slapd becomes non-responsive after several weeks runtime
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
openldap (Ubuntu) |
Expired
|
Low
|
Unassigned |
Bug Description
I use an Ubuntu 10.04 (lucid) server for, among other things, LDAP authentication.
slapd:
Installed: 2.4.21-0ubuntu5
Candidate: 2.4.21-0ubuntu5
Version table:
*** 2.4.21-0ubuntu5 0
500 http://
500 http://
100 /var/lib/
Distributor ID: Ubuntu
Description: Ubuntu 10.04 LTS
Release: 10.04
Codename: lucid
With Ubuntu 9.10 it was stable and well-behaved. Since upgrading to 10.04, however, I've been seeing issues where slapd stops responding to ldap queries after several weeks of uptime. slapd has to be killed and re-launched to get it working again.
I haven't found any way to reproduce this except waiting. The delay before it happens is variable, from a few days to a few weeks.
All other daemons on the server are stable and well behaved. The issue appears to be restricted to slapd.
After the second time this happened, I tried to dump a core from slapd so I could debug it after I'd restarted it to restore services. However, gcore reported the following error:
/root/slapd.
Cannot access memory at address 0x1fe8513
gcore: failed to create slapd.core.25525
and when I attached gdb directly to attempt to just get a backtrace, it failed in a similar manner:
Attaching to process 25525
Reading symbols from /usr/sbin/
done.
Cannot access memory at address 0x1fe8513
... making it rather hard to collect debug information. A backtrace requested after that error is essentially useless, as it contains no symbols, only '???' and addresses.
This time when it happened I collected some information from slapd's /proc entry before re-starting the process. It's all I could come up with, and it didn't tell me much. I'll attach the files from /proc here.
Ideas? Can this be caused by a slapd bug? Or should I be looking for a kernel bug?
Changed in openldap (Ubuntu): | |
importance: | Undecided → Low |
status: | New → Incomplete |
The attachment contains /proc/$ pid/{environ, maps,pagemap, smaps,status} from a failed slapd instance.
I couldn't read /proc/$pid/mem; it reported "no such process".