Hi Simone,
On 26 Nov 2010, at 14:06, Simone Felici wrote:
THe last days I've noticed high load on opsview-web process and in
addiction a HIGH traffic between master server and database server.
I'm able to reproduce the issue, maybe someone can help me to solve.
It seems the problem has something to do with snmptrap-debug. I've
enabled 115 service checks on 3 hosts (every host has 115
service checks) with snmptrap. Enabling the snmptrap-debug on one
host for some time it has collected me ~4000 traps. When I've
finished to debug the issues, I've deleted the debug traps from web
interface with the button: "Delete all debug".
The first problem is that there isn't a page that inform me a delete
is started and I return to the debug page with all my debug
traps. This could point me to repress the button again, involving
maybe more delete routines.
BTW, pressing only one time, this action has started a LOT of sql
commands, logged from db server, in this art:
22 Query SELECT trap, servicecheck, expandedrule, result
FROM snmptrapruledebug WHERE id='<id>'
22 Query DELETE FROM snmptrapruledebug WHERE id='<id>'
The snmptrapruledebug was popolate with ~64.000 records. Innodb
hasn't a structure to maintain precise number of rows. The last id
was 987263 but it's incremental, also we suppose the 64.000 records
are exact.
It seems the delete routine is not optimized. It's impossible the
delete of the debug table causes me a traffic of 10MBit/s. See
the images attached.
The first (biggest) block of traffic (10MBit/s peak) is generated
pressing multiple times the delete routine. The second block
pressing the delete routine only once.
I've stopped the first block of traffic stopping opsview-web
service. Then, I've started it again with mysql debug on and pressed
the delete routine only one time to log a little bit.
Could it be a right reason of this cause?
Is there a better way to handle this or is this known as an issue?
(should be i think!)
We've been aware of something around here, but we haven't got round to
working on this. However, you've clearly demonstrated that this has a
big impact, so I've raised the priority of this issue and we'll look
to get it fixed as part of 3.11 release.
https://secure.opsera.com/jira/browse/OPS-1119
Ton
_______________________________________________
Opsview-users mailing list
[email protected]
http://lists.opsview.org/lists/listinfo/opsview-users
