Codebuild Batch gone extremely slow in eu-west-1
Hi AWS,
Over the past couple months, I have noticed that the builds in the eu-west-1 region (the only one I use to be fair) has gone incredibly slow, especially with batch builds.
2 months ago, the one batch with less than 20 builds in it would complete each x86_x64 based builds in under 5 minutes, and the aarch64 ones would do in about 7 minutes, as noticed that often getting the ARM builds out of the queue takes a couple minutes longer.
This has now gone so slow that the exact same build project, working for over a year, is regularly failing due to timeout. The timeout is set to 1h, which is already an average of 6x longer than it was necessary before.
What's causing such slowness ?
- 최신
- 최다 투표
- 가장 많은 댓글
Hello,
The information provided is quite less to dig deep into the issue. There could be multiple factors responsible for the same. Hence the best scenario would be to open up an AWS Support Case & then provide the relevant information related to memory, latency etc to determine the cause for the same.
Meanwhile, below is the document provided as a best practice for the AWS Code build to improve performance: https://aws.amazon.com/blogs/devops/improve-build-performance-and-save-time-using-local-caching-in-aws-codebuild/
The builds are not slow to execute. They are slow to start!. It takes a good 40 minutes for a "nested" build project to start. They are stuck in the queue all that time, but once they run, no issues during the build is observed. And not, I am not reaching the per account per region limit of concurrency builds.
Also seen similar posts in other AWS Regions that were posted quite close the time I first started to notice the latency in getting the builds from queued to start / provisioning. So this is something that either is going slow in the codebuild backend or, but much more doubtful, AWS is "running out" of capacity for codebuild projects to run in their allocated pool ?