From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7.5) Gecko/20041107 Firefox/1.0 Description of problem: After upgrading from kernel-2.6.10-1.11_FC2smp to 1.14 we experienced that under heavy load the system became unstable, because of too many processes stacking up. In peak times this happened quite frequently, resulting in urgent needs to e.g. shutdown the database to save the machine from crashing. The machine is running 2 processors in hyperthreading mode (meaning 4 cpus for smp). It's running both database and, for historical reasons, NFS-server. Of 2GB RAM there was a bit less than 1GB taken for MySQL, which was planned and worked fairly well before. But even after shutting down MySQL, there was still 1GB of RAM taken for _something_. We didn't find out where it went, actually. During rescue-operations to the database even sshd-processes got killed and in /var/log/messages kernel said it was "out of memory". Since running out of memory the system did not have enough resources for caching and the like, so it became quite slow and more processes stacked up - boosting the system from 300/400 to about 900 processes within minutes sometimes. Sometimes the system was able to run with very low remaining RAM - but only while it was serving with very light load. These problems disappeared, as it seems until now, after rebooting with kernel 1.11. Are there any known "memory-eaters" introduced between 1.11 and 1.14? Version-Release number of selected component (if applicable): kernel-2.6.10-1.14_FC2smp How reproducible: Always Steps to Reproduce: 1.Upgrade system from 1.11 to 1.14 2.Set up database and NFS. 3.Bring system under load. Actual Results: Even under medium load, after a few hours, we could see memory "disappearing" somewhere. Additional info:
yes, there is one known leak in the latest kernel, I'll be putting out another fc2 soon fixing this and a few other problems.
Does this also affect fc3? If it was known, could you maybe send some warning through the announcement-list for not too many additional people running into similar problems?
Dave, does the release of yesterday / today (1.770_FC2 / FC3) fix this issue as well?
hopefully. if it doesn't, I'm unaware of any leaks right now. if you still get OOM kills with that kernel, it could point to a more fundamental problem with the VM.
Dave, it worked for us so far. Afaik we can declare this one fixed.
Fedora Core 2 has now reached end of life, and no further updates will be provided by Red Hat. The Fedora legacy project will be producing further kernel updates for security problems only. If this bug has not been fixed in the latest Fedora Core 2 update kernel, please try to reproduce it under Fedora Core 3, and reopen if necessary, changing the product version accordingly. Thank you.