Occasionally our users report an issue where updates made to child work items do not promptly calculate updated rollup information to the parent artifacts. This has been an elusive problem to solve, so we wanted to provide an update on it.
Background:
The implementation to calculate rollups is not new, and for a fair amount of time it worked well. It needed to be tuned to remain performant, but continued to work well. We had isolated reports of some anomaly (or anomalies) that causing it to intermittently not perform well at all however. Nailing down precisely what that anomaly is, has been challenging to say the least. There were tuning measures we could take that theoretically should aid the issue, which we did, and we did see improvements. There have been multiple occasions where we thought we had it resolved, only to see reports come through yet again.
Currently:
In late May reports of degradation in performance spiked. Following that and despite implementing more changes that looked to resolve the issue, the reports continued. Having exhausted most tuning measures, we have discussed other options, including but not limited to re-architecting the job from the ground up. We are currently implementing additional telemetry to help us suss out the frequency and the severity of the delays. The expectation is that will help us make a decision about which of the options we've considered stand the best chance of succeeding, troubleshoot what conditions cause the more dramatic delays to occur and to give us more of a baseline to measure improvements against.
For anyone impacted by this issue, thanks for your patience. We hope this information is helpful