to make local cluster", "path", path.Path, "error", err)
continue
}
clusters = append(clusters, c)
}
return clusters, nil
}
func makeExposeClusterName(destinationPort int) string {
return fmt.Sprintf("exposed_cluster_%d", destinationPort)
}
// clustersFromSnapshotMeshGateway returns the xDS API representation of the "clusters"
// for a mesh gateway. This will include 1 cluster per remote datacenter as well as
// 1 cluster for each service subset.
2020-03-27 21:57:16 +00:00
func (s *Server) clustersFromSnapshotMeshGateway(cfgSnap *proxycfg.ConfigSnapshot) ([]proto.Message, error) {
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
datacenters := cfgSnap.MeshGateway.Datacenters()
// 1 cluster per remote dc + 1 cluster per local service (this is a lower bound - all subset specific clusters will be appended)
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
clusters := make([]proto.Message, 0, len(datacenters)+len(cfgSnap.MeshGateway.ServiceGroups))
// generate the remote dc clusters
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
for _, dc := range datacenters {
if dc == cfgSnap.Datacenter {
continue // skip local
}
clusterName := connect.DatacenterSNI(dc, cfgSnap.Roots.TrustDomain)
cluster, err := s.makeMeshGatewayCluster(clusterName, cfgSnap)
if err != nil {
return nil, err
}
clusters = append(clusters, cluster)
}
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
if cfgSnap.ServiceMeta[structs.MetaWANFederationKey] == "1" && cfgSnap.ServerSNIFn != nil {
// Add all of the remote wildcard datacenter mappings for servers.
for _, dc := range datacenters {
clusterName := cfgSnap.ServerSNIFn(dc, "")
cluster, err := s.makeMeshGatewayCluster(clusterName, cfgSnap)
if err != nil {
return nil, err
}
clusters = append(clusters, cluster)
}
// And for the current datacenter, send all flavors appropriately.
for _, srv := range cfgSnap.MeshGateway.ConsulServers {
clusterName := cfgSnap.ServerSNIFn(cfgSnap.Datacenter, srv.Node.Node)
cluster, err := s.makeMeshGatewayCluster(clusterName, cfgSnap)
if err != nil {
return nil, err
}
clusters = append(clusters, cluster)
}
}
// generate the per-service clusters
for svc, _ := range cfgSnap.MeshGateway.ServiceGroups {
clusterName := connect.ServiceSNI(svc.ID, "", svc.NamespaceOrDefault(), cfgSnap.Datacenter, cfgSnap.Roots.TrustDomain)
resolver, hasResolver := cfgSnap.MeshGateway.ServiceResolvers[svc]
// Create the cluster for default/unnamed services
var cluster *envoy.Cluster
var err error
if hasResolver {
cluster, err = s.makeMeshGatewayClusterWithConnectTimeout(clusterName, cfgSnap, resolver.ConnectTimeout)
} else {
cluster, err = s.makeMeshGatewayCluster(clusterName, cfgSnap)
}
if err != nil {
return nil, err
}
clusters = append(clusters, cluster)
// if there is a service-resolver for this service then also setup subset clusters for it
if hasResolver {
// generate 1 cluster for each service subset
for subsetName := range resolver.Subsets {
clusterName := connect.ServiceSNI(svc.ID, subsetName, svc.NamespaceOrDefault(), cfgSnap.Datacenter, cfgSnap.Roots.TrustDomain)
cluster, err := s.makeMeshGatewayClusterWithConnectTimeout(clusterName, cfgSnap, resolver.ConnectTimeout)
if err != nil {
return nil, err
}
clusters = append(clusters, cluster)
}
}
}
return clusters, nil
}
func (s *Server) makeAppCluster(cfgSnap *proxycfg.ConfigSnapshot, name, pathProtocol string, port int) (*envoy.Cluster, error) {
var c *envoy.Cluster
var err error
cfg, err := ParseProxyConfig(cfgSnap.Proxy.Config)
if err != nil {
// Don't hard fail on a config typo, just warn. The parse func returns
// default config if there is an error so it's safe to continue.
