Fix ClusterLoadAssignment timeouts dropping endpoints.
When a large number of upstreams are configured on a single envoy
proxy, there was a chance that it would timeout when waiting for
ClusterLoadAssignments. While this doesn't always immediately cause
issues, consul-dataplane instances appear to consistently drop
endpoints from their configurations after an xDS connection is
re-established (the server dies, random disconnect, etc).
This commit adds an `xds_fetch_timeout_ms` config to service registrations
so that users can set the value higher for large instances that have
many upstreams. The timeout can be disabled by setting a value of `0`.
This configuration was introduced to reduce the risk of causing a
breaking change for users if there is ever a scenario where endpoints
would never be received. Rather than just always blocking indefinitely
or for a significantly longer period of time, this config will affect
only the service instance associated with it.
Fix xDS missing endpoint race condition.
This fixes the following race condition:
- Send update endpoints
- Send update cluster
- Recv ACK endpoints
- Recv ACK cluster
Prior to this fix, it would have resulted in the endpoints NOT existing in
Envoy. This occurred because the cluster update implicitly clears the endpoints
in Envoy, but we would never re-send the endpoint data to compensate for the
loss, because we would incorrectly ACK the invalid old endpoint hash. Since the
endpoint's hash did not actually change, they would not be resent.
The fix for this is to effectively clear out the invalid pending ACKs for child
resources whenever the parent changes. This ensures that we do not store the
child's hash as accepted when the race occurs.
An escape-hatch environment variable `XDS_PROTOCOL_LEGACY_CHILD_RESEND` was
added so that users can revert back to the old legacy behavior in the event
that this produces unknown side-effects.
This bug report and fix was mostly implemented by @ksmiley with some minor
tweaks.
Co-authored-by: Derek Menteer <derek.menteer@hashicorp.com>
Co-authored-by: Keith Smiley <ksmiley@salesforce.com>
backport of commit af6045cdf1bfe8589d4ff5f03ffd327a39b62ced
Co-authored-by: Ronald Ekambi <ronekambi@gmail.com>
Co-authored-by: Ronald <roncodingenthusiast@users.noreply.github.com>
* Default "stats_flush_interval" to 1 minute for Consul Telemetry Collector (#19663)
* Set default of 1m for StatsFlushInterval when the collector is setup
* Add documentation on the stats_flush_interval value
* Do not default in two conditions 1) preconfigured sinks exist 2) preconfigured flush interval exists
* Fix wording of docs
* Add changelog
* Fix docs
(cherry picked from commit bfb3a43648634e117c8405d6a6e3cc837fb8e3dc)
* fix pr labeller
---------
Co-authored-by: Ashvitha <ashvitha.sridharan@hashicorp.com>
* backport of commit 06507fe053f8777b12c8a0a54e062df1fb111a52
* backport of commit 14e160573dbc9d644cb045af7eda5da0d03447d0
* backport of commit 088ec70f902e79a101669fbb76830fe356b23982
---------
Co-authored-by: Jared Kirschner <85913323+jkirschner-hashicorp@users.noreply.github.com>
* [Cloud][CC-6925] Updates to pushing server state (#19682)
* Upgrade hcp-sdk-go to latest version v0.73
Changes:
- go get github.com/hashicorp/hcp-sdk-go
- go mod tidy
* From upgrade: regenerate protobufs for upgrade from 1.30 to 1.31
Ran: `make proto`
Slack: https://hashicorp.slack.com/archives/C0253EQ5B40/p1701105418579429
* From upgrade: fix mock interface implementation
After upgrading, there is the following compile error:
cannot use &mockHCPCfg{} (value of type *mockHCPCfg) as "github.com/hashicorp/hcp-sdk-go/config".HCPConfig value in return statement: *mockHCPCfg does not implement "github.com/hashicorp/hcp-sdk-go/config".HCPConfig (missing method Logout)
Solution: update the mock to have the missing Logout method
* From upgrade: Lint: remove usage of deprecated req.ServerState.TLS
Due to upgrade, linting is erroring due to usage of a newly deprecated field
22:47:56 [consul]: make lint
--> Running golangci-lint (.)
