Since about two weeks we can regularly see Bitbucket Server (v7.3.2) create 100% CPU utilization on all of the 8 cores the VM running it has. The website is not accessible when this occurs, even when waiting for 10 minutes. The only way to make it accessible is by restarting the whole application.
One of the last times it happened I was able to connect to the application via JProfiler an get a thread dump. I could see multiple threads running in what seemed to me like an infinite loop. All the threads in org.hibernate.cache.spi.support.SimpleTimestamper.next() (line:38) where running.
As it seems to me that the problem has something to do with user directories I already replaced the direct LDAP directory connection with one that uses our Jira server to access the LDAP directory indirectly, but that did unfortunately not help.
After that I looked into the actual hibernate code which indeed runs in an infinite loop (although I do not really get why every thread says line 38 and not any of the other lines in the loop). Regarding this exact method I found Hibernate issue HHH-13664, but by monitoring the system time when the issue occurred another time I could rule out, that our servers time was set to any time in the past.
Right now we have to restart the application about 2 or 3 times a day (in about 9 office hours).
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.