[Gllug] Interrupt storm

[Paul M]

One of my servers was faced with surges of a really heavy load this
morning with the load average peaking at >10.  They went on for 10
minutes or so at time before tailing off gradually and the whole
episode lasted maybe 1 or 2 hours. I've been keeping a log of the load
average for a while and it appears we have had very occasional times
when the load average has gone up to 3 or so, but nothing this high.
It cleared up by itself eventually and things are back to normal.
Monitoring (using dstat -aim) showed that nic throughput, cpu and
memory usage remained steady and as normal throughout but the inerrupt
rate went sky high. Nothing has showed up in the logs.  Does anyone
have any idea what might have caused this, or how  I can isolate the
problem any further?

The machine is a 2.Ghz pentium with 1/2 a gig of Ram and serves as a
mail and DNS server. It serves the DNS for about 100 sites and has
maybe 30 email accounts on it.
It runs Debain Sarge, Tinydns, Postfix (with Postgrey and Spamassasin)
serving mail via Popa3d for most of the accounts with 8 of the heavily
used accounts served via ssl over imap using uw-imapd. It also has
Apache-ssl running to provide webmail but 3 users at once would be
exceptional. Theres a snapshot of dstat below,  though under normal
conditions only).

thanks,

Paul M



-cpu-total- -net/total- ---paging-- ---system-- ---load-avg---
----interrupts--- ------memory-usage----- ---time---
usr sys idl|_recv _send|__in_ _out_|_int_ _csw_|_1m_ _5m_ 15m_|__4__
__14_ __16_|_used _buff _cach _free
  1   1  99|2662B 2745B|   0     0 | 130   104 |0.03 0.03    0|   0
 0    30 | 198M 18.9M  247M 8148k
  2   1  98|1689B 6835B|   0     0 | 190   105 |0.03 0.03    0|   0
58    32 | 198M 18.9M  247M 8120k
  1   1  98|3717B 4610B|   0     0 | 193   164 |0.03 0.03    0|   0
 0    93 | 198M 18.9M  247M 8120k
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug