[New scheduler] Deflake test heartbeat #11586
Conversation
|
Hmm this failure only happened at Windows build though? I am not sure how this will be related to only Windows failure? But let me approve it first anyway. Let's see the build result. |
|
I believe this is just coincidental because the iteration order over the map is consistent on linux, and consistent but different on windows. I won't say that confidently in case the test doesn't pass though :p |
|
Please resolve issues in the CI (but it at least resolves the Windows issue!) |
|
remaining tests failures seem unrelated |
|
https://travis-ci.com/github/ray-project/ray/jobs/406407223 This failure looks like it is related to this PR. (build failure from cluster_task_manager tests) |
|
CAn you merge the latest master? There are lots of serve failures which must've been resolved in the master branch. |
Why are these changes needed?
The heartbeat test was relying on the order of entries in the
resource_load_by_shapefield, but these were coming from a map. Simple fix is to just compare without worrying about the order.Related issue number
Checks
scripts/format.shto lint the changes in this PR.