[jffnms-users] Still have crazy JFFNMS slowness

[Leigh Porter]

Title: Still have crazy JFFNMS slowness

Hiya all,

I still have this box with crazy JFFNMS slowness. Take a look at this! :

 Alarms:         down: 63
                Hosts:  57      
        Interfaces:     2055

  3:37am  up 11 days 18:07,  1 user,  load average: 33.31, 33.08, 33.86

Polling on some hosts is taking upto 13 minutes!

The box is a Duel processor 3.0Ghz Xeon with 2Gb memory. I'm using MySQL for the database
and a 2.6.11 kernel (SuSE 9.3 box)

I have seperate disks for the MySQL database and the rest of JFFNMS (i.e. rrd). The JFFNMS
disks are a couple of RAID0 fast SCSI disks on a fast RAID controller.

Here are the iostats:

jffnms:/proc # iostat
Linux 2.6.11.4-20a-smp (jffnms)         04/12/06

avg-cpu:  %user   %nice    %sys %iowait   %idle
          73.51    0.00   14.54    0.36   11.59

Device:            tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn
fd0               0.00         0.00         0.00          6          0
sda              10.25         2.10       184.38    2132000  187296360
sdb             133.05        39.99      1386.92   40624856 1408857488
hdc               0.00         0.00         0.00        152          0

There are a lot of tps towards the sdb disk (this is where jffnms lives) and the Block writes
per second are high, but I would not have thought that this would cause the slowness I see.

I see other people on the list monitoring large networks on much smaller hardware, whatever
could be happening here!?

I don't really want to throw more hardware at it until I know where to throw it at. Perhaps
moving MySQL onto a seperate box would help, but then it does not look as though MySQL is
that busy anyway.

If anybody has any ideas, it'd be appreciated!

Thanks,
Leigh Porter