feat(opentelemetry-node): add cloud/container resource detectors #214

david-luna · 2024-06-06T10:47:14Z

Closes #128

This PR adds the container and the cloud resource detectors (alibaba, aws, azure & gcp). By default all are included unless the env var OTEL_NODE_RESOURCE_DETECTORS is set. I that case the distro includes according the value set in the environment. Some highlights:

resolution of which detectors is moved to a separate module instead like it was done for instrumentations
env var affects to all detectors. So if uses sets a list of detectors our distro detector may be excluded.
looking at @opentelemetry/auto-instrumentations-node there are a couple of detectors not related to cloud/containers left to add

Question
This detectors resolution gives more priority to setup via code rather than the environment setup via environment vars. I wonder if that's the behaviour we want 🤔

Vanilla SDK seems to have this behaviour too (link) and its @opentelemetry/auto-instrumentations-node the one looking up to env var OTEL_NODE_RESOURCE_DETECTORS

david-luna · 2024-06-06T11:40:51Z

Tests are failing since some detectors generate spans. An example is HTTP spans being generated by GCP detector

node -r @elastic/opentelemetry-node ./test/fixtures/use-http-server.js

------ trace 350a53 (2 spans) ------
       span 1edd33 "GET" (10.9ms, STATUS_CODE_ERROR, SPAN_KIND_CLIENT, GET http://127.0.0.1:56695/ -> 404)
  +6ms `- span 059043 "GET" (1.9ms, SPAN_KIND_SERVER, GET http://127.0.0.1:56695/ -> 404)
------ trace 68c04e (1 span) ------
       span 190981 "GET" (6.3ms, STATUS_CODE_ERROR, SPAN_KIND_CLIENT, GET http://metadata.google.internal./computeMetadata/v1/instance)
------ trace f32a41 (2 spans) ------
       span 8af26e "POST" (3.3ms, SPAN_KIND_CLIENT, POST http://127.0.0.1:56695/echo -> 200)
  +2ms `- span 2df652 "POST" (0.8ms, SPAN_KIND_SERVER, POST http://127.0.0.1:56695/echo -> 200)
------ trace 3230c3 (1 span) ------
       span b94337 "GET" (3004.0ms, STATUS_CODE_ERROR, SPAN_KIND_CLIENT, GET http://169.254.169.254/computeMetadata/v1/instance)

ba8681a adds a proposal to filter out these spans when testing. To be discussed

david-luna · 2024-06-06T18:34:01Z

As discussed in today's meeting with @trentm

distro resource detector is added always
check auto-instrumentations if GCP spans are also created and create an issue for that
filter out the GCP spans from the sortedSpans getter in tests
change those warnings from AWS & container detectors in upstream

trentm · 2024-06-07T18:25:56Z

@david-luna This is still draft. I wasn't sure if this was ready for review yet.

david-luna · 2024-06-10T08:39:04Z

@david-luna This is still draft. I wasn't sure if this was ready for review yet.

There was a linting issue happening in CI but not in local so it wasn't ready yet. Last commit contains a fix for gen:types npm script.

david-luna · 2024-06-10T09:37:22Z

Note on the lint failure.

npm ls --all

...
 ├─┬ @opentelemetry/[email protected]
  │ ├── @opentelemetry/[email protected] deduped
  │ ├── @opentelemetry/[email protected] deduped
  │ ├── @opentelemetry/[email protected] deduped
  │ ├── @opentelemetry/[email protected] deduped
  │ └─┬ [email protected]
  │   ├─┬ [email protected]
  │   │ ├── [email protected]
  │   │ ├─┬ [email protected]
  │   │ │ ├─┬ [email protected]
  │   │ │ │ └── [email protected] deduped
  │   │ │ └── [email protected] deduped
  │   │ ├── [email protected] deduped
  │   │ ├─┬ [email protected]
  │   │ │ ├── UNMET OPTIONAL DEPENDENCY encoding@^0.1.0
  │   │ │ └─┬ [email protected]
  │   │ │   ├── [email protected].       <----- THIS VERSION OF `tr46` HAS NOT LICENSE FILE
  │   │ │   └── [email protected]
  │   │ └── [email protected]
  │   └─┬ [email protected]
  │     └── [email protected]
 ...

