[Ldapdns-users] ldapaxfr eating copious CPU and mem

[James W. (S. Sysadmin) <sys...@su...>]

G'day,

ldapaxfr just up and made off with all my CPU and memory.  No TCP port 
53 traffic was going through at the time so it's unlikely to have been 
related to any axfr that was actually in progress.  The issue was 
noticed when an in-house script that searches LDAP and receives around 
400 entries each with one attribute specified in the result was left 
waiting forever.  Nothing else LDAP-related appeared to be affected 
adversely which may have been due to the use of indices.  Here's a 
snippet from top:

top - 10:12:08 up 135 days, 11:10,  4 users,  load average: 8.51, 7.82, 6.72
Tasks:  88 total,   3 running,  85 sleeping,   0 stopped,   0 zombie
Cpu(s): 95.7% us,  2.7% sy,  0.0% ni,  0.0% id,  0.0% wa,  0.0% hi,  1.7% si
Mem:    906736k total,   901652k used,     5084k free,     1120k buffers
Swap:  2891576k total,   742552k used,  2149024k free,    13108k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  
COMMAND          
  25213 ldapdns   25   0 1021m  740 3780 R 84.1  0.1 841:48.65 ldapaxfr  

Restarting ldapdns restored everything to normal.

Sorry I can't provide any more info at this time, and not sure I can 
replicate this in a dev. environment.  I'm suspicious of the openldap 
libs though haven't seen this particular behaviour from anything else 
dependant on them; or could it possibly be in the ldapaxfr code?

I'm running 2.06 as packaged in Debian Sarge.

Cheers,

-- 
James Wakefield
Systems Administrator
+61 03 5227 6888

We have now moved head office to 8-12 Pakington Street,
Geelong West.