During an incident investigation, I found the case where an app still had a value for PartitionCount of 4 in their taskhub.json blob even though their app currently was using 16 partitions. This lead our scaling infrastructure to ignore 12/16 of the control queues, meaning the app could not activate for hours until an orchestration happened to fall in one of the 4 partitions the scale controller was looking at.
pushing out of 2.2.2.
The workaround for this is to manually edit the taskhub.json file.
This bug should only impact consumption and premium plans.
@ConnorMcMahon I just ran into this as well. Nothing out of the ordinary, just updated the partitionCount in host.json. It is worth noting the storage account was already created during a previous release.
Is this the same issue where the code is looking for storageOptions instead of storageProvider?
The problem I am facing now is I as a developer do not have access to manually modify taskhub.json.
@mpaul31,
I do not believe this is the same issue, as I found this separately. I will see if we can get this into 2.4.0 releasing at the end of the month.
Unfortunately, this will likely get pushed to October release. Hopefully the fix will be in early October, and we can do a prerelease version of DurableTask.AzureStorage on myget so people can get this fix shortly.
Pull request here.
The pull request linked has been merged, and this will release with v2.4.0