feat(instrumentation-winston): add log sending to Logs API #1837

hectorhdzg · 2023-11-30T00:18:54Z

Winston instrumentation will emit recordLog using Logs API whenever configured on, adding similar configuration options as the ones added for Bunyan in #1713

…g/opentelemetry-js-contrib into hectorhdzg/winstonlogs

codecov · 2023-11-30T00:27:31Z

Codecov Report

Merging #1837 (5e0291e) into main (dfb2dff) will increase coverage by 0.00%.
Report is 12 commits behind head on main.
The diff coverage is 82.71%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1837   +/-   ##
=======================================
  Coverage   90.97%   90.98%           
=======================================
  Files         146      145    -1     
  Lines        7492     7454   -38     
  Branches     1502     1488   -14     
=======================================
- Hits         6816     6782   -34     
+ Misses        676      672    -4

Files	Coverage Δ
packages/winston-transport/src/utils.ts	`100.00% <100.00%> (ø)`
.../winston-transport/src/OpenTelemetryTransportV3.ts	`93.75% <93.75%> (ø)`
...try-instrumentation-winston/src/instrumentation.ts	`69.02% <74.00%> (+11.41%)`	⬆️

... and 4 files with indirect coverage changes

plugins/node/opentelemetry-instrumentation-winston/README.md

seemk · 2023-12-01T09:40:47Z

plugins/node/opentelemetry-instrumentation-winston/src/instrumentation.ts

+        attributes[key] = splat[key];
+      }
+    }
+    const timestamp = Date.now();


In the logs API the comment states that timestamp should be in nanoseconds.

The spec also specifies nanoseconds

I wonder if the logs API timestamp field should be a HrTime instead?

Yeah that make sense, updated to use hrTime and convert it to nanoseconds

I thought the LogRecord API allows for HrTime to be passed in, but looks like the types don't support it yet.

I've opened a PR for it: open-telemetry/opentelemetry-js#4345

plugins/node/opentelemetry-instrumentation-winston/README.md

mhennoch · 2023-12-05T19:10:23Z

plugins/node/opentelemetry-instrumentation-winston/src/instrumentation.ts

+      VERSION,
+      Object.assign({}, DEFAULT_CONFIG, config)
+    );
+    this._logger = logs.getLogger(


I wonder if this works with the example in readme. sdk.start() sets the global log provider but this constructor is ran before it so this._logger will always be noop no?

You are correct, this is related to open-telemetry/opentelemetry-js#3622, same issue apply to LoggerProvider now, I can take care of it in open-telemetry/opentelemetry-js#4314, in the meantime will update the documentation to use registerInstrumentations instead of NodeSDK

blumamir

Thanks for working on this 🥇

Overall, LGTM. Added a few minor comments.

blumamir · 2023-11-30T14:52:24Z

plugins/node/opentelemetry-instrumentation-winston/package.json

@@ -41,7 +41,8 @@
    "access": "public"
  },
  "peerDependencies": {
-    "@opentelemetry/api": "^1.3.0"
+    "@opentelemetry/api": "^1.3.0",
+    "@opentelemetry/api-logs": "^0.45.1"


Since it is also installed as a regular dependency, I think we should not list it as peer dependency.
This is also what bunyan is currently doing.

Now that I think about it again, perhaps they should both use peer dependency to be consistent with "@opentelemetry/api" usage?

Well I believe we should update Bunyan as well, I can create a different PR so people are aware specific changes are happening for that instrumentation, some issues in the readme that were brough up by @mhennoch should apply to bunyan as well

plugins/node/opentelemetry-instrumentation-winston/src/instrumentation.ts

blumamir · 2023-12-06T07:42:43Z

plugins/node/opentelemetry-instrumentation-winston/src/instrumentation.ts

+        attributes[key] = splat[key];
+      }
+    }
+    const timestamp = hrTimeToNanoseconds(hrTime());


I wonder if it makes sense for each log record producer to set theobservedTimestamp and timestamp upon creation.

timestamp

From the specs it reads that:

Field: Timestamp
Description: Time when the event occurred measured by the origin clock, i.e. the time at the source. This field is optional, it may be missing if the source timestamp is unknown.

Since the timestamp is known, I would expect that we set timestamp in LogRecord below.

Regardless of this PR, it seems that timestamp is optional both in the data model spec and in otlp, but is not optional in the sdk where it is null coalesed into now when unset. Is this behaviour following the spec?

observedTimestamp

The spec for data model says:

For events that originate in OpenTelemetry (e.g. using OpenTelemetry Logging SDK) this timestamp is typically set at the generation time and is equal to Timestamp.

