bind 9.3.2

Rok Potočnik r at rula.net
Fri Feb 8 23:14:35 UTC 2008


Hello! I'm a sysadmin for a couple of authoritive & caching name 
servers. There's one box that has some issues...

HP ProLiant DL360 G4
2x Intel Xeon 3GHz - not dual core, just HT :/
2x 1GB DDR 333MHz RAM
2x 72GB SAS in HW RAID 1
running SLES 9.3, Bind 9.3.2
serving ~1k zones, ~60k clients (caching) with ~1k requests per second, 
both authoritive and caching

The problem is that named for an unknown reason hogs the CPU every now 
and then... Sometimes twice per day, sometimes once per couple of days...
The CPU load increments somewhat linear 'till it reaches 1, that's when 
the server becomes kinda sluggish and a lot of users complain about it. 
Restarting named works every time, that's when the cpu load is around 
0.1 - 0.3. It's using about 25% of RAM and I don't think it's a memory 
related issue. The problem was there before adding the second CPU so i 
don't think it's a CPU issue. I ran out of ideas why is that. It's a 
production server and I can't do much testing. I could run gdb and/or 
strace but I'm not that good at debugging with that tools. It would be 
important to resolve this ASAP without too intrusive techniques (like I 
said, it's a production server, so I should keep downtime at minimum).

Any ideas?

-- 
Rok



More information about the bind-users mailing list