kennknowles commented on PR #36615: URL: https://github.com/apache/beam/pull/36615#issuecomment-3451978660
> > Another question: How many Dataflow jobs are in parallel? I think at some point we single-threaded these to avoid quota issues, or some other build issue, but we could run with 10+ threads and just need to adjust our quota. > > It was set to default value > > https://github.com/apache/beam/blob/4325e2b1e01623ffe23a60bbdee421354b490317/runners/google-cloud-dataflow-java/build.gradle#L274 > > Checked in Dataflow console running it's effectively 4 jobs in parallel. > > > but we could run with 10+ threads and just need to adjust our quota. > > We'd definitely need much larger quota if bump this value. Currently one of the quota most easily hit is in-use ipv4 addresses on us-central1. ipv4 quota is not easy to get approved. I wonder if we can run validates runner tests (and supported Dataflow test in general) on ipv6 only or disable external networks (could be follow up) This seems like a good idea. I'm not sure why we would need external networks. Certainly not for all tests. Another possibility could be to shard the test suite for Dataflow. I think probably ParDoTest and CombineTest probably have a ton of pipelines and if you separate some "core ValidatesRunner suite" from "extended ValidatesRunner suite" it might reduce flakes and speed up reruns while reducing quota usage. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