Linting if failing because of a missing license file, was added in https://github.com/jsdom/tr46/tree/v1.0.0. The package.json file has the field "license" set correctly but I'm not sure if we should add a fallback to that field if the file is missing.

Any thoughts @trentm ?

trentm · 2024-06-10T17:44:26Z

The package.json file has the field "license" set correctly but I'm not sure if we should add a fallback to that field if the file is missing.

I think it is fine to add a fallback in the lint script for this. The package.json "license" field says MIT. The very next commit on the tr46 repo adds the license file that is then included in subsequent published versions.

I think it would be fine to try a PR on node-fetch to update their whatwg-url dep, but I don't think that needs to block us here.

david-luna · 2024-06-10T21:47:41Z

@trentm I just realized ./script/gen-notice.sh already has a fallback mechanism for packages that do not have a license file. I've added a new entry for this fallback in 6888419

trentm · 2024-06-10T23:18:49Z

packages/opentelemetry-node/lib/detectors.js

+    const detectorsFromEnv =
+        process.env['OTEL_NODE_RESOURCE_DETECTORS'] || 'all';
+
+    if (detectorsFromEnv === 'none') {
+        return [];
+    }
+
+    const detectorKeys =
+        detectorsFromEnv === 'all'
+            ? Object.keys(defaultDetectors)
+            : detectorsFromEnv.split(',');


The logic is slightly difference in auto-instr-node: https://github.com/open-telemetry/opentelemetry-js-contrib/blob/dadf30828e7008df7b85c8bd7eaacb90770aac5c/metapackages/auto-instrumentations-node/src/utils.ts#L235-L244

E.g. OTEL_NODE_RESOURCE_DETECTORS=host,os,none with a "none" in there the behaviour differs. Also if 'all' is one of the values in there.

Personally, I think it would be nice to .trim() each value so OTEL_NODE_RESOURCE_DETECTORS=host, os works. auto-instr-node does trim in its handling of process.env.OTEL_NODE_ENABLED_INSTRUMENTATIONS, FWIW.

I've applied your suggestions. Thanks :)

packages/opentelemetry-node/lib/detectors.js

trentm · 2024-06-10T23:21:58Z

packages/opentelemetry-node/test/instr-http.test.js

@@ -76,6 +76,7 @@ const testFixtures = [
            //     ------ trace b8467d (2 spans) ------
            //            span bc8a2c "POST" (3.8ms, SPAN_KIND_CLIENT, POST http://127.0.0.1:64972/echo -> 200)
            //      +2ms `- span 4e7adf "POST" (1.1ms, SPAN_KIND_SERVER, POST http://127.0.0.1:64972/echo -> 200)
+            // const spans = col.sortedSpans.slice(RESOURCE_DETECTOR_SPAN_COUNT);


debugging code?

yup, I've missed that

packages/opentelemetry-node/test/testutils.js

Co-authored-by: Trent Mick <[email protected]>

david-luna · 2024-06-11T09:23:19Z

@trentm I've added some tests since we can inspect the warnings when passing bogus values. With this I've realized the SDK was logging some errors that shouldn't be logged.

I'm hesitant if wee need to add more tests for other scenarios like the behaviour when the value all or none is present in the list. WDYT?

…rsing

…ic/elastic-otel-node into dluna/128-cloud-resource-detectors

trentm · 2024-06-11T15:30:20Z

packages/opentelemetry-node/lib/elastic-node-sdk.js

-                // hostDetectorSync is not currently in the OTel default, but may be added
-                hostDetectorSync,
-            ],
+            resourceDetectors: resolveDetectors(opts.resourceDetectors),


Not for this PR. I wonder if eventually we may want to export something similar to getInstrumentations() that enables a user to user our SDK and get the default set of detectors plus one or more custom detectors. Or perhaps this is something that is more generally solved with some concept of SDK plugins or extensions as briefly discussed on the OTel JS SIG.

I'd say if the number of detectors grow enough we will have a case for that API to be exported. We can discuss it and create an issue if we agree it's necessary

trentm · 2024-06-11T15:35:48Z

I'm hesitant if wee need to add more tests for other scenarios like the behaviour when the value all or none is present in the list. WDYT?

