Im going to fire up the server again today and test more, will let you know,
if there anyway of running the poller and but not writing to the DB, I see there is a flag to disable the writing of RRDs, I want to see if the poller performance is the issue, the disk I/O or the DB
On 25 Jul 2016, at 00:30, Adam Armstrong <adama@memetic.orgmailto:adama@memetic.org> wrote:
If do you have sufficient buffers and caching for innodb in mysql?
We have some /very/ large installs which saw little impact from the changes.
Adam.
Sent from BlueMailhttp://www.bluemail.me/r
On 24 Jul 2016, at 20:55, Lyndon Labuschagne <lyndonl@mobiletorque.co.zamailto:lyndonl@mobiletorque.co.za> wrote:
I have been performance tuning all day with little to no affect
its just very strange how a XEN VM out performs bare metal, it may also be Ubuntu 16.04 as that is on the bare metal server and 15.04 is on the VM
On 24 Jul 2016, at 21:47, Adam Armstrong <adama@memetic.orgmailto:adama@memetic.org> wrote:
We've changed the way the ports table works, removing memory tables.
It's likely that you now need more mysql resources, since were now writing to an innodb table instead of a memory table.
Odd, really. It didn't seem to have a huge effect on my dev install or some other very large installs.
Perhaps you could investigate mysql performance tuning? :)
Adam.
Sent from BlueMail
On 24 Jul 2016, at 19:58, Lyndon Labuschagne <lyndonl@mobiletorque.co.zamailto:lyndonl@mobiletorque.co.za> wrote: i went back to use the old database server on the VM and all was much better, running pollers at 10 to 15 seconds per device, until I upgraded the db by using discovery -u
back in the same boat again
On 24 Jul 2016, at 09:56, Lyndon Labuschagne <lyndonl@mobiletorque.co.zamailto:lyndonl@mobiletorque.co.za> wrote:
Ok if I poll a router / switch / server etc it seems to be ok, a full scan on a 7606 takes around 45 seconds,
if I poll a radio, Ubiquiti Airfibre/ M5 . Rocket etc it seems to be slow, even when the radio is directly connected to the hosting switch as a test. The issue pops up when doing the ifindex - ifIndex: 1 2 3 5 6 7 8 9 (that takes 40 to 50 seconds)
o Caching Oids ifDescr ifType ifOperStatus ifAlias ifName ifHighSpeed ifInMulticastPkts ifOutMulticastPkts ifInBroadcastPkts ifOutBroadcastPkts ifHCInOctets ifHCOutOctets ifHCInUcastPkts ifHCOutUcastPkts ifHCInMulticastPkts ifHCOutMulticastPkts ifHCInBroadcastPkts ifHCOutBroadcastPkts ifInOctets ifOutOctets ifInUcastPkts ifOutUcastPkts ifInNUcastPkts ifOutNUcastPkts ifInDiscards ifOutDiscards ifInErrors ifOutErrors ifInUnknownProtos ifMtu ifSpeed ifPhysAddress ifAdminStatus ifLastChange ifPromiscuousMode ifConnectorPresent, ifIndex: 1 2 3 5 6 7 8 9 ADSL-LINE-MIB EtherLike-MIB dot3StatsDuplexStatus
The the other thing that had changed is that I was running a slightly older version of Observium on the VM server, still in the 79xx range not the 8003 version
On 24 Jul 2016, at 09:16, Lyndon Labuschagne <lyndonl@mobiletorque.co.zamailto:lyndonl@mobiletorque.co.za> wrote:
I know its bad form to reply to your own email but here is some more information re drive performance
root@observium:/opt/observium/rrd# hdparm -Tt /dev/mapper/observium--vg-root
/dev/mapper/observium--vg-root: Timing cached reads: 14428 MB in 2.00 seconds = 7218.63 MB/sec Timing buffered disk reads: 950 MB in 3.01 seconds = 315.94 MB/sec root@observium:/opt/observium/rrd# dd if=/dev/zero of=/tmp/output bs=8k count=10k; rm -f /tmp/output 10240+0 records in 10240+0 records out 83886080 bytes (84 MB, 80 MiB) copied, 0.125735 s, 667 MB/s
On 24 Jul 2016, at 09:13, Lyndon Labuschagne <lyndonl@mobiletorque.co.zamailto:lyndonl@mobiletorque.co.za> wrote:
Hi All
I have moved from a VM server to a physical server. as the VM was taking a bit of a pounding, graph dropouts etc The physical server is a dual quad core Xeon, 2.66Ghz with 128GB of RAM and a 32GB ramdisk, and rrdcache I also have 2.5TB or SAS storage in a raid6 config
Will all of this my average poller-wrapper time is around 60 to 70 seconds per device if I manually run an snmpwalk a single device takes around 6 seconds
I have also disabled the ramdisk, rrdcached etc and I am not seeing much of a difference either way
I have around 900 devices and the VM server was way faster, any idea where to start looking for a bottleneck?
Regards Lyndon
observium mailing list observium@observium.orgmailto:observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
observium mailing list observium@observium.orgmailto:observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
observium mailing list observium@observium.orgmailto:observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
observium mailing list observium@observium.orgmailto:observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium ________________________________
observium mailing list observium@observium.orgmailto:observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
________________________________
observium mailing list observium@observium.orgmailto:observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
_______________________________________________ observium mailing list observium@observium.orgmailto:observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium