chore: update core dependencies to 0.25

[dyladan]

What it says on the tin

[codecov]

Codecov Report

Merging #635 (8c0d32f) into main (c25bc38) will decrease coverage by 1.81%.
The diff coverage is 100.00%.

❗ Current head 8c0d32f differs from pull request most recent head a7ba541. Consider uploading reports for the commit a7ba541 to get more accurate results

@@            Coverage Diff             @@
##             main     #635      +/-   ##
==========================================
- Coverage   96.68%   94.87%   -1.82%     
==========================================
  Files          13      208     +195     
  Lines         634    12128   +11494     
  Branches      124     1155    +1031     
==========================================
+ Hits          613    11506   +10893     
- Misses         21      622     +601     

Impacted Files	Coverage Δ
...njection/src/instrumentation/WebInstrumentation.ts	97.22% <ø> (ø)
...ages/opentelemetry-host-metrics/src/BaseMetrics.ts	57.57% <ø> (ø)
...ges/opentelemetry-host-metrics/test/metric.test.ts	96.89% <ø> (ø)
...metry-instrumentation-graphql/test/graphql.test.ts	100.00% <ø> (ø)
...entelemetry-instrumentation-graphql/test/helper.ts	100.00% <ø> (ø)
...e/opentelemetry-instrumentation-pg/test/pg.test.ts	95.01% <ø> (ø)
...pentelemetry-instrumentation-pg/test/utils.test.ts	98.36% <ø> (ø)
...tension-autoinjection/test/instrumentation.test.ts	100.00% <100.00%> (ø)
...ntelemetry-instrumentation-pg/test/pg-pool.test.ts	91.42% <100.00%> (ø)
...opentelemetry-resource-detector-aws/src/version.ts	0.00% <0.00%> (ø)
... and 194 more

[dyladan]

Anyone know why this is failing? i'm not really sure

[Flarna]

Best guess is too high load because of parallel testing. I see no other reason for e.g. "before all" hook in "WinstonInstrumentation" not finishing within 2000ms.

[dyladan]

Best guess is too high load because of parallel testing. I see no other reason for e.g. "before all" hook in "WinstonInstrumentation" not finishing within 2000ms.

I agree. I can't reproduce it locally either.

[dyladan]

Trying to limit concurrency to 4 (there are 2 cpus on the runners) to see if that helps. Probably we haven't seen this yet because we haven't had an update which touches so many packages.

[dyladan]

Seems like the lambda failure is a legitimate failure

[dyladan]

@willarmiros the AWS lambda tests are consistently failing and they pass locally. Any idea what might be up?

[willarmiros]

Taking a look...seems like the problem is that the spans aren't getting exported. Were there any changes to the memory exporter or force flush?

[willarmiros]

I can see no relevant changes were made, and that wouldn't really make sense since the tests are passing locally anyway. I can see that there many other tests that are using the in-memory exporter and then verifying the content of those exported spans. One difference is that we're using the BatchSpanProcessor and all others that I've seen are using the SimpleSpanProcessor, so maybe there's something to that

[dyladan]

Taking a look...seems like the problem is that the spans aren't getting exported. Were there any changes to the memory exporter or force flush?

Not that I'm aware of

[dyladan]

Quick perusal of the release notes from 0.25 shows only this that might have made a difference?

open-telemetry/opentelemetry-js#2396

[willarmiros]

Yeah I saw that, don't think that's the culprit because it should have equally impacted the tests that use SimpleSpanProcessor... it could be something weird with the GitHub runner environment, just really not sure why we'd only see it with these tests on this particular PR.

[willarmiros]

@anuraaga do you have any ideas?

[anuraaga]

@willarmiros Since we don't have package-lock.json is it possible the dependencies aren't at the same version when run locally as what's running in CI? That's what comes to mind as a possible reason for not being able to reproduce.

We have this that relies on some weird behavior of the patching infrastructure but wonder if an update changed it.

https://github.com/open-telemetry/opentelemetry-js-contrib/blob/main/plugins/node/opentelemetry-instrumentation-aws-lambda/src/instrumentation.ts#L109

[dyladan]

@anuraaga @willarmiros I wonder if it has to do with the hoisting? Dependencies are hoisted in CI to make the CI run faster.