open-consul/agent/xds/endpoints.go

555 lines
17 KiB
Go
Raw Normal View History

package xds
import (
"errors"
"fmt"
envoy "github.com/envoyproxy/go-control-plane/envoy/api/v2"
envoycore "github.com/envoyproxy/go-control-plane/envoy/api/v2/core"
envoyendpoint "github.com/envoyproxy/go-control-plane/envoy/api/v2/endpoint"
"github.com/gogo/protobuf/proto"
"github.com/hashicorp/consul/agent/connect"
"github.com/hashicorp/consul/agent/proxycfg"
"github.com/hashicorp/consul/agent/structs"
"github.com/hashicorp/consul/api"
bexpr "github.com/hashicorp/go-bexpr"
)
const (
UnnamedSubset = ""
)
// endpointsFromSnapshot returns the xDS API representation of the "endpoints"
2020-03-27 21:57:16 +00:00
func (s *Server) endpointsFromSnapshot(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.endpointsFromSnapshotConnectProxy(cfgSnap)
case structs.ServiceKindTerminatingGateway:
return s.endpointsFromSnapshotTerminatingGateway(cfgSnap)
case structs.ServiceKindMeshGateway:
2020-03-27 21:57:16 +00:00
return s.endpointsFromSnapshotMeshGateway(cfgSnap)
case structs.ServiceKindIngressGateway:
return s.endpointsFromSnapshotIngressGateway(cfgSnap)
default:
return nil, fmt.Errorf("Invalid service kind: %v", cfgSnap.Kind)
}
}
// endpointsFromSnapshotConnectProxy returns the xDS API representation of the "endpoints"
// (upstream instances) in the snapshot.
2020-03-27 21:57:16 +00:00
func (s *Server) endpointsFromSnapshotConnectProxy(cfgSnap *proxycfg.ConfigSnapshot) ([]proto.Message, error) {
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
resources := make([]proto.Message, 0,
len(cfgSnap.ConnectProxy.PreparedQueryEndpoints)+len(cfgSnap.ConnectProxy.WatchedUpstreamEndpoints))
for _, u := range cfgSnap.Proxy.Upstreams {
id := u.Identifier()
var chain *structs.CompiledDiscoveryChain
if u.DestinationType != structs.UpstreamDestTypePreparedQuery {
chain = cfgSnap.ConnectProxy.DiscoveryChain[id]
}
if chain == nil {
// We ONLY want this branch for prepared queries.
dc := u.Datacenter
if dc == "" {
dc = cfgSnap.Datacenter
}
clusterName := connect.UpstreamSNI(&u, "", dc, cfgSnap.Roots.TrustDomain)
endpoints, ok := cfgSnap.ConnectProxy.PreparedQueryEndpoints[id]
if ok {
la := makeLoadAssignment(
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
clusterName,
[]loadAssignmentEndpointGroup{
{Endpoints: endpoints},
},
cfgSnap.Datacenter,
)
resources = append(resources, la)
}
} else {
// Newfangled discovery chain plumbing.
es := s.endpointsFromDiscoveryChain(
u,
chain,
cfgSnap.Datacenter,
cfgSnap.ConnectProxy.WatchedUpstreamEndpoints[id],
cfgSnap.ConnectProxy.WatchedGatewayEndpoints[id],
)
resources = append(resources, es...)
}
}
return resources, nil
}
func (s *Server) filterSubsetEndpoints(subset *structs.ServiceResolverSubset, endpoints structs.CheckServiceNodes) (structs.CheckServiceNodes, error) {
// locally execute the subsets filter
if subset.Filter != "" {
filter, err := bexpr.CreateFilter(subset.Filter, nil, endpoints)
if err != nil {
return nil, err
}
raw, err := filter.Execute(endpoints)
if err != nil {
return nil, err
}
return raw.(structs.CheckServiceNodes), nil
}
return endpoints, nil
}
func (s *Server) endpointsFromSnapshotTerminatingGateway(cfgSnap *proxycfg.ConfigSnapshot) ([]proto.Message, error) {
2020-04-14 14:59:23 +00:00
return s.endpointsFromServicesAndResolvers(cfgSnap, cfgSnap.TerminatingGateway.ServiceGroups, cfgSnap.TerminatingGateway.ServiceResolvers)
}
2020-03-27 21:57:16 +00:00
func (s *Server) endpointsFromSnapshotMeshGateway(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()
resources := make([]proto.Message, 0, len(datacenters)+len(cfgSnap.MeshGateway.ServiceGroups))
// generate the endpoints for the gateways in the remote datacenters
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 {
// Skip creating endpoints for mesh gateways in local DC and gateways in remote DCs with a hostname as their address
// EDS cannot resolve hostnames so we provide them through CDS instead
if dc == cfgSnap.Datacenter || len(cfgSnap.MeshGateway.HostnameDatacenters[dc]) > 0 {
continue
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
}
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
endpoints, ok := cfgSnap.MeshGateway.GatewayGroups[dc]
if !ok {
endpoints, ok = cfgSnap.MeshGateway.FedStateGateways[dc]
if !ok { // not possible
s.Logger.Error("skipping mesh gateway endpoints because no definition found", "datacenter", dc)
continue
}
}
{ // standard connect
clusterName := connect.DatacenterSNI(dc, cfgSnap.Roots.TrustDomain)
la := makeLoadAssignment(
clusterName,
[]loadAssignmentEndpointGroup{
{Endpoints: endpoints},
},
cfgSnap.Datacenter,
)
resources = append(resources, la)
}
if cfgSnap.ServiceMeta[structs.MetaWANFederationKey] == "1" && cfgSnap.ServerSNIFn != nil {
clusterName := cfgSnap.ServerSNIFn(dc, "")
la := makeLoadAssignment(
clusterName,
[]loadAssignmentEndpointGroup{
{Endpoints: endpoints},
},
cfgSnap.Datacenter,
)
resources = append(resources, la)
}
}
// generate endpoints for our servers if WAN federation is enabled
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 {
var allServersLbEndpoints []envoyendpoint.LbEndpoint
for _, srv := range cfgSnap.MeshGateway.ConsulServers {
clusterName := cfgSnap.ServerSNIFn(cfgSnap.Datacenter, srv.Node.Node)
addr, port := srv.BestAddress(false /*wan*/)
lbEndpoint := envoyendpoint.LbEndpoint{
HostIdentifier: &envoyendpoint.LbEndpoint_Endpoint{
Endpoint: &envoyendpoint.Endpoint{
Address: makeAddressPtr(addr, port),
},
},
HealthStatus: envoycore.HealthStatus_UNKNOWN,
}
cla := &envoy.ClusterLoadAssignment{
ClusterName: clusterName,
Endpoints: []envoyendpoint.LocalityLbEndpoints{{
LbEndpoints: []envoyendpoint.LbEndpoint{lbEndpoint},
}},
}
allServersLbEndpoints = append(allServersLbEndpoints, lbEndpoint)
resources = append(resources, cla)
}
// And add one catch all so that remote datacenters can dial ANY server
// in this datacenter without knowing its name.
resources = append(resources, &envoy.ClusterLoadAssignment{
ClusterName: cfgSnap.ServerSNIFn(cfgSnap.Datacenter, ""),
Endpoints: []envoyendpoint.LocalityLbEndpoints{{
LbEndpoints: allServersLbEndpoints,
}},
})
}
// Generate the endpoints for each service and its subsets
2020-04-14 14:59:23 +00:00
e, err := s.endpointsFromServicesAndResolvers(cfgSnap, cfgSnap.MeshGateway.ServiceGroups, cfgSnap.MeshGateway.ServiceResolvers)
if err != nil {
return nil, err
}
resources = append(resources, e...)
return resources, nil
}
func (s *Server) endpointsFromServicesAndResolvers(
cfgSnap *proxycfg.ConfigSnapshot,
services map[structs.ServiceID]structs.CheckServiceNodes,
resolvers map[structs.ServiceID]*structs.ServiceResolverConfigEntry) ([]proto.Message, error) {
resources := make([]proto.Message, 0, len(services))
// generate the endpoints for the linked service groups
for svc, endpoints := range services {
// Skip creating endpoints for services that have hostnames as addresses
// EDS cannot resolve hostnames so we provide them through CDS instead
if cfgSnap.Kind == structs.ServiceKindTerminatingGateway && len(cfgSnap.TerminatingGateway.HostnameServices[svc]) > 0 {
continue
}
2020-04-14 14:59:23 +00:00
clusterEndpoints := make(map[string][]loadAssignmentEndpointGroup)
clusterEndpoints[UnnamedSubset] = []loadAssignmentEndpointGroup{{Endpoints: endpoints, OnlyPassing: false}}
// Collect all of the loadAssignmentEndpointGroups for the various subsets. We do this before generating
// the endpoints for the default/unnamed subset so that we can take into account the DefaultSubset on the
// service-resolver which may prevent the default/unnamed cluster from creating endpoints for all service
// instances.
