Count of? Triggers

[syndeysider]

Hi Guys

So i'm monitoring IO performance on Linux/Windows Servers. Our heavily over utilized SAN takes a knock every now and than and the Trigger Alerts queue fills up with "io exceeds..." etc. etc.

Good stuff.

I'd like to create an Action based alert to fire when the number of "active and unacknowledged" "io exceeds.." > 20 (e.g.) which interprets into 20 servers alerting on io performance issues.

It is normal on a given day to have 5-10 servers alert at various times on IO performance due to our underachieving SAN. What's not normal is 20+ servers alerting within a 15 minute windows...

Any thoughts?

[steveboyson]

Sorry, I do not really understand the question.

What about "io.load.avg(10min) > $LOADFACTOR" when $LOADFACTOR is a value which can be reached only in total overload of your SAN?

[syndeysider]

So instead of monitoring a metric on a particular host over a period of 10 minutes. I want to monitor a collection of metrics i.e. multiple hosts individual metrics over a 10 minute period.

"io.load.avg(10min) > $LOADFACTOR" works well per host for example, but i'm more concered if this gets breached for 20+ hosts

So 20+ triggers of function io.load.avg(10min) PROBLEM is the actual trigger/action?

[steveboyson]

You might define a specific hostgroup for all of your to-be-monitored hosts and then fiddle around with zabbix' grp* functions.

https://www.zabbix.com/documentation/2.0/manual/config/items/itemtypes/aggregate

[syndeysider]

Excellent! Exactly what I was looking for!

Thank you!