I think we are fine without the tests. Your call if you'd like to add more.

david-luna added 3 commits June 6, 2024 11:54

feat(opentelemetry-node): add cloud resource detectors

019dd24

chore(opentelemetry-node): remove logs

6a04b2d

feat(opentelemetry-node): add container resource detector

46b20e5

david-luna requested a review from trentm June 6, 2024 10:47

david-luna and others added 2 commits June 6, 2024 12:47

Merge branch 'main' into dluna/128-cloud-resource-detectors

6be80b3

chore(opentelemetry-node): fix linting

3b5f096

test(opentelementry-node): filter of resource detector spans proposal

ba8681a

david-luna added 2 commits June 7, 2024 11:32

chore(opentelementry-node): add distro detector with any configuration

d9506af

chore(opentelementry-node): update types

d8175df

chore(opentelementry-node): fix gen:types script

a3c36c6

david-luna marked this pull request as ready for review June 10, 2024 08:39

chore(opentelemetry-node): add fallback license check fro tr46 package

6888419

Merge branch 'main' into dluna/128-cloud-resource-detectors

1bb59e7

trentm reviewed Jun 10, 2024

View reviewed changes

david-luna and others added 2 commits June 11, 2024 09:41

Update packages/opentelemetry-node/lib/detectors.js

9783c25

Co-authored-by: Trent Mick <[email protected]>

chore(opentelemetry-node): update env var parsing for resource detectors

726c816

david-luna mentioned this pull request Jun 11, 2024

Remove span filter for GCP resource detector HTTP calls in tests #222

Open

Update packages/opentelemetry-node/test/testutils.js

f396957

Co-authored-by: Trent Mick <[email protected]>

david-luna requested a review from trentm June 11, 2024 09:21

david-luna added 3 commits June 11, 2024 15:24

chore(opentelemetry-node): add tests for resource detectors envvar pa…

12b6ed8

…rsing

Merge branch 'dluna/128-cloud-resource-detectors' of github.com:elast…

dd7d848

…ic/elastic-otel-node into dluna/128-cloud-resource-detectors

chore(opentelemetry-node): fix linting issues

0fd33e9

chore(opentelemetry-node): update types

da8643f

trentm approved these changes Jun 11, 2024

View reviewed changes

Merge branch 'main' into dluna/128-cloud-resource-detectors

73db47b

david-luna merged commit f89bf32 into main Jun 11, 2024
12 checks passed

david-luna deleted the dluna/128-cloud-resource-detectors branch June 11, 2024 17:16

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(opentelemetry-node): add cloud/container resource detectors #214

feat(opentelemetry-node): add cloud/container resource detectors #214

david-luna commented Jun 6, 2024

david-luna commented Jun 6, 2024 •

edited

Loading

david-luna commented Jun 6, 2024 •

edited

Loading

trentm commented Jun 7, 2024

david-luna commented Jun 10, 2024

david-luna commented Jun 10, 2024

trentm commented Jun 10, 2024

david-luna commented Jun 10, 2024

trentm Jun 10, 2024

david-luna Jun 11, 2024

trentm Jun 10, 2024

david-luna Jun 11, 2024

david-luna commented Jun 11, 2024 •

edited

Loading

trentm Jun 11, 2024

david-luna Jun 11, 2024 •

edited

Loading

trentm commented Jun 11, 2024

feat(opentelemetry-node): add cloud/container resource detectors #214

feat(opentelemetry-node): add cloud/container resource detectors #214

Conversation

david-luna commented Jun 6, 2024

david-luna commented Jun 6, 2024 • edited Loading

david-luna commented Jun 6, 2024 • edited Loading

trentm commented Jun 7, 2024

david-luna commented Jun 10, 2024

david-luna commented Jun 10, 2024

trentm commented Jun 10, 2024

david-luna commented Jun 10, 2024

trentm Jun 10, 2024

Choose a reason for hiding this comment

david-luna Jun 11, 2024

Choose a reason for hiding this comment

trentm Jun 10, 2024

Choose a reason for hiding this comment

david-luna Jun 11, 2024

Choose a reason for hiding this comment

david-luna commented Jun 11, 2024 • edited Loading

trentm Jun 11, 2024

Choose a reason for hiding this comment

david-luna Jun 11, 2024 • edited Loading

Choose a reason for hiding this comment

trentm commented Jun 11, 2024

david-luna commented Jun 6, 2024 •

edited

Loading

david-luna commented Jun 6, 2024 •

edited

Loading

david-luna commented Jun 11, 2024 •

edited

Loading

david-luna Jun 11, 2024 •

edited

Loading