-
Notifications
You must be signed in to change notification settings - Fork 1.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
22.3.0 CE native build crashes #5303
Comments
We just tried a recent-ish dev build (GraalVM CE 23.0.0-dev-20221020_1959), and the native build succeeded. |
I am having this very same issue but it is strange because it only happens at Jenkins. At local I am being able to build the native image. That is weird :S I am using native generation through docker: |
@martincalvodaniel does this build fixes the issue for you too? |
At local, I am using this image:
And its working. At Jenkins I am using that exact digest image and its failing with the very exact message of the PO:
|
I also see this occurring on my both home PCs, while it seems to work on GitHub Actions. The mandrel image works for me as a workaround:
|
Hi @peter-hofer. I was looking into this and noticed 62432ef. Unfortunately I can't reproduce the issue so I can only "look around". Do you think a backport of 62432ef to 22.3 could fix the issue? |
I have a very similar env but I'm unable to reproduce it. I'm running Colima docker daemon but the build tool in use here does not like it:
Docker client does work:
Even after setting the
@lowecg Can you make this reproducible outside of docker? Or fix holy-lamdba? According to this issue, the code seems to assume that a docker user on macos will have |
@galderz Noted on the Docker issue. I've put an alternative Docker test into my holy-lambda fork and have updated the repro repo to use it. Please could you try that? |
@zakkak, this is indeed intended to be a workaround for this issue, so backporting should help. We have seen this problem on very rare occasions without being able to pin it down. I believe @christianhaeubl would appreciate if you had a consistent reproducer. |
@peter-hofer Unfortunately I don't have a reproducer either, but I have heard more than a couple of people having this issue with Quarkus. CCing some of them @rsvoboda, @loicmathieu, @ppalaga, @beikov in case they can provide more details on how to reproduce. |
Just looking into it. I can reproduce the fail on RHEL 8 machine running in OpenStack.
Steps to reproduce:
|
When using local GraalVM ( So the failures we see are related to the usage of CC @cescoffier, @gsmet |
It could be a missing dependency, but I will need to know what new dependency needs to be added to the image (while mandrel does not need it) |
Still no luck, I don't have
|
I've replicated it using this instructions on my Fedora box /cc @zakkak. If I can only replicate it with builder image, I can quickly build a patched version with the commit that @zakkak thinks might fix it and verify. |
Note the instructions that worked to replicate are the ones from previous @rsvoboda comment #5303 (comment). |
@galderz Good news that you're able to reproduce the issue with quarkus. I patched holy-lambda to remove the |
FYI I'm trying to verify @zakkak's theory about the patch that might avoid this issue. I'll update when I've found out something. |
I think @zakkak's theory has legs. I built |
I've sent a PR to backport the fix to 22.3: #5392 |
Thanks @galderz! |
Due to a bug in GraalVM, let's default to using Mandrel containers for now. See: - quarkusio#29124 - oracle/graal#5303
Due to a bug in GraalVM, let's default to using Mandrel containers for now. See: - quarkusio#29124 - oracle/graal#5303
@peter-hofer @christianhaeubl I think I found a way to reproduce the issue! Running:
Note the results in:
With the above command I am able to reproduce the issue with both rootless and rootfull docker instances. Environment:
|
I think this is reproducible with
The reason why 22.3.0 CE builds crash seems to be because
|
Should we create a user for the uid we are using in the container? |
Due to a bug in GraalVM, let's default to using Mandrel containers for now. See: - quarkusio#29124 - oracle/graal#5303 (cherry picked from commit 4583de3)
You would need to do this each time the container is created, since the uid we use is the one of the user starting the container, so it's not fixed. BTW Quarkus builder images ship with a quarkus user with uid 1001 which might be another reason some people (who happen to have users with uid 1001) don't observe the failure. |
45bf93a is the fix in the 22.3 tree. This should be fixed once 22.3.1 is released. |
Due to a bug in GraalVM, let's default to using Mandrel containers for now. See: - quarkusio#29124 - oracle/graal#5303 (cherry picked from commit 4583de3)
Disable monitoring due to oracle/graal#5303, reported via projectnessie#5900.
Disable monitoring due to oracle/graal#5303, reported via projectnessie#5900.
Disable monitoring due to oracle/graal#5303, reported via #5900.
Since GraalVM 22.3.1 CE has been released and according to my test, the use case of the OP against this release works properly (as you can see below), maybe this issue can now be considered fixed, anything else to test/do?
|
Closing as fixed. |
This change effectively reverts projectnessie#5903 (projectnessie#5900), since oracle/graal#5303 is marked resolved for GraalVM 22.3.1, which is not also used for our native image builds via `quay.io/quarkus/[email protected]` pointing to `22.3.1`.
This change effectively reverts #5903 (#5900), since oracle/graal#5303 is marked resolved for GraalVM 22.3.1, which is not also used for our native image builds via `quay.io/quarkus/[email protected]` pointing to `22.3.1`.
FWIW, I'm still seeing this issue with 22.3.1, whilst building a docker image which downloads graalvm and then follows that up by running
|
@rubin55 That's unrelated to this particular issue, sorry. Your issue is likely caused by a bug in the container detection code in substrate. Likely the graal equivalent of https://bugs.openjdk.org/browse/JDK-8272124 |
@rubin55 did you find a solution to the problem? I've got similar problem reported here. |
@pompiuses no, I never delved deeper and couldn't resolve at the time, sorry! |
Why: The build failed for me during at quarkus-maven-plugin:2.16.5.Final:build (default) @ kafka-server with Fatal error: Failed while initializing the performance data. Which appears to be a known issue with 22.3.0 oracle/graal#5303 Fixed for me by upgrading to the latest image
Describe the issue
I’m trying to build a Clojure-based lambda using the native image compilation, but the build immediately crashes with the following:
Fatal error: Failed while initializing the performance data.
Steps to reproduce the issue
Please include both build steps as well as run steps
Describe GraalVM and your environment:
More details
I’m trying to build a Clojure-based lambda using the native image compilation, but the build immediately crashes with:
Fatal error: Failed while initializing the performance data.
The text was updated successfully, but these errors were encountered: