Comments on: Network / System Monitoring Smorgasbord http://kylecordes.com/2008/network-system-monitoring-smorgasbord Software, Business, and Life Tue, 20 Dec 2011 16:22:18 +0000 hourly 1 http://wordpress.org/?v=3.2.1 By: po http://kylecordes.com/2008/network-system-monitoring-smorgasbord/comment-page-1#comment-28124 po Mon, 01 Dec 2008 22:14:35 +0000 http://kylecordes.com/2008/10/19/network-system-monitoring-smorgasbord/#comment-28124 I've eval'd all of the aforementioned tools and find some are more suited to their intended roles: trending and others monitoring. Sure, the lines have been blurred (nagios plugins for ganglia, etc), but personally I think they better left as two distinct roles (until the next one comes along...). Nagios has addons to chart performance data, but it's first and foremost a monitoring & alerting system. It can scale to large numbers of clients by the use of nrpe or even nsca to push results (and even distributed nagios instances, etc). I prefer to push the result back to avoid holding a client up , rather than polling via snmp, etc. Alerting is flexible. I wish it was simpler to setup and use, but saying it's the least worst of a bad bunch does it a disservice. Ganglia, which I also use is ideal for trending. Basically you can chart anything provided you can supply a value to the gmond program (pick your language). It's great for seeing whether MySQL setting change really did the trick than the one two weeks ago and whether disk i/o dropped. This is very useful for capturing business/app metrics in addition to the typical lamp stack metrics. If deployed with multicast, then you have some resilience for your data/gui (provided you have more than one host running gmetad - i.e more than one host storing the rrds). I find it useful for capacity planning. It's also pretty easy if you're not using snmp to get custom data into Nagios / Ganglia - I'm sure the other have similar provisions, but it was trickier imo. Others: I've used hyperic (ent demo), zenoss, zabbix, cacti, but I found them all to be restrictive. I settled on a Nagios / Ganglia combination Hyperic - (I tested 3.0, but is now v4 I believe) was feature limited in the demo, but I found the agents bulky. Zenoss - install was awkward, but the interface was tricky to use. Cacti - only found it useful for small installs. Painful if you need to do repetitive tasks (templates in cacti can be tricky). Used it with cactid, otherwise polling could take a while. Zabbix - disliked for similar reasons to cacti, largely related to admin/ config tasks. I also tried OpenNMS - didn't eval in as much depth (suffered from memory issues even after tuning the server) Another one which may be of interest: http://www.collectd.org - it also support multicast, and is purely intended to be a stats collector (in so far that you choose your own tool to display the resulting rrds), but very small footprint. (A collectd > ganglia plugin would be nice!) I’ve eval’d all of the aforementioned tools and find some are more suited to their intended roles: trending and others monitoring.
Sure, the lines have been blurred (nagios plugins for ganglia, etc), but personally I think they better left as two distinct roles (until the next one comes along…).

Nagios has addons to chart performance data, but it’s first and foremost a monitoring & alerting system. It can scale to large numbers of clients by the use of nrpe or even nsca to push results (and even distributed nagios instances, etc). I prefer to push the result back to avoid holding a client up , rather than polling via snmp, etc. Alerting is flexible. I wish it was simpler to setup and use, but saying it’s the least worst of a bad bunch does it a disservice.

Ganglia, which I also use is ideal for trending. Basically you can chart anything provided you can supply a value to the gmond program (pick your language). It’s great for seeing whether MySQL setting change really did the trick than the one two weeks ago and whether disk i/o dropped. This is very useful for capturing business/app metrics in addition to the typical lamp stack metrics. If deployed with multicast, then you have some resilience for your data/gui (provided you have more than one host running gmetad – i.e more than one host storing the rrds). I find it useful for capacity planning.

It’s also pretty easy if you’re not using snmp to get custom data into Nagios / Ganglia – I’m sure the other have similar provisions, but it was trickier imo.

