SmallRye Fault Tolerance: add support for OpenTelemetry Metrics #46636
Conversation
|
Hi @brunobat, this is what we discussed earlier today. The failing TCK test is a TCK bug; I submitted a PR (microprofile/microprofile-fault-tolerance#655), but until that is merged and released, I've kept an exclusion. EDIT: I also filed a TCK challenge: microprofile/microprofile-fault-tolerance#656 |
This comment has been minimized.
This comment has been minimized.
|
Hmm, the failing tests might be relevant. I'll take a look. |
This requires propagating the "metrics enabled at build time" information through the `OpenTelemetrySdkBuildItem`. I took the liberty to also add the tracing and logging build time enablement information, even though I have no use for it at the moment. One TCK test remains excluded, because it has a bug [1][2]. [1] microprofile/microprofile-fault-tolerance#656 [2] microprofile/microprofile-fault-tolerance#655
Ladicek
force-pushed
the
fault-tolerance-opentelemetry
branch
from
bff6ca9 to
49fa4b9
Compare
quarkus-bot
bot
added
area/arc
|
Hi @brunobat, I had to add some "build time enabled" information to the |
Status for workflow
|
| Status | Name | Step | Failures | Logs | Raw logs | Build scan |
|---|---|---|---|---|---|---|
| ✔️ | JVM Tests - JDK 17 | Logs | Raw logs | 🔍 | ||
| ✔️ | JVM Tests - JDK 21 | Logs | Raw logs | 🔍 | ||
| ✖ | JVM Tests - JDK 17 Windows | Build |
Failures | Logs | Raw logs | 🔍 |
Full information is available in the Build summary check run.
You can consult the Develocity build scans.
Failures
⚙️ JVM Tests - JDK 17 Windows #
- Failing: extensions/vertx-http/deployment
! Skipped: devtools/cli extensions/agroal/deployment extensions/amazon-lambda-http/deployment and 146 more📦 extensions/vertx-http/deployment
✖ io.quarkus.vertx.http.proxy.TrustedForwarderDnsResolveTest.testTrustedProxyResolved - History - More details - Source on GitHub
java.io.IOException:
Error while binding on /127.0.0.1:53530
original message : Address already in use: bind
at org.apache.mina.transport.socket.nio.NioDatagramAcceptor.open(NioDatagramAcceptor.java:708)
at org.apache.mina.transport.socket.nio.NioDatagramAcceptor.registerHandles(NioDatagramAcceptor.java:228)
at org.apache.mina.transport.socket.nio.NioDatagramAcceptor.access$300(NioDatagramAcceptor.java:68)
at org.apache.mina.transport.socket.nio.NioDatagramAcceptor$Acceptor.run(NioDatagramAcceptor.java:164)
at org.apache.mina.util.NamePreservingRunnable.run(NamePreservingRunnable.java:64)
| metricsProviders++; | ||
| } | ||
| if (openTelemetrySdk.map(OpenTelemetrySdkBuildItem::isMetricsBuildTimeEnabled).orElse(false)) { | ||
| builder.addBeanClass("io.smallrye.faulttolerance.metrics.OpenTelemetryProvider"); |
There was a problem hiding this comment.
This will only be affected by the build time configuration. We are still going to instrument if the OTel SDK is disabled at runtime...
I suggest you skip the processFaultToleranceAnnotations() method (just return or skip all) in case OpenTelemetry is used and if disabled at runtime.
There was a problem hiding this comment.
I cannot really do that, because I try to do a lot of things at build time and there's no runtime config at that time.
What I could do is read the runtime config and remove the OTEL metrics provider if runtime-disabled. I'm not yet sure what the best way would be -- clearly the RuntimeValue is a Quarkus-specific concept and so it cannot be in SmallRye Fault Tolerance. On the other hand, SmallRye Fault Tolerance has configuration for disabling the specific metric providers at runtime, so there might be a way to translate the OTEL config to the SRye FT config, somehow. Not yet sure how.
There was a problem hiding this comment.
Also, that build item is only produced at runtime, even if it's published with build time data... Not sure if this will work properly and if we need different build items for OTel build and run times.
There was a problem hiding this comment.
Build items are never produced at runtime. I mean, at runtime, the concept of build items and build steps doesn't even exist.
There was a problem hiding this comment.
Right, the config values the build item holds might come from runtime.
There was a problem hiding this comment.
Can the Otel Provider be disabled?
See: https://github.com/smallrye/smallrye-fault-tolerance/blob/8da7dae36f195608291687793ca9f12fd9567d93/implementation/fault-tolerance/src/main/java/io/smallrye/faulttolerance/metrics/OpenTelemetryProvider.java#L27
The current code doesn't support that, but with some changes...
There was a problem hiding this comment.
Yeah, that's the configuration I mentioned above. There might be a way, I just didn't have the time to explore it yet.
One TCK test remains excluded, because it has a bug [1].
[1] microprofile/microprofile-fault-tolerance#655