Discovery Disks & IO/Performance

[dvuong]

I found the following articles:

http://rals-madlab.blogspot.com/2013/07/in-spirit-of-zabbix-and-good-monitoring.html

http://www.muck.net/19/getting-hard-disk-performance-stats-from-zabbix

I'm having problem importing the template.

Currently running: Zabbix 2.0.6
Trying to get disk discovery (like filesystem size but for I/O)
And automatically generate I/O graph for disk ...

Anyone have such a template to share?

[cidnurg]

Here you go

Encountered the same problem today and decided to share solution - https://github.com/grundic/zabbix-disk-performance

[dkanbier]

I've created a template with LLD support for Zabbix 2.2.1:

http://www.denniskanbier.nl/blog/monitoring/monitoring-disk-io-using-zabbix

I don't have access to a hardware raid device at this very moment, when I do I'll try to add that as well.

[joshuamcdo]

Reisk i/o

You mean to tell me that zabbix still doesn't support this natively?

[kloczek]

You mean to tell me that zabbix still doesn't support this natively?

Creating list of block devices can be done by system.run[] like:

system.run["awk 'BEGIN {print \"{\"; print \"\\"data\\":[\"; ORS=\"\"} {if (NR!=1) {print \",\n\"}; print \" { \\"{#DISK}\\":\\"\" $3 \"\\" }\"} END {print \"\n ]\n}\n\"}' /proc/diskstats"]

complicating LLD by use external UserParametr and script usually adds unnecessary level of complexity.
Filtering off some part of LLD json string can be done natively by using zabbix regexp. With LLD key using system.run[] and regexp done of scripts is needed.

If you have few hosts to monitor it may be not a problem to use some script executed by agent. However on scale hundredths or thousands hosts it is real nightmare.

[jan.garaj]

You mean to tell me that zabbix still doesn't support this natively?

Yes, it supports natively, keys: vfs.dev.read, vfs.dev.write *
You don't have access to all block devices metrics (/proc/diskstats), because IMHO it's problem with different OS and their IOps metric implementation. Some kernel versions (e.g. 2.6.18-53.el5.centos.plus, 2.4-) can be also problem, because they don't provide all "standard" block device metrics.

For me is important question which monitoring system can provide extended IOps monitoring (avgqu-sz, %util, ...) natively?
Does someone know answer?