First of all, The timeout configuration setting that's described in STASH-4129 does not apply to /rest/api/latests/projects/<project>/repos/<repo>/commits endpoint. If you want to tune the process timeout setting, you should tune either process.execution.timeout or process.idle.timeout as described in the Process Execution section.
As to why the operation times out, it's hard to say just from the stack trace. There could be a number of reasons:
- your script(?) that does the REST call does not consume any of the output and the git process times out because its output hasn't been read for more than 60s, or the script does read some, but not all of the data and does not close the HTTP connection?
- for some reason the git rev-list command that is started takes more than 120s to complete.
- for some reason the git rev-list command takes more than 60s before the first output appears.
Does this happen to all repositories or is it just a single repository? Try to do the same REST call using curl from the command line and time how long it takes and whether you get any output.
If the git command really is slow, there might still be a host of reasons. The disk could be slow, or if you're using a network filesystem such as NFS, the latency could be high resulting in slow disk I/O. Or the repository could have a large number of loose objects or lots of small pack files, which can also slow down git operations.
If the problem persists, I'd suggest:
- Enable profiling (Administration > Logging and Profiling)
- Reproduce the problem a few times
- Leave on profiling for some time to also collect performance information for other operations
- Create a support zip (Administration > Atlassian Support Tools > Support Zip)
- Open a support ticket on https://support.atlassian.com and upload the support zip.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.