This PR adds cluster members to the metrics API. The number of members per
segment are reported as well as the total number of members.
Tested by running a multi-node cluster locally and ensuring the numbers were
correct. Also added unit test coverage to add the new expected gauges to
existing test cases.
We have seen test flakes caused by 'concurrent map read and map write', and the race detector
reports the problem as well (prevent us from running some tests with -race).
The root of the problem is the grpc expects resolvers to be registered at init time
before any requests are made, but we were using a separate resolver for each test.
This commit introduces a resolver registry. The registry is registered as the single
resolver for the consul scheme. Each test uses the Authority section of the target
(instead of the scheme) to identify the resolver that should be used for the test.
The scheme is used for lookup, which is why it can no longer be used as the unique
key.
This allows us to use a lock around the map of resolvers, preventing the data race.
* fix tests to use a dummy nodeName and not fail when hostname is not a valid nodeName
* remove conditional testing
* add test when node name is invalid
* debug: remove the CLI check for debug_enabled
The API allows collecting profiles even debug_enabled=false as long as
ACLs are enabled. Remove this check from the CLI so that users do not
need to set debug_enabled=true for no reason.
Also:
- fix the API client to return errors on non-200 status codes for debug
endpoints
- improve the failure messages when pprof data can not be collected
Co-Authored-By: Dhia Ayachi <dhia@hashicorp.com>
* remove parallel test runs
parallel runs create a race condition that fail the debug tests
* Add changelog
Co-authored-by: Daniel Nephin <dnephin@hashicorp.com>
The bulk of this commit is moving the LeaderRoutineManager from the agent/consul package into its own package: lib/gort. It also got a renaming and its Start method now requires a context. Requiring that context required updating a whole bunch of other places in the code.
This refactor is to make it easier to see how serf feature flags are
encoded as serf tags, and where those feature flags are read.
- use constants for both the prefix and feature flag name. A constant
makes it much easier for an IDE to locate the read and write location.
- isolate the feature-flag encoding logic in the metadata package, so
that the feature flag prefix can be unexported. Only expose a function
for encoding the flags into tags. This logic is now next to the logic
which reads the tags.
- remove the duplicate `addEnterpriseSerfTags` functions. Both Client
and Server structs had the same implementation. And neither
implementation needed the method receiver.
The prior solution to call reply.Reset() aged poorly since newer fields
were added to the reply, but not added to Reset() leading serial
blocking query loops on the server to blend replies.
This could manifest as a service-defaults protocol change from
default=>http not reverting back to default after the config entry
reponsible was deleted.
* Save exposed HTTP or GRPC ports to the agent's store
* Add those the health checks API so we can retrieve them from the API
* Change redirect-traffic command to also exclude those ports from inbound traffic redirection when expose.checks is set to true.
TestACLEndpoint_Login_with_TokenLocality was reguardly being reported as failed even though
it was not failing. I took another look and I suspect it is because t.Parllel was being
called in a goroutine.
This would lead to strange behaviour which apparently confused the 'go test' runner.
- return errors in TestAgent.Start so that the retry works correctly
- remove duplicate logging, the error is returned already
- add a missing t.Helper() to retry.Run
- properly set a.Agent to nil so that subsequent retry attempts will actually try to start
Also accept an RPCInfo instead of interface{}. Accepting an interface
lead to a bug where the caller was expecting the arg to be the response
when in fact it was always passed the request. By accepting RPCInfo
it should indicate that this is actually the request value.
One caller of canRetry already passed an RPCInfo, the second handles
the type assertion before calling canRetry.
A recent change in 1.9.x inverted the order of these two lines, which caused the
X-Consul-Effective-Consistency header to be missing for the servie health endpoints
Only default to the user token and agent token for the sync. Change the
exported methods to only return the stored tokens associated with a
specific check or service.
Also fixes a bug with listing kind=mesh config entries. ValidateConfigEntryKind was only being used by
the List endpoint, and was yet another place where we have to enumerate all the kinds.
This commit removes ValidateConfigEntryKind and uses MakeConfigEntry instead. This change removes
the need to maintain two separate functions at the cost of creating an instance of the config entry which will be thrown away immediately.
