James Hutchison created BEAM-7975: ------------------------------------- Summary: error syncing pod - failed to start container artifact (python SDK) Key: BEAM-7975 URL: https://issues.apache.org/jira/browse/BEAM-7975 Project: Beam Issue Type: Bug Components: sdk-py-harness Affects Versions: 2.13.0 Reporter: James Hutchison
{code:java} Error syncing pod 5966....e59c ("<container name>-08131110-7hcg-harness-fbm2_default(5966....e59c)"), skipping: failed to "StartContainer" for "artifact" with CrashLoopBackOff: "Back-off 5m0s restarting failed container=artifact pod=<container name>-08131110-7hcg-harness-fbm2_default(5966.....e59c)"{code} Seeing these in streaming pipeline. Running pipeline in batch mode I'm not seeing anything. Messages appear about every 0.5 - 5 seconds I've been trying to efficiently scale my streaming pipeline and found that adding more workers / dividing into more groups seems to have minimal improvement. Perhaps this is part of the problem? One pipeline which never completed (got to one of the last steps and then log messages simply ceased without error on the workers) had this going on in the kubelet logs. I checked some of my other streaming pipelines and found the same thing going on, even though they would complete. In a couple of my streaming pipelines, I've gotten the following error message, despite the pipeline eventually finishing: {code:java} Processing stuck in step s01 for at least 05m00s without outputting or completing in state process{code} Perhaps they are related? This is running with 7 workers in streaming mode. I haven't checked to see if number of workers is a factor. The pipeline uses requirements.txt and setup.py, as well as using an extra package and using save_main_session. -- This message was sent by Atlassian JIRA (v7.6.14#76016)