Observium CPU Usage

PrincessOfCats

Does anyone have any idea how much CPU Observium is supposed to use?

Seems a bit high to me:

Load (15-min avg) is 0.45-0.35, with 20 devices and 85 ports. rrdcached is enabled. CPU is a Intel(R) Xeon(R) CPU E3-1230 V2 @ 3.30GHz

Side note: I enabled observium Friday morning, so the first part of the graph can be ignored.

Edit: Corrected load with wrong decimal position

cassa

Much.

linuxthefish

SNMP uses quite a lot yeah, but if you are having CPU issues with your install you could reduce the number of hosts being polled in the crontab.

PrincessOfCats

@linuxthefish said:
SNMP uses quite a lot yeah, but if you are having CPU issues with your install you could reduce the number of hosts being polled in the crontab.

This is already a reduction from 10->5, should I be reducing it more?

This is my monitoring server, so I'm looking to run a few other things on there (Icinga2, nfsen)

MarkTurner

SSDs alleviate some of this pain. We have a few Observium instances spun up with a good few thousand elements and distributed collection and SSDs moved it from sluggish to performant again.

PrincessOfCats

Huh.

I guess that this is the load it's supposed to be at. This server doesn't have SSDs. I will just wait a bit to see if I can find a SSD offer with a host in Incero. Meanwhile, I will turn down the number of pollers to see if I can still get data without missing points. I have a few slow-to-respond servers, mainly due to distance.

MarkTurner

What does iostat show? Does it show IOWAIT?

PrincessOfCats

Linux 3.13.0-24-generic (Sarah)         04/03/2015      _x86_64_        (1 CPU)

avg-cpu:  %user   %nice %system %iowait  %steal   %idle
           2.54    0.01    1.97    0.58    0.04   94.86

Device:            tps    kB_read/s    kB_wrtn/s    kB_read    kB_wrtn
xvdap1           13.28         2.21       103.61     758637   35543708
xvdap2            0.00         0.01         0.05       4724      17380

Thinking of it now, it shouldn't be very high due to use of rrdcached.

MarkTurner

For comparison:

Linux 2.6.32-358.el6.x86_64 (observium07-slc01)     03/04/15    x86_64    (24 CPU)

avg-cpu:  %user   %nice %system %iowait  %steal   %idle

           7.88    0.00   10.92    0.24    0.00   80.96

Device:         rrqm/s   wrqm/s     r/s     w/s   rsec/s   wsec/s avgrq-sz avgqu-sz   await  svctm  %util

sdb               0.00     0.72    0.01    0.01     1.58     5.77   473.79     0.01  353.95   3.46   0.01

sda               0.00    88.24    0.03  133.19     1.04  1686.33    12.67     0.08    0.62   0.31   4.14

PrincessOfCats

In that case, SSDs look like it could pull the load down a bit more than what I have currently.

For now, I've yanked out two more pollers to sit at 3 now, and the load is now a bit more respectable until I can find a nice offer.

MarkTurner

@StartledPhoenix - what is your budget for this? Maybe able to get you a low cost dedicated with SSD.

FrankZ

@StartledPhoenix - Just trying to offer perspective. I have 18 devices with 62 ports running a ~15% cpu load on a crissic 256mb with 2 cores. Load average of .26 .14 .10. No SSD. I have had no issues for many months.

s10.postimg.org/7y2ccbvu1/obserload.jpg

PrincessOfCats

@MarkTurner said:
StartledPhoenix - what is your budget for this? Maybe able to get you a low cost dedicated with SSD.

I'm sure you can't offer anything in my budget - this is a $5 VPS

PrincessOfCats

@FrankZ said:
StartledPhoenix - Just trying to offer perspective. I have 18 devices with 62 ports running a ~15% cpu load on a crissic 256mb with 2 cores. Load average of .26 .14 .10. No SSD. I have had no issues for many months.

s10.postimg.org/7y2ccbvu1/obserload.jpg

How many pollers are you using?

I have 3, so my load average of 0.28, 0.43, 0.35 is around right

FrankZ

How many pollers are you using?

two

J1021

Who are you hosting this with? Does your host provide CPU usage graphs?

PrincessOfCats

HostSumo, and no, they don't provide CPU usage graphs.