Monitoring update - it's working for us!

I had a post about monitoring a while back, here's a progress update...

Some ideas behind what we're trying to achieve:
  1. Gain insight into how our systems are functioning (from an application point of view, secondary from a server point of view)
  2. Know that something is broken before it breaks (minimum immediately)
  3. Show people how our systems are functioning
  4. Use data (metrics/logs/etc) to aid in trending, RCAs, etc
Again, our big catch is that we run 95% Windows in our 'production' environments.  An MS shop, if you will.  Another big catch is very small budget for this sort of thing.  Another big catch is very little support from developers (who are busy working on revenue-generating projects).  Another big catch is the existing monitoring infrastructure consisted of WhatsupGold and hard-coded email alerts inside the system (i.e. we have to dump it).

So...the internet provided ideas and tools, we just put them together.
  • Data sources: Windows event log & IIS logs (NXlog CE), Windows perf counters (graphite-powershell), Nagios data (Check_MK agent), Elasticsearch data, firewall/LB logs c/o syslog
  • Monitoring & notification: Nagios (actually Check_MK c/o OMD)
  • Data orchestration (haha I made that up): Logstash (ELK)
  • Visualization: Graphios/Graphite/Grafana, Thruk Panorama, simple-nagios-dashboard
Amusingly, our monitoring infrastructure contains only a single Windows box (one machine that runs custom SQL queries that output Nagios-readable check/perfdata).

So here's what it looks like....

We're also now dumping Chef into the mix (haha!) to apply all the agent-stuff automagically.

Some links:
  • OMD: http://omdistro.org/start
  • OMD:Thruk Panorama: http://www.thruk.org/dashboard.html
  • OMD:Check_MK: http://mathias-kettner.com/check_mk_introduction.html
  • NXlog: http://nxlog.org/
  • ELK stack: http://www.elasticsearch.org/webinars/elk-stack-devops-environment/
  • Graphite: https://graphite.readthedocs.org/en/latest/overview.html
  • Graphite-powershell: https://github.com/MattHodge/Graphite-PowerShell-Functions
  • Graphios: https://github.com/shawn-sterling/graphios
  • Grafana 1.9.0: https://github.com/grafana/grafana
  • Simple-nagios-dashboard: https://github.com/JobV/simple-nagios-dashboard
All that put together and you get this (it's our first prototype):

Comments

  1. Hi,

    Nice schema there. Got me started on ELK :)
    Could I ask what dashboards are you using in Kibana?

    Thanks.

    ReplyDelete
  2. Kibana is vanilla at the moment - we're trying to get our support folk using it as a tool first (traditionally they do not use IIS/Event Logs). To be honest, wasn't even aware of Kibana-specific (i.e. ELK-specific) dashboards. Good to know!

    Our ELK stack is now up to 5 nodes of ES and one LS/Kibana box. I really want to get load-balancing in-house, but since all of this is brand new, there is no open-source ($0 budget) load balancer tech already in place.

    ReplyDelete
  3. Hm, I re-thought your comment - we're generally having the support folk look at environment-specific /IIS or Event logs, and those are grouped by info/warn/crit level (or 200/300,400,500 codes). So one dashboard would show trending and a single graph that details IIS codes for all production web servers.

    Logstash sends all IIS traffic to passive Nagios monitors that are host-specific. So the host goes 'ok' with 200/300 traffic, 'warn' for 400s, and 'crit' for 500s. It means it's a 'self-healing' monitor in a sense - it notifies if we get 500s. This also is a new thing for us...maybe that'll change.

    Hope that helps. Also, we're checking this out: http://www.networkassassin.com/elk-for-network-operations/

    ReplyDelete

Post a Comment

Popular posts from this blog

DFSR - eventid 4312 - replication just won't work

Fixing duplicate SPNs (service principal name)

Logstash to Nagios - alerting based on Windows Event ID