So I guess we should not set it when calling the API? or set both timestamp and observedTimestamp to the same value in the logRecord?

Currently each time is captured at a different location thus they will probably not be equal, right?

I believe it make sense to send the timestamp if the actual log had some information about it, not applicable in this case, I'm updating this to not send time parameter and let the SDK handle it.

blumamir · 2023-12-06T08:15:00Z

plugins/node/opentelemetry-instrumentation-winston/src/instrumentation.ts

+            let msgArguments = 0;
+            for (let i = 1; i < args.length; i++) {
+              if (typeof args[i] === 'string') {
+                msgArguments++;
+              }
+            }


nit: You can transform this code to a no-side-effects, shorter version like so:

Suggested change

let msgArguments = 0;

for (let i = 1; i < args.length; i++) {

if (typeof args[i] === 'string') {

msgArguments++;

}

}

const msgArguments = args.slice(1).filter(a => typeof a === 'string').length

or (if the first argument is always string):

Suggested change

let msgArguments = 0;

for (let i = 1; i < args.length; i++) {

if (typeof args[i] === 'string') {

msgArguments++;

}

}

const msgArguments = args.filter(a => typeof a === 'string').length - 1

I updated this code a little bit, so it doesn't apply anymore, let me know if some other code can be transformed as well.

blumamir · 2023-12-06T08:24:01Z

plugins/node/opentelemetry-instrumentation-winston/src/instrumentation.ts

+                const values = args.slice(2, msgArguments + 1);
+                logRecord['msg'] = util.format(args[1], ...values);


I didn't run it myself, but it seems winston describes that this behavior is not always applied, and if it does, the arguments might be non-string:

The log method provides the string interpolation using util.format. It must be enabled using format.splat().

and:

logger.log('info', 'test message %d', 123);

Updated the code, util.format support multiple types so passing them as they are now, FYI this code is only applicable to Winston 2 https://github.com/winstonjs/winston/tree/2.x, string interpolation in newer version is handled before write is called.

blumamir · 2023-12-06T08:29:56Z

plugins/node/opentelemetry-instrumentation-winston/src/instrumentation.ts

+            for (let i = args.length - 1; i >= 0; i--) {
+              if (typeof args[i] === 'object') {
+                logRecord['meta'] = args[i];
+                break;
+              }
+            }


you can do it in one go like so:

logRecord['meta'] = args.findLast(a => typeof a === 'object')

I would need to cast args to an array to use that, I usually don't risk casting for args out of my control, also added some other logic to calculate the index to quickly determine the string interpolation parameters.

plugins/node/opentelemetry-instrumentation-winston/src/instrumentation.ts

blumamir · 2023-12-06T08:39:19Z

plugins/node/opentelemetry-instrumentation-winston/src/instrumentation.ts

-            injectRecord(spanContext, record);
-            instrumentation._callHook(span, record);
-            return original.apply(this, args);
+              args.splice(insertAt, 0, record);


we are manipulating the args here, and then use the updated value below for the log bridge. Is it ok?

Well the span information would be duplicated in this case I guess if Logger is also adding correlation information into the LogRecords

martinkuba · 2023-12-09T18:58:48Z

I think it would be better to implement this as a Winston transport rather than an instrumentation, similar to how it has been implemented for Bunyan as a stream. The Logs Bridge API was intended for log appenders, not instrumentations.

hectorhdzg · 2023-12-11T18:17:27Z

@martinkuba my initial approach was using Winston transports, and the problem I found was related to having a direct dependency to winston 3x and 2x to make it work, transports require to extend a specific class in winston is not just to follow some interface, so that would mean people installing @opentelemetry-instrumentation-winston will get winston 3x and winston 2x installed as well, I believe this is undesirable and that is why I decided to use existing patch in write/log methods instead. Related to log appenders or instrumentations, there was some discussion about this, we can create another new appender package just for Winston, Bunyan and others, but I believe the agreement was to use the instrumentation instead, I don't have a problem creating a new appender project here, but is just some maintenance overhead in my opinion and possible causing users confusion.

martinkuba · 2023-12-11T20:12:27Z

@hectorhdzg That makes sense. I just wanted to highlight that the Logs API was not intended to be used in instrumentations (per my interpretation of the spec). Maybe it's fine to approach it this way, but I am also thinking in terms of allowing (or encouraging) instrumentations to generate logs. The PR linked below is related.
open-telemetry/opentelemetry-js#4314