Others:
I’ve used hyperic (ent demo), zenoss, zabbix, cacti, but I found them all to be restrictive. I settled on a Nagios / Ganglia combination
Hyperic – (I tested 3.0, but is now v4 I believe) was feature limited in the demo, but I found the agents bulky.
Zenoss – install was awkward, but the interface was tricky to use.
Cacti – only found it useful for small installs. Painful if you need to do repetitive tasks (templates in cacti can be tricky). Used it with cactid, otherwise polling could take a while.
Zabbix – disliked for similar reasons to cacti, largely related to admin/ config tasks.
I also tried OpenNMS – didn’t eval in as much depth (suffered from memory issues even after tuning the server)

Another one which may be of interest: http://www.collectd.org – it also support multicast, and is purely intended to be a stats collector (in so far that you choose your own tool to display the resulting rrds), but very small footprint. (A collectd > ganglia plugin would be nice!)

]]> By: Matt Ray http://kylecordes.com/2008/network-system-monitoring-smorgasbord/comment-page-1#comment-28115 Matt Ray Mon, 01 Dec 2008 18:05:25 +0000 http://kylecordes.com/2008/10/19/network-system-monitoring-smorgasbord/#comment-28115 Zenoss is remote agentless monitoring, and we're making a push into more OS monitoring via SSH in our next release: http://forums.zenoss.com/viewtopic.php?t=8137 There is a community ZenPack for PostgreSQL, (http://www.zenoss.com/community/projects/zenpacks/postgresql) and I'm working on opening up Community ZenPack efforts to more collaboration very soon. I'd be more than happy to get a more robust PostgreSQL ZenPack worked on. Karl: Sorry to hear you had issues with Zenoss' installation, was this recent? We've done a lot of installer work and documentation around this recently. Feel free to email me for followup. Thanks, Matt Ray Zenoss Community Manager mray@zenoss.com Zenoss is remote agentless monitoring, and we’re making a push into more OS monitoring via SSH in our next release: http://forums.zenoss.com/viewtopic.php?t=8137

There is a community ZenPack for PostgreSQL, (http://www.zenoss.com/community/projects/zenpacks/postgresql) and I’m working on opening up Community ZenPack efforts to more collaboration very soon. I’d be more than happy to get a more robust PostgreSQL ZenPack worked on.

Karl: Sorry to hear you had issues with Zenoss’ installation, was this recent? We’ve done a lot of installer work and documentation around this recently. Feel free to email me for followup.

Thanks,
Matt Ray
Zenoss Community Manager
mray@zenoss.com

]]> By: Jonas B. http://kylecordes.com/2008/network-system-monitoring-smorgasbord/comment-page-1#comment-28101 Jonas B. Mon, 01 Dec 2008 07:56:14 +0000 http://kylecordes.com/2008/10/19/network-system-monitoring-smorgasbord/#comment-28101 Nagios graphs the data you want with the nagiosgraph plugin, You use nrpe to gather information form Nagios plugins remotely. It's really very easy to set up, but you need to familiarize yourself with its architecture before you dive in. A host has services and contacts, a service has contacts, etc. When it comes to monitoring you really want mature products. Give it a chance, it's well worth it! Nagios graphs the data you want with the nagiosgraph plugin, You use nrpe to gather information form Nagios plugins remotely. It’s really very easy to set up, but you need to familiarize yourself with its architecture before you dive in. A host has services and contacts, a service has contacts, etc. When it comes to monitoring you really want mature products. Give it a chance, it’s well worth it!

]]> By: Karl Katzke http://kylecordes.com/2008/network-system-monitoring-smorgasbord/comment-page-1#comment-26646 Karl Katzke Sun, 19 Oct 2008 18:49:07 +0000 http://kylecordes.com/2008/10/19/network-system-monitoring-smorgasbord/#comment-26646 I had an impossible time getting Zenoss up and running properly. It's configuration is halfway to impossible -- and I'm coming from the Nagios/Cacti world here. I definitely want to see further reviews of ganglia and zabbix, though. I had an impossible time getting Zenoss up and running properly. It’s configuration is halfway to impossible — and I’m coming from the Nagios/Cacti world here.

I definitely want to see further reviews of ganglia and zabbix, though.

]]>