Upgrade open-telemetry/opentelemetry-collector-contrib to v0.89.0

[gebn]

make vendor-check fmt passing

What this PR does: Upgrades the open-telemetry/opentelemetry-collector-contrib dependency to v0.89.0, including transitive dependencies as necessary. This version includes open-telemetry/opentelemetry-collector-contrib#26022, which we would like to benefit from in Tempo. N.B. one code change in cmd/tempo/main.go, due to NewTracer() being removed in go.opentelemetry.io/otel/bridge/[email protected].

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

[CLAassistant]

All committers have signed the CLA.

[gebn]

Seeking approval for the CLA, then I can rebase with my work email. Please let me know of any thoughts in the meantime.

[joe-elliott]

Thanks for the PR. If you can get the CLA signed then we can merge.

[joe-elliott]

Linter issue seems like a simple fix. Tempodb test failure is likely transient. I will re-run tests. However, there does appear to be a panic on startup in integration tests:

13:23:29 tempo: panic: runtime error: invalid memory address or nil pointer dereference
13:23:29 tempo: [signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x207cf8b]
13:23:29 tempo: goroutine 336 [running]:
13:23:29 tempo: go.opentelemetry.io/collector/internal/sharedcomponent.(*SharedComponent[...]).Start.func1()
13:23:29 tempo: /home/runner/work/tempo/tempo/vendor/go.opentelemetry.io/collector/internal/sharedcomponent/sharedcomponent.go:92 +0xcb
13:23:29 tempo: sync.(*Once).doSlow(0xc00093d340?, 0x0?)
13:23:29 tempo: /opt/hostedtoolcache/go/1.21.3/x64/src/sync/once.go:74 +0xbf
13:23:29 tempo: sync.(*Once).Do(...)
13:23:29 tempo: /opt/hostedtoolcache/go/1.21.3/x64/src/sync/once.go:65
13:23:29 tempo: go.opentelemetry.io/collector/internal/sharedcomponent.(*SharedComponent[...]).Start(0x1da[264](https://github.com/grafana/tempo/actions/runs/6898014680/job/18785339327?pr=3148#step:4:265)7?, {0x2f28fe8?, 0xc000a11090?}, {0x2f296b0?, 0xc000268770?}
13:23:29 tempo: )
13:23:29 tempo: /home/runner/work/tempo/tempo/vendor/go.opentelemetry.io/collector/internal/sharedcomponent/sharedcomponent.go:87 +0x9b
13:23:29 tempo: github.com/grafana/tempo/modules/distributor/receiver.(*receiversShim).starting(0xc000[268](https://github.com/grafana/tempo/actions/runs/6898014680/job/18785339327?pr=3148#step:4:269)770, {0x2f28fe8, 0xc000a11090})
13:23:29 tempo: /home/runner/work/tempo/tempo/modules/distributor/receiver/shim.go:293 +0x83
13:23:29 tempo: github.com/grafana/dskit/services.(*BasicService).main(0xc0004bf360)
13:23:29 tempo: /home/runner/work/tempo/tempo/vendor/github.com/grafana/dskit/services/basic_service.go:157 +0x5a
13:23:29 tempo: created by github.com/grafana/dskit/services.(*BasicService).StartAsync.func1 in goroutine 331
13:23:29 tempo: /home/runner/work/tempo/tempo/vendor/github.com/grafana/dskit/services/basic_service.go:119 +0x105
    api_test.go:26: 
        	Error Trace:	/home/runner/work/tempo/tempo/integration/e2e/api_test.go:26
        	Error:      	Received unexpected error:
        	            	docker container tempo failed to start: exit status 1
        	Test:       	TestSearchTagValuesV2

Take a look and see what you can find. Let me know if you have any questions and I can help dig some.

[knylander-grafana]

Thank you for updating the changelog

[gebn]

I've fixed the low-hanging fruit. There's still a SIGSEGV panic in vendor/go.opentelemetry.io/collector/internal/sharedcomponent/sharedcomponent.go:92 when running the e2e tests, however I'm not sure where it's coming from - an if _ == nil before each dereference is not entered.

[joe-elliott]

The nil pointer dereference is b/c ReportComponentStatus is nil on the line referenced above (vendor/go.opentelemetry.io/collector/internal/sharedcomponent/sharedcomponent.go:92).

I can find two places this field is set:

https://github.com/open-telemetry/opentelemetry-collector/blob/main/service/service.go#L125

and

https://github.com/open-telemetry/opentelemetry-collector/blob/0ae738f0c8eebf669967cd1bdda5072929df4b9d/service/internal/servicetelemetry/telemetry_settings.go#L18

Maybe we need to create our receivers through one of these code paths to make sure the field is set correctly?

Running the following will show the issue. Integration tests not required:

go run ./cmd/tempo --storage.trace.backend=local --storage.trace.local.path=/tmp/tempo/traces --storage.trace.wal.path=/tmp/tempo/wal --server.http-listen-port=8080

[gebn]

Thanks @joe-elliott, I didn't realise it was a nil-able field as opposed to a method. I've dug through, and the simple fix is:

diff --git a/modules/distributor/receiver/shim.go b/modules/distributor/receiver/shim.go
index cf9ecd41d..11fc47c8d 100644
--- a/modules/distributor/receiver/shim.go
+++ b/modules/distributor/receiver/shim.go
@@ -228,11 +228,16 @@ func New(receiverCfg map[string]interface{}, pusher TracesPusher, middleware Mid
 
        // todo: propagate a real context?  translate our log configuration into zap?
        ctx := context.Background()
-       params := receiver.CreateSettings{TelemetrySettings: component.TelemetrySettings{
-               Logger:         zapLogger,
-               TracerProvider: tracenoop.NewTracerProvider(),
-               MeterProvider:  metricnoop.NewMeterProvider(),
-       }}
+       params := receiver.CreateSettings{
+               TelemetrySettings: component.TelemetrySettings{
+                       Logger:         zapLogger,
+                       TracerProvider: tracenoop.NewTracerProvider(),
+                       MeterProvider:  metricnoop.NewMeterProvider(),
+                       ReportComponentStatus: func(*component.StatusEvent) error {
+                               return nil
+                       },
+               },
+       }
 
        for componentID, cfg := range conf.Receivers {
                factoryBase := receiverFactories[componentID.Type()]

As you mention, it looks like the OTel Collector authors assumed Service would be used to indirectly create the TelemetrySettings struct. There are no exported abstractions for doing this elsewhere. This struct appears to effectively represent the entire collector - it houses build info, creates a logger from scratch, and uses status changes to signal shutdown. Indeed the whole feature seems to have been introduced purely to ease implementation of the official collector, which is presumably not what Tempo is interested in. It may be desirable to hook Tempo's existing lifecycle management into this callback, however given Tempo currently handles errors correctly, perhaps that can come later.

I've pushed the no-op implementation. Please let me know your thoughts.

[gebn]

Please hold off on merging, I'm still going through approvals for AGPL contribution. I'll take a look at the tempodb tests

[joe-elliott]

Please hold off on merging, I'm still going through approvals for AGPL contribution.

Will do.

I'll take a look at the tempodb tests

@mapno was having this issue yesterday. He solved it by rebasing onto main (vs. merging main into his branch). It's a really weird test failure and he was unable to track down the root cause.

[mapno]

$ git pull --rebase origin main fixed it for me

[gebn]

Rebased off main. I now have approval for contributing, so also switched commit author to my work email 👍

[joe-elliott]

Great update! Thanks.