I see that Java has different packages for different versions of Log4j appenders
https://github.com/open-telemetry/opentelemetry-java-instrumentation/tree/main/instrumentation/log4j

hectorhdzg · 2023-12-11T22:20:05Z

@martinkuba I understand appender term makes sense as is what the specification mentions, I created a discussion issue to talk about this in a single place, as this conversation has been happening in several places. I don't have strong opinions about one or the other, I just want to allow this functionality and move this forward.
open-telemetry/opentelemetry-js#4355

This PR updates the telemetry subsystem to support 1) formatting log entries to OTLP and 2) export logs/traces to configurable OTLP backends. For instance, with a service that accepts both traces and logs (like the OTel collector), the DBOS SDK can be configure to export both signals to this service: ```yaml telemetry: OTLPExporters: - tracesEndpoint: 'http://localhost:4318/v1/traces' - logsEndpoint: 'http://localhost:4318/v1/logs' ``` Note this setup works out of the box with Jaeger and the OTel collector (see screenshots bellow.) Also note `opentelemetry-js` logs API is experimental, so I fixed the version we use to `0.41.2` **Details** - Use a wrapper around the Winston logger everywhere. - The wrapper adds a new OTLP transport to the wrapped winston logger if it detects configured exporters - This transport simply formats an OTel `LogRecord` and push it to our collector - Massively simplify our typing by only using opentelemetry `LogRecord` and `Span` **Others** - Metadata are injected to all Winston transports and the transport logic decides whether to include them or not **Thoughts** - There exists a neater way to integrate the active span context to each OTLP log records, which we can consider in the future. Talk to me for details or see PR comments about having a context manager - The Winston folks are actively working on an OTLP transport/exporter themselves, see [here](open-telemetry/opentelemetry-js-contrib#1558) and [there](open-telemetry/opentelemetry-js-contrib#1837). In the future we might want to use it. **Tests** ☑️ Unit tests: ideally we would be able to intercept outgoing HTTP requests and expect the payload to conform our expectations. This seems to be quite some work so pushing that for later. ✅ Tested with DBOS cloud integration tests. ✅ Logs w/o metadata: ![Screenshot 2023-12-11 at 11 57 42](https://github.com/dbos-inc/dbos-ts/assets/3437048/093efe7a-6ff2-4ec0-bf6b-f4d9e30f7ed9) ✅ Logs w/ metadata: ![Screenshot 2023-12-11 at 12 01 02](https://github.com/dbos-inc/dbos-ts/assets/3437048/8a8ed3ff-37a7-4632-aee4-d140bfb0a549) ✅ Logs & traces exported to a local OTel collector ![Screenshot 2023-12-10 at 14 23 25](https://github.com/dbos-inc/dbos-ts/assets/3437048/292acc9e-a7f6-47e5-bb58-f7216d8afcfd) ✅ Traces exported to a local Jaeger: <img width="1784" alt="jaeger" src="https://github.com/dbos-inc/dbos-ts/assets/3437048/89b6e8dc-a2da-43c6-9f4f-821ed31c2159">

hectorhdzg · 2024-01-03T01:04:46Z

Updated the code to use Winston transports instead, we discussed log instrumentations in SIG meeting a few weeks ago, and the consensus was to provide a log appender class, a winston transport in this case, and also allow auto instrumentation injecting the transport automatically, please take a look when you have a chance. @martinkuba @blumamir @seemk @mhennoch

pichlermarc

Looks like all comments are now addressed. Thanks for sticking with us through the long review process @hectorhdzg.

@trentm I see there's still a changes requested review from your side - would you mind having another look to check if your comments have been addressed? 🙂

pichlermarc · 2024-02-21T16:48:50Z

packages/winston-transport/src/OpenTelemetryTransportV3.ts

+
+  constructor(options?: Transport.TransportStreamOptions) {
+    super(options);
+    this._logger = logs.getLogger('@opentelemetry/winston', VERSION);


I'd prefer @opentelemetry/winston-transport, I think that's the one that makes the most sense given that this can be used as a stand-alone

…g/opentelemetry-js-contrib into hectorhdzg/winstonlogs

hectorhdzg · 2024-03-05T21:25:20Z

@trentm let me know if there is any concern pending

trentm

I'm really sorry for the block and long delay. I have a couple small changes that I think should be done (e.g. the package-lock.json vestigial entry), a few nits (i.e. feel free to ignore "nits" from me at your discretion), and then other Qs can be deferred to separate issues.

