-
Suggestion
-
Resolution: Done
-
None
-
JDK 1.6.0_29, CentOS 5.8, MySQL 5.5
Recently we found our Bamboo server disk space hitting particular free space thresholds. We configured stricter artifact retention policies for some of our more artifact-heavy plans, and increased base restrictions for the global expiry.
While we can see the build expiry process running, it seems to taking more than a week to get through a full build expiry run, even though we've set Bamboo to run build expiry every night.
The expiry process needs to be more performant. It seems to be a serial process, so multi-threading it would be a start.
Additionally, instead of an expiry process kicking off at the global expiry time, it might be a possibility to have each build run the plan expiry process at the end of the build.
- copied to
-
BDEV-1054 Failed to load
[BAM-11711] Improve performance of build expiry process
Here's some other stats from the system:
Instance Statistics Number of plans4838 Number of builds (approx)879758 Full reindex time (approx)16 hours, 17
Thanks for the report David. I agree, Multi-threading the operation might be a start. What is the current size of your artifacts directory at the moment?
Hi David,
Because the instance is very big (4838 plans), the problem may not have its root cause in the build expiry feature.
Depending on the size of the builds (number of test results, plan definition complexity, etc) and the overall system performance, it could be that the bottle neck is Bamboo taking very long to remove the DB content before removing the disk files.
I've raised the following support ticket so that we can request more details:
https://support.atlassian.com/browse/BSP-7511
The ticket will allow us to keep confidential data available to you and us only. Lets continue the communication at BSP-7511.
Cheers,
Renan