s.Logger.Warn("failed to parse Connect.Proxy.Config", "error", err)
}
// If we have overridden local cluster config try to parse it into an Envoy cluster
if cfg.LocalClusterJSON != "" {
return makeClusterFromUserConfig(cfg.LocalClusterJSON)
}
addr := cfgSnap.Proxy.LocalServiceAddress
if addr == "" {
addr = "127.0.0.1"
}
c = &envoy.Cluster{
Name: name,
ConnectTimeout: time.Duration(cfg.LocalConnectTimeoutMs) * time.Millisecond,
ClusterDiscoveryType: &envoy.Cluster_Type{Type: envoy.Cluster_STATIC},
LoadAssignment: &envoy.ClusterLoadAssignment{
ClusterName: name,
Endpoints: []envoyendpoint.LocalityLbEndpoints{
{
LbEndpoints: []envoyendpoint.LbEndpoint{
makeEndpoint(name,
addr,
port),
},
},
},
},
}
protocol := pathProtocol
if protocol == "" {
protocol = cfg.Protocol
}
if protocol == "http2" || protocol == "grpc" {
c.Http2ProtocolOptions = &envoycore.Http2ProtocolOptions{}
}
return c, err
}
func (s *Server) makeUpstreamClusterForPreparedQuery(upstream structs.Upstream, cfgSnap *proxycfg.ConfigSnapshot) (*envoy.Cluster, error) {
var c *envoy.Cluster
var err error
dc := upstream.Datacenter
if dc == "" {
dc = cfgSnap.Datacenter
}
sni := connect.UpstreamSNI(&upstream, "", dc, cfgSnap.Roots.TrustDomain)
cfg, err := ParseUpstreamConfig(upstream.Config)
if err != nil {
// Don't hard fail on a config typo, just warn. The parse func returns
// default config if there is an error so it's safe to continue.
s.Logger.Warn("failed to parse", "upstream", upstream.Identifier(), "error", err)
}
if cfg.ClusterJSON != "" {
c, err = makeClusterFromUserConfig(cfg.ClusterJSON)
if err != nil {
return c, err
}
// In the happy path don't return yet as we need to inject TLS config still.
}
if c == nil {
c = &envoy.Cluster{
Name: sni,
ConnectTimeout: time.Duration(cfg.ConnectTimeoutMs) * time.Millisecond,
ClusterDiscoveryType: &envoy.Cluster_Type{Type: envoy.Cluster_EDS},
EdsClusterConfig: &envoy.Cluster_EdsClusterConfig{
EdsConfig: &envoycore.ConfigSource{
ConfigSourceSpecifier: &envoycore.ConfigSource_Ads{
Ads: &envoycore.AggregatedConfigSource{},
},
},
},
CircuitBreakers: &envoycluster.CircuitBreakers{
Thresholds: makeThresholdsIfNeeded(cfg.Limits),
},
// Having an empty config enables outlier detection with default config.
OutlierDetection: &envoycluster.OutlierDetection{},
}
if cfg.Protocol == "http2" || cfg.Protocol == "grpc" {
c.Http2ProtocolOptions = &envoycore.Http2ProtocolOptions{}
}
}
// Enable TLS upstream with the configured client certificate.
c.TlsContext = &envoyauth.UpstreamTlsContext{
CommonTlsContext: makeCommonTLSContext(cfgSnap),
Sni: sni,
}
return c, nil
}
func (s *Server) makeUpstreamClustersForDiscoveryChain(
upstream structs.Upstream,
chain *structs.CompiledDiscoveryChain,
cfgSnap *proxycfg.ConfigSnapshot,
) ([]*envoy.Cluster, error) {
if chain == nil {
return nil, fmt.Errorf("cannot create upstream cluster without discovery chain for %s", upstream.Identifier())
}
cfg, err := ParseUpstreamConfigNoDefaults(upstream.Config)
if err != nil {
// Don't hard fail on a config typo, just warn. The parse func returns
// default config if there is an error so it's safe to continue.