packages/winston-transport/README.md

plugins/node/opentelemetry-instrumentation-winston/package.json

package-lock.json

packages/winston-transport/package.json

packages/winston-transport/src/utils.ts

trentm · 2024-03-14T00:48:05Z

packages/winston-transport/src/utils.ts

+  logger: Logger
+): void {
+  const { message, msg, level, meta, ...splat } = record;
+  const attributes = Object.assign(meta ?? {}, {});


re-up on this question: https://github.com/open-telemetry/opentelemetry-js-contrib/pull/1837/files#r1484898511

trentm · 2024-03-14T00:51:19Z

packages/winston-transport/src/utils.ts

+      attributes[key] = splat[key];
+    }
+  }
+  const logRecord: LogRecord = {


Note: This is where instrumentation-bunyan intentionally drops the trace_id, span_id, and trace_flags fields that might have been added by the enabled "log correlation" functionality.

However, to not hold up this PR, this could be deferred to a separate issue/PR along with the outstanding question -- #1837 (comment) -- about the LogRecord traceId, spanId, and traceFlags fields being incorrectly set in cases where there is stream buffering.

Created #2010 to follow up on this

trentm · 2024-03-14T17:44:56Z

@hectorhdzg I see you resolved a number of the conversations above, but I don't see any new commits pushed to your feature branch. Did you perhaps forget to push up your local changes? (Sorry if I'm jumping the gun. I'm trying to not be the slow reviewer here again. :)

Update: Ah, the latest comment update was only minutes ago, so I think I'm jumping the gun.

hectorhdzg · 2024-03-14T17:57:51Z

@trentm sorry I usually start resolving whenever I update locally, and push changes after, everything should be up to date now.

trentm

LGTM. Thanks!

Add recordLog emit

638c36d

hectorhdzg requested a review from a team November 30, 2023 00:18

hectorhdzg added 3 commits November 29, 2023 16:19

Merge branch 'main' into hectorhdzg/winstonlogs

d461556

Readme lint

4ec2196

Merge branch 'hectorhdzg/winstonlogs' of https://github.com/hectorhdz…

20f58ab

…g/opentelemetry-js-contrib into hectorhdzg/winstonlogs

seemk reviewed Dec 1, 2023

View reviewed changes

hectorhdzg added 3 commits December 1, 2023 16:36

Update

01286b5

Lint

adc8f7a

Merge branch 'main' into hectorhdzg/winstonlogs

7d24291

mhennoch reviewed Dec 5, 2023

View reviewed changes

plugins/node/opentelemetry-instrumentation-winston/README.md Show resolved Hide resolved

mhennoch reviewed Dec 5, 2023

View reviewed changes

blumamir reviewed Dec 6, 2023

View reviewed changes

hectorhdzg added 6 commits December 8, 2023 13:40

Update

fe2bdbb

Merge branch 'main' into hectorhdzg/winstonlogs

1d4b91a

Update

e76321b

Merge branch 'main' into hectorhdzg/winstonlogs

d80c76a

Lint

683c578

Tests

ffa9f9b

maxdml mentioned this pull request Dec 9, 2023

OTLP telemetry exporter dbos-inc/dbos-transact-ts#213

Merged

hectorhdzg mentioned this pull request Dec 11, 2023

Log Bridge API Appender packages or use current Instrumentation packages open-telemetry/opentelemetry-js#4355

Closed

hectorhdzg added 3 commits December 29, 2023 15:52

Added transports

441f72a

Add log appenders(Winston transports)

