AWS Codepipeline is very slow(source to build stage)

1

Hello, We have an AWS Codepipeline where it has Source stage as Codecommit and build stage as Jenkins(Custom) and then ECS deployment. We observed that moving from Source stage to build stage taking around 40 to 45 min time. The size of our repository is 2.4 GB. We see that everytime it zip the code and upload the S3 bucket which acts as source repository for Build stage. We are trying to understand why it is taking that much time ? is this behaviour is expected based on repository size ?
Note: the above mentioned time only to move to build stage( the build time is not counted here)

已提問 5 年前檢視次數 2194 次
2 個答案
1

Can you explain what actually happens during the source stage? Why does anything need to get copied to or from S3? CodeBuild can read directly from CodeCommit, so if the source is CodeCommit and the build action is CodeBuild, what actually needs to happen during the source stage?

已回答 4 年前
0

Hi,

Regretfully, this is likely the expected time for the repository source stage action currently based on the repository. We understand the pain this causes and will take this post as a customer request to improve performance on this aspect of the CodeCommit experience.

The CodeCommit source stages tend to be particularly slow packaging large individual files into the ZIP archives. As a workaround, if you can move some of your largest file assets out into other locations (like S3, for example), then it would very likely have a positive performance impact on the source stage.

Another potential for slowness comes from long delta chains that Git may be creating. If you repack your repository with a smaller delta chain length and re-push to CodeCommit, this may also have some positive performance impact on your Source stage. To do this, run:

git repack -adf --depth 15
git push

Note that there must be a new content to push here.

Thanks!

AWS
已回答 5 年前

您尚未登入。 登入 去張貼答案。

一個好的回答可以清楚地回答問題並提供建設性的意見回饋,同時有助於提問者的專業成長。

回答問題指南