[collectd] loss of data on collectd-server

eric fauser ef_cd<span style="display: none;">.trailing-username</span>(a)<span style="display: none;">leading-domain.</span>apa.at
Mon Jan 8 12:45:31 CET 2007


> How much
> traffic do you recieve from collectd clients?

the machine receives 48 collectd-streams, so ~500kBit networkIO generates 
2MByte of random IO
to the disk-subsystem, which consits of 6x36GB 10k Raid5 and the cpu reports 
~17% IO-waitung

((the old system (same hardware) used only a mirrored 15k 148GB Disks which 
generated 35% IO-waitung))

are there other users with similar system ?

> This is weird, too: The same timestep is sent twice, which
> shouldn't happen either. However the values differ, so the
> problem is not the same packet being received twice. Do you
> have more than one instance running on that host?

(no there is only one instance, but...)
a veritas-clusters runs on this machine using fibre-channel connected disks, 
use /dev/vx as a (tmpfs)-mapping-service and also df shows  2x /dev/vx 
devices :(
(btw. no vx-devices are listed in /etc/fstab, so i think a kernel-module 
from the veritas-cluster do this)

                      70376888   6837704  59964196  11% /
tmpfs                        4         0         4   0% /dev/vx
tmpfs                        4         0         4   0% /dev/vx
                      41445944  29669832   9670768  76% /appl
/dev/cciss/c0d0p1       101086     19828     76039  21% /boot
none                   2074632         0   2074632   0% /dev/shm
                      31453184  25359747   5712674  82% /appl/oraDB

the only bad thing is, that rrd_update reports this every 10sec to the 
syslog ;(

br eric

btw: i should change the thread-subject to "lots of data..." ;) 

More information about the collectd mailing list