6731e97

Merge remote-tracking branch 'upstream/main' into hectorhdzg/winstonlogs

1c7ac82

Addressing comments

2405492

pichlermarc approved these changes Feb 22, 2024

View reviewed changes

mhennoch approved these changes Feb 22, 2024

View reviewed changes

hectorhdzg added 6 commits February 26, 2024 15:01

Merge remote-tracking branch 'upstream/main' into hectorhdzg/winstonlogs

c2dc4eb

Update logger name

95d8a6a

Merge remote-tracking branch 'upstream/main' into hectorhdzg/winstonlogs

fb41f93

Merge branch 'hectorhdzg/winstonlogs' of https://github.com/hectorhdz…

ec9da89

…g/opentelemetry-js-contrib into hectorhdzg/winstonlogs

Update

1fe760d

Merge branch 'main' into hectorhdzg/winstonlogs

615c702

hectorhdzg added 6 commits March 7, 2024 14:20

Merge remote-tracking branch 'upstream/main' into hectorhdzg/winstonlogs

8cf9b57

Update logs api dependency

064cba7

Merge remote-tracking branch 'upstream/main' into hectorhdzg/winstonlogs

9e9e7d1

Update

fae82c9

Merge branch 'main' into hectorhdzg/winstonlogs

1750b3e

Merge branch 'main' into hectorhdzg/winstonlogs

a36d4e8

trentm requested changes Mar 14, 2024

View reviewed changes

hectorhdzg added 3 commits March 14, 2024 10:46

Addressing comments

a5e1ad6

Merge branch 'main' into hectorhdzg/winstonlogs

32fc1a2

Lint

5e0291e

hectorhdzg mentioned this pull request Mar 14, 2024

[instrumentation-winston] LogRecord traceId and spanId values are not getting set correctly #2010

Open

trentm approved these changes Mar 14, 2024

View reviewed changes

hectorhdzg merged commit 358345f into open-telemetry:main Mar 14, 2024
17 checks passed

dyladan mentioned this pull request Mar 14, 2024

chore: release main #2007

Merged

trentm mentioned this pull request May 30, 2024

feat(instrumentation-pino): add log sending to Logs Bridge API #2249

Merged

6 tasks

This was referenced Jun 19, 2024

[instrumentation-winston] May duplicate transport (and logs) without notice #2283

Closed

auto-instrumentation with winston doesn't seem to report logs to open collector #2090

Open

trentm mentioned this pull request Nov 20, 2024

Winston logging not working with node js #2280

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(instrumentation-winston): add log sending to Logs API #1837

feat(instrumentation-winston): add log sending to Logs API #1837

hectorhdzg commented Nov 30, 2023 •

edited

Loading

codecov bot commented Nov 30, 2023 •

edited

Loading

seemk Dec 1, 2023

hectorhdzg Dec 2, 2023

seemk Dec 3, 2023

mhennoch Dec 5, 2023

hectorhdzg Dec 8, 2023

blumamir left a comment

blumamir Nov 30, 2023

hectorhdzg Dec 8, 2023

blumamir Dec 6, 2023

hectorhdzg Dec 8, 2023

blumamir Dec 6, 2023

hectorhdzg Dec 8, 2023

blumamir Dec 6, 2023

hectorhdzg Dec 8, 2023

blumamir Dec 6, 2023

hectorhdzg Dec 8, 2023

blumamir Dec 6, 2023

hectorhdzg Dec 8, 2023

martinkuba commented Dec 9, 2023

hectorhdzg commented Dec 11, 2023

martinkuba commented Dec 11, 2023

hectorhdzg commented Dec 11, 2023

hectorhdzg commented Jan 3, 2024

pichlermarc left a comment

pichlermarc Feb 21, 2024

hectorhdzg commented Mar 5, 2024

trentm left a comment

trentm Mar 14, 2024 •

edited

Loading

trentm Mar 14, 2024

hectorhdzg Mar 14, 2024

trentm commented Mar 14, 2024 •

edited

Loading

hectorhdzg commented Mar 14, 2024

trentm left a comment

		const values = args.slice(2, msgArguments + 1);
		logRecord['msg'] = util.format(args[1], ...values);

feat(instrumentation-winston): add log sending to Logs API #1837

feat(instrumentation-winston): add log sending to Logs API #1837

Conversation

hectorhdzg commented Nov 30, 2023 • edited Loading

codecov bot commented Nov 30, 2023 • edited Loading

Codecov Report

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

blumamir left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

timestamp

observedTimestamp

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

martinkuba commented Dec 9, 2023

hectorhdzg commented Dec 11, 2023

martinkuba commented Dec 11, 2023

hectorhdzg commented Dec 11, 2023

hectorhdzg commented Jan 3, 2024

pichlermarc left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hectorhdzg commented Mar 5, 2024

trentm left a comment

Choose a reason for hiding this comment

trentm Mar 14, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

trentm commented Mar 14, 2024 • edited Loading

hectorhdzg commented Mar 14, 2024

trentm left a comment

Choose a reason for hiding this comment

hectorhdzg commented Nov 30, 2023 •

edited

Loading

codecov bot commented Nov 30, 2023 •

edited

Loading

`timestamp`

`observedTimestamp`

trentm Mar 14, 2024 •

edited

Loading

trentm commented Mar 14, 2024 •

edited

Loading