2020-04-14 14:59:23 +00:00
if resolver, hasResolver := resolvers[svc]; hasResolver {
for subsetName, subset := range resolver.Subsets {
subsetEndpoints, err := s.filterSubsetEndpoints(&subset, endpoints)
if err != nil {
return nil, err
}
2020-04-14 14:59:23 +00:00
groups := []loadAssignmentEndpointGroup{{Endpoints: subsetEndpoints, OnlyPassing: subset.OnlyPassing}}
clusterEndpoints[subsetName] = groups
// if this subset is the default then override the unnamed subset with this configuration
if subsetName == resolver.DefaultSubset {
2020-04-14 14:59:23 +00:00
clusterEndpoints[UnnamedSubset] = groups
}
}
}
// now generate the load assignment for all subsets
2020-04-14 14:59:23 +00:00
for subsetName, groups := range clusterEndpoints {
clusterName := connect.ServiceSNI(svc.ID, subsetName, svc.NamespaceOrDefault(), cfgSnap.Datacenter, cfgSnap.Roots.TrustDomain)
la := makeLoadAssignment(
clusterName,
2020-04-14 14:59:23 +00:00
groups,
cfgSnap.Datacenter,
)
resources = append(resources, la)
}
}
return resources, nil
}
func (s *Server) endpointsFromSnapshotIngressGateway(cfgSnap *proxycfg.ConfigSnapshot) ([]proto.Message, error) {
var resources []proto.Message
createdClusters := make(map[string]bool)
for _, upstreams := range cfgSnap.IngressGateway.Upstreams {
for _, u := range upstreams {
id := u.Identifier()
// If we've already created endpoints for this upstream, skip it. Multiple listeners may
// reference the same upstream, so we don't need to create duplicate endpoints in that case.
if createdClusters[id] {
continue
}
es := s.endpointsFromDiscoveryChain(
u,
cfgSnap.IngressGateway.DiscoveryChain[id],
cfgSnap.Datacenter,
cfgSnap.IngressGateway.WatchedUpstreamEndpoints[id],
cfgSnap.IngressGateway.WatchedGatewayEndpoints[id],
)
resources = append(resources, es...)
createdClusters[id] = true
}
}
return resources, nil
}
func makeEndpoint(clusterName, host string, port int) envoyendpoint.LbEndpoint {
return envoyendpoint.LbEndpoint{
HostIdentifier: &envoyendpoint.LbEndpoint_Endpoint{
Endpoint: &envoyendpoint.Endpoint{
Address: makeAddressPtr(host, port),
},
},
}
}
func (s *Server) endpointsFromDiscoveryChain(
upstream structs.Upstream,
chain *structs.CompiledDiscoveryChain,
datacenter string,
upstreamEndpoints, gatewayEndpoints map[string]structs.CheckServiceNodes,
) []proto.Message {
var resources []proto.Message
if chain == nil {
return resources
}
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 resources
}
} 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)
}
}
// Find all resolver nodes.
for _, node := range chain.Nodes {
if node.Type != structs.DiscoveryGraphNodeTypeResolver {
continue
}
failover := node.Resolver.Failover
targetID := node.Resolver.Target
target := chain.Targets[targetID]
clusterName := CustomizeClusterName(target.Name, chain)
if escapeHatchCluster != nil {
clusterName = escapeHatchCluster.Name
}
s.Logger.Debug("generating endpoints for", "cluster", clusterName)
// Determine if we have to generate the entire cluster differently.
failoverThroughMeshGateway := chain.WillFailoverThroughMeshGateway(node)
if failoverThroughMeshGateway {
actualTargetID := firstHealthyTarget(
chain.Targets,
upstreamEndpoints,
targetID,
failover.Targets,
)
if actualTargetID != targetID {
targetID = actualTargetID
}
failover = nil
}
primaryGroup, valid := makeLoadAssignmentEndpointGroup(
chain.Targets,
upstreamEndpoints,
gatewayEndpoints,
targetID,
datacenter,
)
if !valid {
continue // skip the cluster if we're still populating the snapshot
}
var endpointGroups []loadAssignmentEndpointGroup
if failover != nil && len(failover.Targets) > 0 {
endpointGroups = make([]loadAssignmentEndpointGroup, 0, len(failover.Targets)+1)
endpointGroups = append(endpointGroups, primaryGroup)
for _, failTargetID := range failover.Targets {
failoverGroup, valid := makeLoadAssignmentEndpointGroup(
chain.Targets,
upstreamEndpoints,
gatewayEndpoints,
failTargetID,
datacenter,
)
if !valid {
continue // skip the failover target if we're still populating the snapshot
}
endpointGroups = append(endpointGroups, failoverGroup)
}
} else {
endpointGroups = append(endpointGroups, primaryGroup)
}
la := makeLoadAssignment(
clusterName,
endpointGroups,
datacenter,
)
resources = append(resources, la)
}
return resources
}
type loadAssignmentEndpointGroup struct {
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
Endpoints structs.CheckServiceNodes
OnlyPassing bool
OverrideHealth envoycore.HealthStatus
}
func makeLoadAssignment(clusterName string, endpointGroups []loadAssignmentEndpointGroup, localDatacenter string) *envoy.ClusterLoadAssignment {
cla := &envoy.ClusterLoadAssignment{
ClusterName: clusterName,
Endpoints: make([]envoyendpoint.LocalityLbEndpoints, 0, len(endpointGroups)),
}
if len(endpointGroups) > 1 {
cla.Policy = &envoy.ClusterLoadAssignment_Policy{
// We choose such a large value here that the failover math should
// in effect not happen until zero instances are healthy.
