Closed reservoir-dogs closed 4 years ago
@reservoir-dogs Thanks for the report. This looks like a bug in the retry code recently introduced in #3022. Can you confirm this problem is intermittent for you? It looks like things should work normally when a retry isn't attempted.
cc @RuslanRasulovAkv. It looks like the ProcessInvoker isn't designed to be reused in this way.
Indeed, we observed in the past this error at different stages of the Checkout and if we restarted, then it ended up working.
But since Friday, the retry is systematic on this stage of the Checkout. So, we can no longer deliver. Do you have any idea of the initial error? or how to have more information about it?
@reservoir-dogs Based on the logs you posted the error on the first attempt is:
2020-07-20T09:31:47.0757482Z ##[debug]No workspace matching ws_8_8;Project Collection Build Service (IntralogConsulting) on computer * found in Team Foundation Server https://dev.azure.com/xxx/.
Could you please report this at Developer Community? The error in the second attempt is an agent bug in the retry code, but the root issue (the error on the first attempt) is not necessarily in the agent. Opening a Developer Community issue will get a wider range of support staff looking into it and made it easier to route to the best people to help.
We are working on this bug (_proc should be null), thanks for reporting!
For your information, the agent hosted with the same version works great! The cause could be the permissions of the Azure agent (but we haven't made any changes in the past week) or the hosting system. Finally, we decided to work around this issue by migrating to git source control and we did that this morning which was in our plans.
@RuslanRasulovAkv The version of the agent with the retry logic reverted for now (2.172.2) has been fully rolled out, so I'm going to close this out.
Agent Version and Platform
Version of your agent? 2.172.1
OS of the machine running the agent? Linux
Azure DevOps Type and Version
dev.azure.com
What's not working?
The pipeline worked last wednesday with agent's version 2.171.1. The pipeline has no changes and Friday, it was in failure with agent's version 2.172.1.
Agent and Worker's Diagnostic Logs
Full logs :