Comment 4 for bug 1323165

Revision history for this message
Nikhil Chhaochharia (nikhilchh) wrote :

We have some more information which may help.

Once this problem happens, then the system goes into some kind of unstable state. Existing console or ssh sessions continue working as long as we don't try to access the java process. However, new SSH sessions don't start. Logging in from the console leads to the server information getting displayed but the command prompt does not show up after that.

The task that we execute on the server has a setting for number of threads to use and it is set to 16 by default which consistently leads to this bug after 24-48 hours of processing. We tried to run the same task with 8 threads and it has been running without any problem for days.

We have multiple servers with the exact same hardware and software where we are seeing this bug. We have downgraded one of them to Ubuntu 12.04.0 and that server has been working fine even with 16 threads. We have now upgraded the other servers to the new kernel released yesterday (3.13.0-27) and will report back if the issue is fixed there. If not, then we will try the latest mainline kernel.