Open Telemetry #2706
Open Telemetry #2706
Conversation
| // SdkMeterProvider.builder | ||
| // .registerMetricReader(prometheusHttpServer) | ||
| // .setResource(resource) | ||
| // .build |
There was a problem hiding this comment.
I am leaving these comments in here for now because I think we should switch to open telemetry for metrics reporting too
There was a problem hiding this comment.
Is turning this on mutually exclusive with switching over the metrics reporting stuff? I'm wondering if this can be a gradual switchover with the prometheus server being on but exposing nothing, and then moving metrics over more and more until we can fully deactivate the statsd sidecar.
There was a problem hiding this comment.
it can be gradual with the statsd stuff. Unfortunately the old OpenTelemetryMetrics class from workbench-libs (which actually uses open census) is in use and sitting on the prometheus port
| httpRequestTimer(ExpandedMetricBuilder.empty)(request, response).update(elapsed, TimeUnit.MILLISECONDS) | ||
| response | ||
| def instrumentRequest: Directive1[Context] = | ||
| (traceRequest & extractRequest).tflatMap { case (otelContext, request) => |
There was a problem hiding this comment.
note that this is where tracing starts
|
jenkins retest |
| // SdkMeterProvider.builder | ||
| // .registerMetricReader(prometheusHttpServer) | ||
| // .setResource(resource) | ||
| // .build |
There was a problem hiding this comment.
Is turning this on mutually exclusive with switching over the metrics reporting stuff? I'm wondering if this can be a gradual switchover with the prometheus server being on but exposing nothing, and then moving metrics over more and more until we can fully deactivate the statsd sidecar.
| val workspaceManager = clientLibExclusions("bio.terra" % "workspace-manager-client" % "0.254.997-SNAPSHOT") | ||
| val dataRepo = clientLibExclusions("bio.terra" % "datarepo-jakarta-client" % "1.568.0-SNAPSHOT") | ||
| val resourceBufferService = clientLibExclusions("bio.terra" % "terra-resource-buffer-client" % "0.198.42-SNAPSHOT") | ||
| val billingProfileManager = clientLibExclusions("bio.terra" % "billing-profile-manager-client" % "0.1.502-SNAPSHOT") |
There was a problem hiding this comment.
Nice, we can get rid of the bpm and WSM javax clients once this rolls out.
| ) | ||
|
|
||
| try | ||
| traceNakedWithParent("executeGoogleRequest", childContext) { _ => |
There was a problem hiding this comment.
does this lose the googleProjectId and billingAccount attributes on the span?
| // WithOtelContextFilter must run before JakartaTracingFilter so give it a higher priority | ||
| val priority = 1 | ||
| client.getHttpClient.register(new WithOtelContextFilter(otelContext), priority) | ||
| client.getHttpClient.register(new JakartaTracingFilter(GlobalOpenTelemetry.get()), priority + 1) |
There was a problem hiding this comment.
I see this repeated a few times; is it worth a reusable method?
|
jenkins retest |
1 similar comment
|
jenkins retest |
Ticket: https://broadworkbench.atlassian.net/browse/WOR-1462
updates to use open telemetry
updates to use jakarta client libs
Most of the changes are minor. A few places added trace contexts where they were missing. A new
RawlsTracingContextclass used when a user is not present.PR checklist
model/, then you should publish a new officialrawls-modeland updaterawls-modelin Orchestration's dependencies.