[
https://issues.apache.org/jira/browse/AVRO-2630?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16998085#comment-16998085
]
Ryan Skraba commented on AVRO-2630:
-----------------------------------
Quick update: since we're now using JAVA=8 and JAVA=11 in the same image for
now, I tried caching the docker image in travis in a "warmup" stage before the
builds.
In the worst case (cache is present, but nothing reused for the docker build),
the end-to-end time to get a checkmark on a PR was significantly worse (57
minutes versus 40). In the best case (Dockerfile *mostly* built from cached
files from a previous build) was 42 minutes, slightly worse than today.
Small hiccup in the POC: it looks like the ruby layer is always rebuilt instead
of being taken from the docker build cache. I'll give that a check before
entirely abandoning this approach.
> Travis build is nearing timeout limits.
> ---------------------------------------
>
> Key: AVRO-2630
> URL: https://issues.apache.org/jira/browse/AVRO-2630
> Project: Apache Avro
> Issue Type: Improvement
> Components: build
> Reporter: Ryan Skraba
> Priority: Major
>
> Single "jobs" on travis are [limited to 50
> minutes|https://docs.travis-ci.com/user/customizing-the-build/#build-timeouts].
> A "normal" build takes 47-48 minutes and can occasionally bump over the
> limit, causing timeout errors.
> We can see that this happens occasionally on master and on pull requests.
> We should find a strategy to limit the length of time for a of the jobs being
> run.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