Co-authored-by: R.B. Boyer <4903+rboyer@users.noreply.github.com>
Previously we would associate the address of a discovery chain target
with the discovery chain's filter chain. This was broken for a few reasons:
- If the upstream is a virtual service, the client proxy has no way of
dialing it because virtual services are not targets of their discovery
chains. The targets are distinct services. This is addressed by watching
the endpoints of all upstream services, not just their discovery chain
targets.
- If multiple discovery chains resolve to the same target, that would
lead to multiple filter chains attempting to match on the target's
virtual IP. This is addressed by only matching on the upstream's virtual
IP.
NOTE: this implementation requires an intention to the redirecting
virtual service and not just to the final destination. This is how
we can know that the virtual service is an upstream to watch.
A later PR will look into traversing discovery chains when computing
upstreams so that intentions are only required to the discovery chain
targets.
* WIP reloadable raft config
* Pre-define new raft gauges
* Update go-metrics to change gauge reset behaviour
* Update raft to pull in new metric and reloadable config
* Add snapshot persistance timing and installSnapshot to our 'protected' list as they can be infrequent but are important
* Update telemetry docs
* Update config and telemetry docs
* Add note to oldestLogAge on when it is visible
* Add changelog entry
* Update website/content/docs/agent/options.mdx
Co-authored-by: Matt Keeler <mkeeler@users.noreply.github.com>
Co-authored-by: Matt Keeler <mkeeler@users.noreply.github.com>
* Give descriptive error if auth method not found
Previously during a `consul login -method=blah`, if the auth method was not found, the
error returned would be "ACL not found". This is potentially confusing
because there may be many different ACLs involved in a login: the ACL of
the Consul client, perhaps the binding rule or the auth method.
Now the error will be "auth method blah not found", which is much easier
to debug.
Initially we were loading every potential upstream address into Envoy
and then routing traffic to the logical upstream service. The downside
of this behavior is that traffic meant to go to a specific instance
would be load balanced across ALL instances.
Traffic to specific instance IPs should be forwarded to the original
destination and if it's a destination in the mesh then we should ensure
the appropriate certificates are used.
This PR makes transparent proxying a Kubernetes-only feature for now
since support for other environments requires generating virtual IPs,
and Consul does not do that at the moment.
No config entry needs a Kind field. It is only used to determine the Go type to
target. As we introduce new config entries (like this one) we can remove the kind field
and have the GetKind method return the single supported value.
In this case (similar to proxy-defaults) the Name field is also unnecessary. We always
use the same value. So we can omit the name field entirely.
The only thing that needed fixing up pertained to this section of the 1.18.x release notes:
> grpc_stats: the default value for stats_for_all_methods is switched from true to false, in order to avoid possible memory exhaustion due to an untrusted downstream sending a large number of unique method names. The previous default value was deprecated in version 1.14.0. This only changes the behavior when the value is not set. The previous behavior can be used by setting the value to true. This behavior change by be overridden by setting runtime feature envoy.deprecated_features.grpc_stats_filter_enable_stats_for_all_methods_by_default.
For now to maintain status-quo I'm explicitly setting `stats_for_all_methods=true` in all versions to avoid relying upon the default.
Additionally the naming of the emitted metrics for these gRPC requests changed slightly so the integration test assertions for `case-grpc` needed adjusting.
This ensures that if someone does include some extension Consul does not currently make use of, that extension is actually usable. Without linking these envoy protobufs into the main binary it can't round trip the escape hatches to send them down to envoy.
Whenenver the go-control-plane library is upgraded next we just have to re-run 'make envoy-library'.
This adds support for the Incremental xDS protocol when using xDS v3. This is best reviewed commit-by-commit and will not be squashed when merged.
Union of all commit messages follows to give an overarching summary:
xds: exclusively support incremental xDS when using xDS v3
Attempts to use SoTW via v3 will fail, much like attempts to use incremental via v2 will fail.
