You know to be honest, I run into this a lot. I monitor many metrics that I don't think have a known good value or bad value. In those cases I just trend it. If no one is barking about performance I assume the way its been running is acceptable and watch for activity that is outside the normal trend. In one datacenter I monitored air flow by measuring the voltage produced by a modified pc fan placed in the air stream, as long as I had around 8.5 volts of air I knew it was running normally. I couldn't tell you the CFM or barometric pressure developed under the floor but I could tell you when one of the belts needed replaced or the filters were getting dirty.
So store up a few weeks of normal load, make notes of times things seemed slow, try to correlate that to the I/O and take it from there. If you have the time figure out what the theoretical max is and stress your system to see how close you can get to that max. I guess it really depends on what you intend to use the data for.
ucd/net snmp Device (Disk) I/O templates - Updated - v3.1
Moderators: Developers, Moderators
-
- Posts: 24
- Joined: Fri Aug 05, 2011 9:21 am
Re: ucd/net snmp Device (Disk) I/O templates - Updated - v3.
good answer, thanks.psyber wrote:You know to be honest, I run into this a lot. I monitor many metrics that I don't think have a known good value or bad value. In those cases I just trend it. If no one is barking about performance I assume the way its been running is acceptable and watch for activity that is outside the normal trend. In one datacenter I monitored air flow by measuring the voltage produced by a modified pc fan placed in the air stream, as long as I had around 8.5 volts of air I knew it was running normally. I couldn't tell you the CFM or barometric pressure developed under the floor but I could tell you when one of the belts needed replaced or the filters were getting dirty.
So store up a few weeks of normal load, make notes of times things seemed slow, try to correlate that to the I/O and take it from there. If you have the time figure out what the theoretical max is and stress your system to see how close you can get to that max. I guess it really depends on what you intend to use the data for.
some days I use my VPS to do something and some days I use my VPS to do something else.
I'm curious to know what kind of tasks requires more I/O and I wanted to graph it with cacti.
As far as I'm understanding there is no way to do that.
Re: ucd/net snmp Device (Disk) I/O templates - Updated - v3.
Actually there are many ways to measure I/O.
SNMP is one way (the way this template works):
http://oss.oetiker.ch/mrtg-trac/wiki/Net-SNMP
something like iozone is another way
http://www.thegeekstuff.com/2011/05/iozone-examples/
http://blog.dbadojo.com/2007/10/iozone- ... s-ec2.html
There's also bonnie++
http://www.coker.com.au/bonnie++/
For quick peeks at whats going on you have
vmstat
iotop
If your talking about a vps at some big hosting facility like rackspace your backing is probably a SAN or FAS or NAS, so your I/O may be somewhat dependent on what all the other vps's in their datacenter are doing at that moment in time, how their network is performing, how much of the cache is hit or miss, ... which is something you just can't know without having direct access to that back end storage. Results of trying to measure I/O and correlate that to something you're doing may seem a bit erratic.
SNMP is one way (the way this template works):
http://oss.oetiker.ch/mrtg-trac/wiki/Net-SNMP
something like iozone is another way
http://www.thegeekstuff.com/2011/05/iozone-examples/
http://blog.dbadojo.com/2007/10/iozone- ... s-ec2.html
There's also bonnie++
http://www.coker.com.au/bonnie++/
For quick peeks at whats going on you have
vmstat
iotop
If your talking about a vps at some big hosting facility like rackspace your backing is probably a SAN or FAS or NAS, so your I/O may be somewhat dependent on what all the other vps's in their datacenter are doing at that moment in time, how their network is performing, how much of the cache is hit or miss, ... which is something you just can't know without having direct access to that back end storage. Results of trying to measure I/O and correlate that to something you're doing may seem a bit erratic.
Re: ucd/net snmp Device (Disk) I/O templates - Updated - v3.
Hi !
I installed cacti 0.8.8a on a Ubuntu 11.04 server 2 weeks ago. I monitore different devices (pfSense v2 routers, Windows servers, Linux servers, dell PowerConnect (switchs), etc.).
Today I added the Gandalf's IO disk script
It work for ubuntu server, but not for windows server and pfSense (freeBSD).
I obtain "Success [0 Items, 0 Rows]" after added "SNMP - Get Disk IO" data query for this device.
This is the verbose query :
Do you have any idea?
thanks
I installed cacti 0.8.8a on a Ubuntu 11.04 server 2 weeks ago. I monitore different devices (pfSense v2 routers, Windows servers, Linux servers, dell PowerConnect (switchs), etc.).
Today I added the Gandalf's IO disk script
It work for ubuntu server, but not for windows server and pfSense (freeBSD).
I obtain "Success [0 Items, 0 Rows]" after added "SNMP - Get Disk IO" data query for this device.
This is the verbose query :
I understand there is no data for the OID 1.3.6.1.... but I don't know/understand how to make to work well with all device.+ Running data query [10].
+ Found type = '3' [SNMP Query].
+ Found data query XML file at '/var/www/cacti/resource/snmp_queries/disk_io.xml'
+ XML file parsed ok.
+ <oid_num_indexes> missing in XML file, 'Index Count Changed' emulated by counting oid_index entries
+ Executing SNMP walk for list of indexes @ '.1.3.6.1.4.1.2021.13.15.1.1.1' Index Count: 0
+ No SNMP data returned
Do you have any idea?
thanks
Who is online
Users browsing this forum: No registered users and 5 guests