agent/hcp/testing.go:157:24: SA1019: req.ServerState.TLS is deprecated: use server_tls.internal_rpc instead. (staticcheck)
time.Until(time.Time(req.ServerState.TLS.CertExpiry)).Hours()/24,
^
* From upgrade: adjust oidc error message
From the upgrade, this test started failing:
=== FAIL: internal/go-sso/oidcauth TestOIDC_ClaimsFromAuthCode/failed_code_exchange (re-run 2) (0.01s)
oidc_test.go:393: unexpected error: Provider login failed: Error exchanging oidc code: oauth2: "invalid_grant" "unexpected auth code"
Prior to the upgrade, the error returned was:
```
Provider login failed: Error exchanging oidc code: oauth2: cannot fetch token: 401 Unauthorized\nResponse: {\"error\":\"invalid_grant\",\"error_description\":\"unexpected auth code\"}\n
```
Now the error returned is as below and does not contain "cannot fetch token"
```
Provider login failed: Error exchanging oidc code: oauth2: "invalid_grant" "unexpected auth code"
```
* Update AgentPushServerState structs with new fields
HCP-side changes for the new fields are in:
https://github.com/hashicorp/cloud-global-network-manager-service/pull/1195/files
* Minor refactor for hcpServerStatus to abstract tlsInfo into struct
This will make it easier to set the same tls-info information to both
- status.TLS (deprecated field)
- status.ServerTLSMetadata (new field to use instead)
* Update hcpServerStatus to parse out information for new fields
Changes:
- Improve error message and handling (encountered some issues and was confused)
- Set new field TLSInfo.CertIssuer
- Collect certificate authority metadata and set on TLSInfo.CertificateAuthorities
- Set TLSInfo on both server.TLS and server.ServerTLSMetadata.InternalRPC
* Update serverStatusToHCP to convert new fields to GNM rpc
* Add changelog
* Feedback: connect.ParseCert, caCerts
* Feedback: refactor and unit test server status
* Feedback: test to use expected struct
* Feedback: certificate with intermediate
* Feedback: catch no leaf, remove expectedErr
* Feedback: update todos with jira ticket
* Feedback: mock tlsConfigurator
* Run make proto for files in 1.16 not in main
* update licensing per 1.16 licensing
* [NET-5688] APIGateway UI Topology Fixes (#19657)
* Update catalog and ui endpoints to show APIGateway in gateway service
topology view
* Added initial implementation for service view
* updated ui
* Fix topology view for gateways
* Adding tests for gw controller
* remove unused args
* Undo formatting changes
* Fix call sites for upstream/downstream gw changes
* Add config entry tests
* Fix function calls again
* Move from ServiceKey to ServiceName, cleanup from PR review
* Add additional check for length of services in bound apigateway for
IsSame comparison
* fix formatting for proto
* gofmt
* Add DeepCopy for retrieved BoundAPIGateway
* gofmt
* gofmt
* Rename function to be more consistent
* Remove busl license
docs: Fix Kubernetes CRD example configs (#18878)
Fixes configuration examples for several Consul Kubernetes CRDs. The
CRDs were missing required fields such as `apiVersion`, `metadata`,
and `spec`.
Co-authored-by: Tu Nguyen <im2nguyen@gmail.com>
Add grpc keepalive configuration. (#19339)
Prior to the introduction of this configuration, grpc keepalive messages were
sent after 2 hours of inactivity on the stream. This posed issues in various
scenarios where the server-side xds connection balancing was unaware that envoy
instances were uncleanly killed / force-closed, since the connections would
only be cleaned up after ~5 minutes of TCP timeouts occurred. Setting this
config to a 30 second interval with a 20 second timeout ensures that at most,
it should take up to 50 seconds for a dead xds connection to be closed.