s.Logger.Warn("failed to parse", "upstream", upstream.Identifier(),
"error", err)
}
var escapeHatchCluster *envoy.Cluster
if cfg.ClusterJSON != "" {
if chain.IsDefault() {
// If you haven't done anything to setup the discovery chain, then
// you can use the envoy_cluster_json escape hatch.
escapeHatchCluster, err = makeClusterFromUserConfig(cfg.ClusterJSON)
if err != nil {
return nil, err
}
} else {
s.Logger.Warn("ignoring escape hatch setting, because a discovery chain is configued for",
"discovery chain", chain.ServiceName, "upstream", upstream.Identifier(),
"envoy_cluster_json", chain.ServiceName)
}
}
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
id := upstream.Identifier()
chainEndpointMap, ok := cfgSnap.ConnectProxy.WatchedUpstreamEndpoints[id]
if !ok {
// this should not happen
return nil, fmt.Errorf("no endpoint map for upstream %q", id)
}
var out []*envoy.Cluster
for _, node := range chain.Nodes {
if node.Type != structs.DiscoveryGraphNodeTypeResolver {
continue
}
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
failover := node.Resolver.Failover
targetID := node.Resolver.Target
target := chain.Targets[targetID]
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
// Determine if we have to generate the entire cluster differently.
failoverThroughMeshGateway := chain.WillFailoverThroughMeshGateway(node)
sni := target.SNI
clusterName := CustomizeClusterName(target.Name, chain)
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
if failoverThroughMeshGateway {
actualTargetID := firstHealthyTarget(
chain.Targets,
chainEndpointMap,
targetID,
failover.Targets,
)
if actualTargetID != targetID {
actualTarget := chain.Targets[actualTargetID]
sni = actualTarget.SNI
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
}
}
s.Logger.Debug("generating cluster for", "cluster", clusterName)
c := &envoy.Cluster{
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
Name: clusterName,
AltStatName: clusterName,
ConnectTimeout: node.Resolver.ConnectTimeout,
ClusterDiscoveryType: &envoy.Cluster_Type{Type: envoy.Cluster_EDS},
CommonLbConfig: &envoy.Cluster_CommonLbConfig{
HealthyPanicThreshold: &envoytype.Percent{
Value: 0, // disable panic threshold
},
},
EdsClusterConfig: &envoy.Cluster_EdsClusterConfig{
EdsConfig: &envoycore.ConfigSource{
ConfigSourceSpecifier: &envoycore.ConfigSource_Ads{
Ads: &envoycore.AggregatedConfigSource{},
},
},
},
CircuitBreakers: &envoycluster.CircuitBreakers{
Thresholds: makeThresholdsIfNeeded(cfg.Limits),
},
// Having an empty config enables outlier detection with default config.
OutlierDetection: &envoycluster.OutlierDetection{},
}
proto := cfg.Protocol
if proto == "" {
proto = chain.Protocol
}
if proto == "" {
proto = "tcp"
}
if proto == "http2" || proto == "grpc" {
c.Http2ProtocolOptions = &envoycore.Http2ProtocolOptions{}
}
// Enable TLS upstream with the configured client certificate.
c.TlsContext = &envoyauth.UpstreamTlsContext{
CommonTlsContext: makeCommonTLSContext(cfgSnap),
Sni: sni,
}
out = append(out, c)
}
if escapeHatchCluster != nil {
if len(out) != 1 {
return nil, fmt.Errorf("cannot inject escape hatch cluster when discovery chain had no nodes")
}
defaultCluster := out[0]
// Overlay what the user provided.
escapeHatchCluster.TlsContext = defaultCluster.TlsContext
out = []*envoy.Cluster{escapeHatchCluster}
}
return out, nil
}
// makeClusterFromUserConfig returns the listener config decoded from an
// arbitrary proto3 json format string or an error if it's invalid.
