-
Bug
-
Resolution: Fixed
-
Critical
-
None
-
Platform: All, OS: All
Hi,
recently we started having problems with projects, which have SCM polling
configured but never get triggered. When I looked at the polling items list
(SCMTrigger.DESCRIPTOR.getItemsBeingPolled()) I saw dozens of threads waiting
for a reply from remote computer in workspace.exists(). This is my theory about
what happened (might be wrong since I am not much familiar with Hudson remoting
stuff):
1) Polling thread asked the slave whether the project workspace exists;
2) Slave got disconnected for whatever reason;
3) Slave got online and started working, but didn't know about the polling
thread's request anymore;
4) Polling thread remained waiting, until the end of the world.
All in all, this isn't very convenient since polling threads block scm
operations for the project (and it is very confusing for users because of lack
of any output).