Disk I/O on RH Linux Systems

Our sysads would like to see a five minute average of disk I/O; reads and writes in bytes/second.

RHEL 5.3 w/ net-snmp makes disk I/O data available via the UCD-DISKIO-MIB as Counter data (number of bytes read from device since boot, etc.), but not as averages.

We thought about making a universal poller to poll disk reads and writes every 5 minutes and then use a transform to compute (current poller reading minus previous poller reading) but could not figure it out.

Any ideas?

Find more posts tagged with

SNMP

disk_io

net-snmp

apm_2.0

Accepted answers

All comments

Craig651

How come I keep seeing my question in here with no answers. Is there no solution to this?

FormerMember

How come I keep seeing my question in here with no answers. Is there no solution to this?

You haven't asked a question. The previous poster (which I guess might be another alias of yours) did ask a question. It's hard to say why some questions don't get an answer. Sometimes its because no one knows. Sometimes, it's because it's too open-ended.

You ask:

I think that might work, but really, I don't think I know enough about what you've done to comment more than that.

CaptainOrion

Denny, I am not Craig651. I for one really do want to get this figured out.

I just cannot figure out how to use the universal poller to be able to take a sample at time period 0 (a value of say 5 disk reads) then another sample 5 minutes later (perhaps 12 disk reads) and then calculate (12 reads minus 5 reads) divided by 5 minutes to get 1.4 reads/minute. The delta is the new reading minus the previous reading combined with time or, as a percent change the delta is the new reading minus the previous reading divided by the previous reading. But I cannot get this to work since the counters for disk read and disk write only go up in value and then roll over. The counters are disk reads since boot, or disk writes since boot.

The system admins here do this with RRDB, we are trying to do this with APM.

Thanks

Craig651

Denny, I found that response slightly offensive.

I was simply stating that I posted a similar question to the one that CaptainOrion had posted in another thread and he did not get a response. I don't think I invited that response, especially from a member of the solarwinds staff

FormerMember

I just cannot figure out how to use the universal poller to be able to take a sample at time period 0 (a value of say 5 disk reads) then another sample 5 minutes later (perhaps 12 disk reads) and then calculate (12 reads minus 5 reads) divided by 5 minutes to get 1.4 reads/minute. The delta is the new reading minus the previous reading combined with time or, as a percent change the delta is the new reading minus the previous reading divided by the previous reading. But I cannot get this to work since the counters for disk read and disk write only go up in value and then roll over. The counters are disk reads since boot, or disk writes since boot.

To do this with the UnDP, you need to set it as a counter type. You can do this under "Show Advanced Options" on the first page of the UnDP wizard.

c.gura