OverprovisioningFactor: makeUint32Value(100000),
}
}
for priority, endpointGroup := range endpointGroups {
endpoints := endpointGroup.Endpoints
es := make([]envoyendpoint.LbEndpoint, 0, len(endpoints))
for _, ep := range endpoints {
// TODO (mesh-gateway) - should we respect the translate_wan_addrs configuration here or just always use the wan for cross-dc?
addr, port := ep.BestAddress(localDatacenter != ep.Node.Datacenter)
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
healthStatus, weight := calculateEndpointHealthAndWeight(ep, endpointGroup.OnlyPassing)
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 endpointGroup.OverrideHealth != envoycore.HealthStatus_UNKNOWN {
healthStatus = endpointGroup.OverrideHealth
}
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
es = append(es, envoyendpoint.LbEndpoint{
HostIdentifier: &envoyendpoint.LbEndpoint_Endpoint{
Endpoint: &envoyendpoint.Endpoint{
Address: makeAddressPtr(addr, port),
},
},
HealthStatus: healthStatus,
LoadBalancingWeight: makeUint32Value(weight),
})
}
cla.Endpoints = append(cla.Endpoints, envoyendpoint.LocalityLbEndpoints{
Priority: uint32(priority),
LbEndpoints: es,
})
}
return cla
}
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
func makeLoadAssignmentEndpointGroup(
targets map[string]*structs.DiscoveryTarget,
targetHealth map[string]structs.CheckServiceNodes,
gatewayHealth map[string]structs.CheckServiceNodes,
targetID string,
currentDatacenter string,
) (loadAssignmentEndpointGroup, bool) {
realEndpoints, ok := targetHealth[targetID]
if !ok {
// skip the cluster if we're still populating the snapshot
return loadAssignmentEndpointGroup{}, false
}
target := targets[targetID]
var gatewayDatacenter string
switch target.MeshGateway.Mode {
case structs.MeshGatewayModeRemote:
gatewayDatacenter = target.Datacenter
case structs.MeshGatewayModeLocal:
gatewayDatacenter = currentDatacenter
}
if gatewayDatacenter == "" {
return loadAssignmentEndpointGroup{
Endpoints: realEndpoints,
OnlyPassing: target.Subset.OnlyPassing,
}, true
}
// If using a mesh gateway we need to pull those endpoints instead.
gatewayEndpoints, ok := gatewayHealth[gatewayDatacenter]
if !ok {
// skip the cluster if we're still populating the snapshot
return loadAssignmentEndpointGroup{}, false
}
// But we will use the health from the actual backend service.
overallHealth := envoycore.HealthStatus_UNHEALTHY
for _, ep := range realEndpoints {
health, _ := calculateEndpointHealthAndWeight(ep, target.Subset.OnlyPassing)
if health == envoycore.HealthStatus_HEALTHY {
overallHealth = envoycore.HealthStatus_HEALTHY
break
}
}
return loadAssignmentEndpointGroup{
Endpoints: gatewayEndpoints,
OverrideHealth: overallHealth,
}, true
}
func calculateEndpointHealthAndWeight(
ep structs.CheckServiceNode,
onlyPassing bool,
) (envoycore.HealthStatus, int) {
healthStatus := envoycore.HealthStatus_HEALTHY
weight := 1
if ep.Service.Weights != nil {
weight = ep.Service.Weights.Passing
}
for _, chk := range ep.Checks {
if chk.Status == api.HealthCritical {
healthStatus = envoycore.HealthStatus_UNHEALTHY
}
if onlyPassing && chk.Status != api.HealthPassing {
healthStatus = envoycore.HealthStatus_UNHEALTHY
}
if chk.Status == api.HealthWarning && ep.Service.Weights != nil {
weight = ep.Service.Weights.Warning
}
}
// Make weights fit Envoy's limits. A zero weight means that either Warning
// (likely) or Passing (weirdly) weight has been set to 0 effectively making
// this instance unhealthy and should not be sent traffic.
if weight < 1 {
healthStatus = envoycore.HealthStatus_UNHEALTHY
weight = 1
}
if weight > 128 {
weight = 128
}
return healthStatus, weight
}