You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on May 6, 2022. It is now read-only.
As we can see we were always really close to job timeout. The build process also depends on network connection because we are downloading a docker image for each arch, then building our binaries and pushing result images to quey.
In version 0.3.0 we are building additionally migration and cleaner binaries which also affected building time.
As a result, the new Service Catalog version cannot be released because we are hitting timeout:
The job exceeded the maximum time limit for jobs, and has been terminated.
Solution
The Travis job timeout cannot be customized. We are also limmited by the open-source global queue and sometimes we need to wait for jobs to be started.
As "hot-fix" we should parallelize our deploy stage by arch type. In the long term, we should move our release pipeline to Prow infra where we have more control.
The text was updated successfully, but these errors were encountered:
Description
Currently, the Service Catalog release is done by Travis CI. There is one job called Deploy. In that stage, we are building and pushing:
Release history with time consumption:
Travis builds are timeouted when a job on a public repository takes longer than 50 minutes.
source: https://docs.travis-ci.com/user/customizing-the-build/#build-timeouts
Problem
As we can see we were always really close to job timeout. The build process also depends on network connection because we are downloading a docker image for each arch, then building our binaries and pushing result images to quey.
In version 0.3.0 we are building additionally migration and cleaner binaries which also affected building time.
As a result, the new Service Catalog version cannot be released because we are hitting timeout:
Solution
The Travis job timeout cannot be customized. We are also limmited by the open-source global queue and sometimes we need to wait for jobs to be started.
As "hot-fix" we should parallelize our deploy stage by arch type. In the long term, we should move our release pipeline to Prow infra where we have more control.
The text was updated successfully, but these errors were encountered: