Details
-
Bug
-
Resolution: Fixed
-
Low
-
3.0.0
-
None
-
None
Description
Due to a greedy repository scheduling algorithm we use when scheduling repository scanning, if the requested poll period is short (e.g. default = 60 seconds) and checking for changes takes a time that's longer than the poll period (slurp), a new request IncrementalPingRequest will be scheduled as soon as the previous one is done.
This combined with RepositoryHandle#processPingRequests not releasing the executor until all ping requests are processed, will cause the repository never to release the executor, and prevent other repositories from being updated.
Workaround
- Increase the poll period of affected repositories, e.g. from the default 60 seconds to 5 minutes.
- Increase the number of incremental indexing threads (if your system has enough resources to do so).
Attachments
Issue Links
- is caused by
-
FE-3507 Polling mechanism skips repositories when there are a lot of them
- Closed
- is related to
-
FE-6508 Long initial indexing may lead to repository starvation
- Closed
- mentioned in
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...