iowait graphics

[mtsm]

Does anyone know how to generate an IOWAIT graphic with cacti..im using linux and solaris (via snmp)

thanks

[gandalf]

Yep, we're doing so. What parms are you interested in?
Reinhard

[mtsm]

how do i generate the iowait graphic? i just have cpu usage and load average.....i want an iowait graphic (io activity) i could do that in mrtg..

[mtsm]

it would be the wa% field in linux top and iowait% field in solaris top...

[gandalf]

You may execute vmstat or iostat as well
Reinhard

[mtsm]

sure...but i'd like to do that via snmp...i've got some 200 machines to monitor...and would be easier than creating shell scripts.. is there any template that uses snmp to monitor iowait?

[gandalf]

No, as there is not SNMP OID for IOWAIT, AFAIK. That's why I referred to iostat/vmstat
Reinhard

[mtsm]

Nice....i knew about the snmp oid , the problem is..as im new to cacti i dunno how to build a template for iowait with snmp , do you know if there's one ready to use over here??? i've searched the forum but just found otemplates that use scripts using iostat/vmstat commands ..one of them was by yourself i belive.

thanks again

[gandalf]

... i dunno how to build a template for iowait with snmp , do you know if there's one ready to use over here?

You've missed it. AFAIK, there is no OID for IOWAIT, so it is not possible to get IOWAIT data via SNMP. This is not cacti related. You would have to build your own SNMP agent to report IOWAIT data.
Reinhard

[mtsm]

the old admin from here was graphing iowait with snmp with mrtg like this:

Target[tetis.cpuiw]:.1.3.6.1.4.1.2021.11.53.0&.1.3.6.1.4.1.2021.11.54.0:l4mb1d4@tetis
MaxBytes[tetis.cpuiw]: 100
Title[tetis.cpuiw]: CPU Idle X IOWait - tetis

i wanted to get a similar graph with cacti...

isnt this the real iowait graph?

[gandalf]

Oops, seems I've to correct myself

Code: Select all

snmpwalk -c <community string> -v 1  localhost .1.3.6.1.4.1.2021.11
...
UCD-SNMP-MIB::ssCpuRawUser.0 = Counter32: 10890
UCD-SNMP-MIB::ssCpuRawNice.0 = Counter32: 358
UCD-SNMP-MIB::ssCpuRawSystem.0 = Counter32: 2836
UCD-SNMP-MIB::ssCpuRawIdle.0 = Counter32: 74599
UCD-SNMP-MIB::ssCpuRawWait.0 = Counter32: 15059
UCD-SNMP-MIB::ssCpuRawKernel.0 = Counter32: 2696
UCD-SNMP-MIB::ssCpuRawInterrupt.0 = Counter32: 140
...

From this, it should be possible to develop the wanted template or to extend the already existing one (IOWAIT = Wait???)
Reinhard

[mtsm]

is there already an existing one ? (iowait = wait in linux)...

[gandalf]

AFAIK, no. But you may extend the Graph Template ucd/net - CPU Usage and add the needed Data Templates similar to ucd/net - CPU Usage - Nice
Reinhard

[mtsm]

nice!! it's working...i extended ucd/net cpu usage as you suggested...the problem now is: how to add the iowait to all my hosts? do i have to add by hand the new data source (iowait) host by host? is there any trick i could do?

[gandalf]

Wow, fine! In this case, you're lucky cause all OIDs are stored in separate rrd files. Otherwise, it would be more than tricky ...
For all newly defined graphs, there should be no problem.
You said, you modified the existing Graph Template? And added the needed Data Templates just like the existing ones? That's important, so please confirm this assumption.

Then, for each host, perform the following:
Goto Data Sources, filter your host, Add a new Data Source and select the "ucd/net WAIT" one and "Create" to get this OID polled for that device
Then, goto Graph Management and select the already existing Graph. You'll find one of the Graph Items set to none (the new one!). select the WAIT Data Source from the dropdown and "Save".

Of course, you'll have to wait 2 polling cycles for data to get filled. And you surely will have no historical data for this item
Reinhard