//
// For now we only support embedding in JSON strings because of the hcl parsing
// pain (see config.go comment above call to PatchSliceOfMaps). Until we
// refactor config parser a _lot_ user's opaque config that contains arrays will
// be mangled. We could actually fix that up in mapstructure which knows the
// type of the target so could resolve the slices to singletons unambiguously
// and it would work for us here... but we still have the problem that the
// config would render incorrectly in general in our HTTP API responses so we
// really need to fix it "properly".
//
// When we do that we can support just nesting the config directly into the
// JSON/hcl naturally but this is a stop-gap that gets us an escape hatch
// immediately. It's also probably not a bad thing to support long-term since
// any config generated by other systems will likely be in canonical protobuf
// from rather than our slight variant in JSON/hcl.
func makeClusterFromUserConfig(configJSON string) (*envoy.Cluster, error) {
var jsonFields map[string]*json.RawMessage
if err := json.Unmarshal([]byte(configJSON), &jsonFields); err != nil {
fmt.Println("Custom error", err, configJSON)
return nil, err
}
var c envoy.Cluster
if _, ok := jsonFields["@type"]; ok {
// Type field is present so decode it as a types.Any
var any types.Any
err := jsonpb.UnmarshalString(configJSON, &any)
if err != nil {
return nil, err
}
// And then unmarshal the listener again...
err = proto.Unmarshal(any.Value, &c)
if err != nil {
return nil, err
}
return &c, err
}
// No @type so try decoding as a straight listener.
err := jsonpb.UnmarshalString(configJSON, &c)
return &c, err
}
func (s *Server) makeMeshGatewayCluster(clusterName string, cfgSnap *proxycfg.ConfigSnapshot) (*envoy.Cluster, error) {
return s.makeMeshGatewayClusterWithConnectTimeout(clusterName, cfgSnap, 0)
}
// makeMeshGatewayClusterWithConnectTimeout initializes a mesh gateway cluster
// with the specified connect timeout. If the timeout is 0, the connect timeout
// defaults to use the mesh gateway timeout.
func (s *Server) makeMeshGatewayClusterWithConnectTimeout(clusterName string, cfgSnap *proxycfg.ConfigSnapshot,
connectTimeout time.Duration) (*envoy.Cluster, error) {
cfg, err := ParseGatewayConfig(cfgSnap.Proxy.Config)
if err != nil {
// Don't hard fail on a config typo, just warn. The parse func returns
// default config if there is an error so it's safe to continue.
s.Logger.Warn("failed to parse mesh gateway config", "error", err)
}
if connectTimeout <= 0 {
connectTimeout = time.Duration(cfg.ConnectTimeoutMs) * time.Millisecond
}
return &envoy.Cluster{
Name: clusterName,
ConnectTimeout: connectTimeout,
ClusterDiscoveryType: &envoy.Cluster_Type{Type: envoy.Cluster_EDS},
EdsClusterConfig: &envoy.Cluster_EdsClusterConfig{
EdsConfig: &envoycore.ConfigSource{
ConfigSourceSpecifier: &envoycore.ConfigSource_Ads{
Ads: &envoycore.AggregatedConfigSource{},
},
},
},
// Having an empty config enables outlier detection with default config.
OutlierDetection: &envoycluster.OutlierDetection{},
}, nil
}
func makeThresholdsIfNeeded(limits UpstreamLimits) []*envoycluster.CircuitBreakers_Thresholds {
var empty UpstreamLimits
// Make sure to not create any thresholds when passed the zero-value in order
// to rely on Envoy defaults
if limits == empty {
return nil
}
threshold := &envoycluster.CircuitBreakers_Thresholds{}
// Likewise, make sure to not set any threshold values on the zero-value in
// order to rely on Envoy defaults
if limits.MaxConnections != nil {
threshold.MaxConnections = makeUint32Value(*limits.MaxConnections)
}
if limits.MaxPendingRequests != nil {
threshold.MaxPendingRequests = makeUint32Value(*limits.MaxPendingRequests)
}
if limits.MaxConcurrentRequests != nil {
threshold.MaxRequests = makeUint32Value(*limits.MaxConcurrentRequests)
}
return []*envoycluster.CircuitBreakers_Thresholds{threshold}
}