Work around a strange older envoy behavior involving empty CDS responses over incremental xDS.
xds: various cleanups and refactors that don't strictly concern the addition of incremental xDS support
Dissolve the connectionInfo struct in favor of per-connection ResourceGenerators instead.
Do a better job of ensuring the xds code uses a well configured logger that accurately describes the connected client.
xds: pull out checkStreamACLs method in advance of a later commit
xds: rewrite SoTW xDS protocol tests to use protobufs rather than hand-rolled json strings
In the test we very lightly reuse some of the more boring protobuf construction helper code that is also technically under test. The important thing of the protocol tests is testing the protocol. The actual inputs and outputs are largely already handled by the xds golden output tests now so these protocol tests don't have to do double-duty.
This also updates the SoTW protocol test to exclusively use xDS v2 which is the only variant of SoTW that will be supported in Consul 1.10.
xds: default xds.Server.AuthCheckFrequency at use-time instead of construction-time
This config entry is being renamed primarily because in k8s the name
cluster could be confusing given that the config entry applies across
federated datacenters.
Additionally, this config entry will only apply to Consul as a service
mesh, so the more generic "cluster" name is not needed.
Previous getFromView would call view.Result when the result may not have been returned
(because the index is updated past the minIndex. This would allocate a slice and sort it
for no reason, because the values would never be returned.
Fix this by re-ordering the operations in getFromView.
The test changes in this commit were an attempt to cover the case where
an update is received but the index does not exceed the minIndex.
Also rename it to readEntry now that it doesn't return the entire entry. Based on feedback
in PR review, the full entry is not used by the caller, and accessing the fields wouldn't be
safe outside the lock, so it is safer to return only the Materializer
grpclog.SetLoggerV2 is meant to be called only once before any gRPC requests are received, but
each test that uses TestAgent will call NewBaseDeps again. Use a sync.Once to prevent the grpc
logging from being re-initialized by each test.
This will mean that a test can't use a fake logger to capture logs from the gRPC server.
These tests can flake when we get a notification for an earlier event.
Retry the read from update channel a few times to make sure we get the
event we expect.
Split the TestStreamingClient into the two logical components the real
client uses. This allows us to test multiple clients properly.
Previously writing of ctx from multiple Subscribe calls was showing a
data race.
Once this was fixed a test started to fail because the request had to be
made with a greater index, so that the store.Get call did not return
immediately.
The idleTTL was being written and read concurrently. Instead move the idleTTL to a struct
field so that when one test patches the TTL it does not impact others.
The background goroutines for the store can outlive a test because context cancellation
is async.
Setting this field to a value is equivalent to using the 'near' query paramter.
The intent is to sort the results by proximity to the node requesting
them. However with connect we send the results to envoy, which doesn't
care about the order, so setting this field is increasing the work
performed for no gain.
It is necessary to unset this field now because we would like connect
to use streaming, but streaming does not support sorting by proximity.
Remove View.Result error return value, it was always nil, and seems like it will likely always remain nill
since it is simply reading a stored value.
Also replace some cache types with local types.
And only start expiration time when the last request ends. This makes tracking expiry simpler, and
ensures that no entry can be expired while there are active requests.
Previously canRetry was attempting to retrieve this error from args, however there was never
any callers that would pass an error to args.
With the change to raftApply to move this error to the error return value, it is now possible
to receive this error from the err argument.
This commit updates canRetry to check for ErrChunkingResubmit in err.
Previously we were inconsistently checking the response for errors. This
PR moves the response-is-error check into raftApply, so that all callers
can look at only the error response, instead of having to know that
errors could come from two places.
This should expose a few more errors that were previously hidden because
in some calls to raftApply we were ignoring the response return value.
Also handle errors more consistently. In some cases we would log the
error before returning it. This can be very confusing because it can
result in the same error being logged multiple times. Instead return
a wrapped error.
Previously only a single auth method would be saved to the snapshot. This commit fixes the typo
and adds to the test, to show that all auth methods are now saved.
This way we avoid serializing these when empty. Otherwise users of the
latest version of the api submodule cannot interact with older versions
of Consul, because a new api client would send keys that the older Consul
doesn't recognize yet.
Paul Ewing