Red Hat Bugzilla – Bug 89249
size-512 in /proc/slabinfo keeps growing and machine dies after 10 day
Last modified: 2007-04-18 12:53:11 EDT
Description of problem:
Machine serves http/dns/smtp. After 10 days, services stop responding while
ping is still ok. fork/vfork error appear in the logs. Machine ends
unusable. apache is configured to regularly restart and no process shows
significant memory usage.
Version-Release number of selected component (if applicable):
The machine should stop within 2/3 days for the third time.
Steps to Reproduce:
so something is leaking memory; and since you're the first report of this I have
to assume it's something uncommon in your configuration; could you post lsmod
output and maybe describe things that could be uncommon in your setup ?
Created attachment 91214 [details]
slab-512 output samples for 25 minutes
Created attachment 91215 [details]
Created attachment 91216 [details]
Created attachment 91217 [details]
ps auwx output
Created attachment 91218 [details]
Created attachment 91219 [details]
Created attachment 91220 [details]
sch_sfq 4288 6 (autoclean)
cls_u32 5444 1 (autoclean)
sch_cbq 13248 1 (autoclean)
are 3 modules that are "uncommon" a bit; are you actively using them?
The machine is limiting bandwidth at ~1.8Mbps. Setup script sent as attachment.
Created attachment 91221 [details]
> sch_sfq 4288 6 (autoclean)
> cls_u32 5444 1 (autoclean)
> sch_cbq 13248 1 (autoclean)
> are 3 modules that are "uncommon" a bit; are you actively using them?
Yes, see http://pulsar.zoreil.com/rrdtool/eth0-weekly.gif.
Previous setup (2.4.18-3) didn't show these symptoms but I can't claim that
the bandwidth was limited in the exact same way.