Description
Description
In this project, we use a pool of self-hosted runners (paid for by Microsoft) to run some Linux CI jobs in Azure DevOps:
Line 93 in f91dcfe
These stopped accepting jobs about 2 weeks ago: #6905 (comment)
Opening this to track the work of fixing this.
Benefits of this work
Unblocks development here. It has been more than 2 weeks since the last commit (f91dcfe) was merged to master here.
Acceptance criteria
- LightGBM's CI matrix is fully working
Approach
Per offline messages, it seems that the pool of machines we're using is no longer supported by Azure DevOps.
I believe this could be fixed by upgrading to a new pool, being worked on in #6866
Notes
This issue is just the most recent in a series of several multi-week disruptions to development here. As a result of those disruptions, development has significantly slowed down here... there have been only 11 commits merged to master in the last 60 days 😬
These issues have lasted so long because their resolutions often require administrative actions that can only be taken by a Microsoft employee, and there's only 1 Microsoft employee working on this project. It has been this way for several years and the long disruptions to CI are really frustrating and demotivating for external contributor (and also for long-time maintainers like me and @StrikerRUS) 😫
@shiyu1994 it would really be appreciated if you could get Microsoft to commit some more maintainers to the project so these things do not always depend solely on you. We always appreciate your help but also know you're busy with a lot of other things! For example since SynapseML depends on LightGBM, maybe @mhamilton723 or someone else from that project could help out here?