Who has the largest Zabbix installation?

[SAT QPass]

And to get the ball rolling, we are proud to have 18.2K items (note this is just the ACTIVE) in our database and we are about to add several thousand more. Before we trimmed some of our bloat and redundant data we were running with 23K+ ACTIVE items. I fully expect to top that number easily in the coming weeks as we add more data collection as we have hundreds of node points per server in some cases.

Hardware is currently a Quad Xeon for mysql attached to 300GB slice on a NetApp filier with a dedicated server (running the actual zabbix_server process).

I would be interested to know what size other installations are, ideas, anecdotes, suggestions for optimization, etc.

[cameronsto]

Our install is very small and currently only doing monitoring of trends no triggers or alerts. We need to upgrade to 1.1 (from 1.1beta5) and then I'm going to start using the rest of the features provided by Zabbix.

What kind of load does your box have measuring that many item?

-cameron

[SAT QPass]

What kind of load does your box have measuring that many item?

Good question. Believe it or not.... almost zero. <.20 The zabbix_server itself is sitting on a dual xeon and total %user CPU is <2 as an average. To give you and idea of context we are running 32 suckers and 198 trappers. I would say 95% of the monitored items are zabbix_agent (active) vs. zabbix_agent.

Load on the individual monitored nodes is unmeasurable as well. Something we are actually very sensitive to because of our app requirements.

The mysql server load is also almost zero once you upgrade to 1.1 final. Prior to that running beta8 load was near 20. So there are HUGE performance gains to be had by upgrading.

[cameronsto]

The mysql server load is also almost zero once you upgrade to 1.1 final. Prior to that running beta8 load was near 20. So there are HUGE performance gains to be had by upgrading.

Wow. 1.1 final helped that much? Definitely need to get around to upgrading then.

-cameron

[SAT QPass]

It helps that much when you have 23K items in your database to monitor

[ruckus37]

You guys win hand down, cant wait for our next fusion, we will have around 500 servers to monitor, don’t even want too guess how many triggers or alerts we will have running, probably have an it department +- 60. We will most likely have a few zabbix servers at different locations in a kinda clustered setup.

[SAT QPass]

We use a remote install to act as a heart beat monitor of the primary installation. It also validates our alert path for notifications and triggers. Nothing like having your monitoring app running and alerting just fine, but no route to alert

[SAT QPass]

Anybody who does not think Zabbix is viable for the enterprise should talk to us. Our core business rests squarely on the shoulders of Zabbix and its ability to let us know when our production infrastructure is hosed.

[ruckus37]

Nothing like having your monitoring app running and alerting just fine, but no route to alert

come again?

[SAT QPass]

If you rely on external routes for alerts (such as public IPs) having your monitor up and running does nothing for you if it cannot actually send the alert because your outbound connectivity is down. That is why we have an installation that sits OUTSIDE of the environment on a remote site to validate the alert path is functional.

[ruckus37]

Roger that...

[cybaspace]

We use zabbix 1.1.1 to monitor all kinds of tings like routers, servers workstations, UPSes and printers. We also monitor our batch processing in zabbix. Here is the current status:

Number of hosts 164
Number of items 1679
Number of triggers 1404
Values stored 45877430

For servers there is one HP Proliant DL360 raid5 for the Mysql server. It's a Xeon 3 Ghz with 2 Gb mem. For the webserver / sucker part there is a DL320 SATA RAID. These servers are internally connected with a crosslink 1 Gb link.

The amount of triggers / hosts / items is growing every day!