Prior to #13244, connect proxies and gateways could only be configured by an
xDS session served by the local client agent.
In an upcoming release, it will be possible to deploy a Consul service mesh
without client agents. In this model, xDS sessions will be handled by the
servers themselves, which necessitates load-balancing to prevent a single
server from receiving a disproportionate amount of load and becoming
overwhelmed.
This introduces a simple form of load-balancing where Consul will attempt to
achieve an even spread of load (xDS sessions) between all healthy servers.
It does so by implementing a concurrent session limiter (limiter.SessionLimiter)
and adjusting the limit according to autopilot state and proxy service
registrations in the catalog.
If a server is already over capacity (i.e. the session limit is lowered),
Consul will begin draining sessions to rebalance the load. This will result
in the client receiving a `RESOURCE_EXHAUSTED` status code. It is the client's
responsibility to observe this response and reconnect to a different server.
Users of the gRPC client connection brokered by the
consul-server-connection-manager library will get this for free.
The rate at which Consul will drain sessions to rebalance load is scaled
dynamically based on the number of proxies in the catalog.
* draft commit
* add changelog, update test
* remove extra param
* fix test
* update type to account for nil value
* add test for custom passive health check
* update comments and tests
* update description in docs
* fix missing commas
This is the OSS portion of enterprise PR 2339.
It improves our handling of "irrecoverable" errors in proxycfg data sources.
The canonical example of this is what happens when the ACL token presented by
Envoy is deleted/revoked. Previously, the stream would get "stuck" until the
xDS server re-checked the token (after 5 minutes) and terminated the stream.
Materializers would also sit burning resources retrying something that could
never succeed.
Now, it is possible for data sources to mark errors as "terminal" which causes
the xDS stream to be closed immediately. Similarly, the submatview.Store will
evict materializers when it observes they have encountered such an error.
* add golden files
* add support to http in tgateway egress destination
* fix slice sorting to include both address and port when using server_names
* fix listener loop for http destination
* fix routes to generate a route per port and a virtualhost per port-address combination
* sort virtual hosts list to have a stable order
* extract redundant serviceNode
Now that peered upstreams can generate envoy resources (#13758), we need a way to disambiguate local from peered resources in our metrics. The key difference is that datacenter and partition will be replaced with peer, since in the context of peered resources partition is ambiguous (could refer to the partition in a remote cluster or one that exists locally). The partition and datacenter of the proxy will always be that of the source service.
Regexes were updated to make emitting datacenter and partition labels mutually exclusive with peer labels.
Listener filter names were updated to better match the existing regex.
Cluster names assigned to peered upstreams were updated to be synthesized from local peer name (it previously used the externally provided primary SNI, which contained the peer name from the other side of the peering). Integration tests were updated to assert for the new peer labels.
Peered upstreams has a separate loop in xds from discovery chain upstreams. This PR adds similar but slightly modified code to add filters for peered upstream listeners, clusters, and endpoints in the case of transparent proxy.
Previously, public referred to gRPC services that are both exposed on
the dedicated gRPC port and have their definitions in the proto-public
directory (so were considered usable by 3rd parties). Whereas private
referred to services on the multiplexed server port that are only usable
by agents and other servers.
Now, we're splitting these definitions, such that external/internal
refers to the port and public/private refers to whether they can be used
by 3rd parties.
This is necessary because the peering replication API needs to be
exposed on the dedicated port, but is not (yet) suitable for use by 3rd
parties.
Because peerings are pairwise, between two tuples of (datacenter,
partition) having any exported reference via a discovery chain that
crosses out of the peered datacenter or partition will ultimately not be
able to work for various reasons. The biggest one is that there is no
way in the ultimate destination to configure an intention that can allow
an external SpiffeID to access a service.
This PR ensures that a user simply cannot do this, so they won't run
into weird situations like this.
When the protocol is http-like, and an intention has a peered source
then the normal RBAC mTLS SAN field check is replaces with a joint combo
of:
mTLS SAN field must be the service's local mesh gateway leaf cert
AND
the first XFCC header (from the MGW) must have a URI field that matches the original intention source
Also:
- Update the regex program limit to be much higher than the teeny
defaults, since the RBAC regex constructions are more complicated now.
- Fix a few stray panics in xds generation.
This is only configured in xDS when a service with an L7 protocol is
exported.
They also load any relevant trust bundles for the peered services to
eventually use for L7 SPIFFE validation during mTLS termination.
When converting from Consul intentions to xds RBAC rules, services imported from other peers must encode additional data like partition (from the remote cluster) and trust domain.
This PR updates the PeeringTrustBundle to hold the sending side's local partition as ExportedPartition. It also updates RBAC code to encode SpiffeIDs of imported services with the ExportedPartition and TrustDomain.
Mesh gateways can use hostnames in their tagged addresses (#7999). This is useful
if you were to expose a mesh gateway using a cloud networking load balancer appliance
that gives you a DNS name but no reliable static IPs.
Envoy cannot accept hostnames via EDS and those must be configured using CDS.
There was already logic when configuring gateways in other locations in the code, but
given the illusions in play for peering the downstream of a peered service wasn't aware
that it should be doing that.
Also:
- ensuring that we always try to use wan-like addresses to cross peer boundaries.
Mesh gateways will now enable tcp connections with SNI names including peering information so that those connections may be proxied.
Note: this does not change the callers to use these mesh gateways.
This is the OSS portion of enterprise PR 1994
Rather than directly interrogating the agent-local state for HTTP
checks using the `HTTPCheckFetcher` interface, we now rely on the
config snapshot containing the checks.
This reduces the number of changes required to support server xDS
sessions.
It's not clear why the fetching approach was introduced in
931d167ebb2300839b218d08871f22323c60175d.
Envoy's SPIFFE certificate validation extension allows for us to
validate against different root certificates depending on the trust
domain of the dialing proxy.
If there are any trust bundles from peers in the config snapshot then we
use the SPIFFE validator as the validation context, rather than the
usual TrustedCA.
The injected validation config includes the local root certificates as
well.
For mTLS to work between two proxies in peered clusters with different root CAs,
proxies need to configure their outbound listener to use different root certificates
for validation.
Up until peering was introduced proxies would only ever use one set of root certificates
to validate all mesh traffic, both inbound and outbound. Now an upstream proxy
may have a leaf certificate signed by a CA that's different from the dialing proxy's.
This PR makes changes to proxycfg and xds so that the upstream TLS validation
uses different root certificates depending on which cluster is being dialed.
malizz