When traversing an exported peered service, the discovery chain
evaluation at the other side may re-route the request to a variety of
endpoints. Furthermore we intend to terminate mTLS at the mesh gateway
for arriving peered traffic that is http-like (L7), so the caller needs
to know the mesh gateway's SpiffeID in that case as well.
The following new SpiffeID values will be shipped back in the peerstream
replication:
- tcp: all possible SpiffeIDs resulting from the service-resolver
component of the exported discovery chain
- http-like: the SpiffeID of the mesh gateway
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 our peer deletes the peering it is locally marked as terminated.
This termination should kick off deleting all imported data, but should
not delete the peering object itself.
Keeping peerings marked as terminated acts as a signal that the action
took place.
Once a peering is marked for deletion a new leader routine will now
clean up all imported resources and then the peering itself.
A lot of the logic was grabbed from the namespace/partitions deferred
deletions but with a handful of simplifications:
- The rate limiting is not configurable.
- Deleting imported nodes/services/checks is done by deleting nodes with
the Txn API. The services and checks are deleted as a side-effect.
- There is no "round rate limiter" like with namespaces and partitions.
This is because peerings are purely local, and deleting a peering in
the datacenter does not depend on deleting data from other DCs like
with WAN-federated namespaces. All rate limiting is handled by the
Raft rate limiter.
1. Fix a bug where the peering leader routine would not track all active
peerings in the "stored" reconciliation map. This could lead to
tearing down streams where the token was generated, since the
ConnectedStreams() method used for reconciliation returns all streams
and not just the ones initiated by this leader routine.
2. Fix a race where stream contexts were being canceled before
termination messages were being processed by a peer.
Previously the leader routine would tear down streams by canceling
their context right after the termination message was sent. This
context cancelation could be propagated to the server side faster
than the termination message. Now there is a change where the
dialing peer uses CloseSend() to signal when no more messages will
be sent. Eventually the server peer will read an EOF after receiving
and processing the preceding termination message.
Using CloseSend() is actually not enough to address the issue
mentioned, since it doesn't wait for the server peer to finish
processing messages. Because of this now the dialing peer also reads
from the stream until an error signals that there are no more
messages. Receiving an EOF from our peer indicates that they
processed the termination message and have no additional work to do.
Given that the stream is being closed, all the messages received by
Recv are discarded. We only check for errors to avoid importing new
data.
When deleting a peering we do not want to delete the peering and all
imported data in a single operation, since deleting a large amount of
data at once could overload Consul.
Instead we defer deletion of peerings so that:
1. When a peering deletion request is received via gRPC the peering is
marked for deletion by setting the DeletedAt field.
2. A leader routine will monitor for peerings that are marked for
deletion and kick off a throttled deletion of all imported resources
before deleting the peering itself.
This commit mostly addresses point #1 by modifying the peering service
to mark peerings for deletion. Another key change is to add a
PeeringListDeleted state store function which can return all peerings
marked for deletion. This function is what will be watched by the
deferred deletion leader routine.
Previously, imported data would never be deleted. As
nodes/services/checks were registered and deregistered, resources
deleted from the exporting cluster would accumulate in the imported
cluster.
This commit makes updates to replication so that whenever an update is
received for a service name we reconcile what was present in the catalog
against what was received.
This handleUpdateService method can handle both updates and deletions.
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.
Require use of mesh gateways in order for service mesh data plane
traffic to flow between peers.
This also adds plumbing for envoy integration tests involving peers, and
one starter peering test.
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.
There are a handful of changes in this commit:
* When querying trust bundles for a service we need to be able to
specify the namespace of the service.
* The endpoint needs to track the index because the cache watches use
it.
* Extracted bulk of the endpoint's logic to a state store function
so that index tracking could be tested more easily.
* Removed check for service existence, deferring that sort of work to ACL authz
* Added the cache type
Given that the exported-services config entry can use wildcards, the
precedence for wildcards is handled as with intentions. The most exact
match is the match that applies for any given service. We do not take
the union of all that apply.
Another update that was made was to reflect that only one
exported-services config entry applies to any given service in a
partition. This is a pre-existing constraint that gets enforced by
the Normalize() method on that config entry type.
Proxies dialing exporting services need to know the SPIFFE ID of
services dialed so that the upstream's SANs can be validated.
This commit attaches the SPIFFE ID to all connect proxies exported over
the peering stream so that they are available to importing clusters.
The data in the SPIFFE ID cannot be re-constructed in peer clusters
because the partition of exported services is overwritten on imports.
Signed-off-by: acpana <8968914+acpana@users.noreply.github.com>
Co-authored-by: Chris S. Kim <ckim@hashicorp.com>
Co-authored-by: Freddy <freddygv@users.noreply.github.com>
The importing peer will need to know what SNI and SPIFFE name
corresponds to each exported service. Additionally it will need to know
at a high level the protocol in use (L4/L7) to generate the appropriate
connection pool and local metrics.
For replicated connect synthetic entities we edit the `Connect{}` part
of a `NodeService` to have a new section:
{
"PeerMeta": {
"SNI": [
"web.default.default.owt.external.183150d5-1033-3672-c426-c29205a576b8.consul"
],
"SpiffeID": [
"spiffe://183150d5-1033-3672-c426-c29205a576b8.consul/ns/default/dc/dc1/svc/web"
],
"Protocol": "tcp"
}
}
This data is then replicated and saved as-is at the importing side. Both
SNI and SpiffeID are slices for now until I can be sure we don't need
them for how mesh gateways will ultimately work.
Treat each exported service as a "discovery chain" and replicate one
synthetic CheckServiceNode for each chain and remote mesh gateway.
The health will be a flattened generated check of the checks for that
mesh gateway node.
By changing to use WatchCtx we will actually block for changes to the peering list. WatchCh creates a goroutine to collect errors from WatchCtx and returns immediately.
The existing behavior wouldn't result in a tight loop because of the rate limiting in the surrounding function, but it would still lead to more work than is necessary.
R.B. Boyer