On Friday I went to Scale Summit, an unconference for ops and dev professionals working on scalable, high-performance systems.
There were many knowledgeable people so I've written up a couple of the sessions that I found most fruitful. There's also a helpful write-up of sessions I've missed out.
Once again Open Space Technology proved a good format. I haven't yet run an OST event, but Scale Summit has further cemented my wish to do so.
However, it's dependent on everyone wanting to contribute in some way to the proceedings. While I think this was the case for most, I wonder if having fewer people would have concentrated some of the discussions.
All in all though, it was a real success. Thanks to the organisers for pulling off an excellent day.
Points of interest:
The usual collection of monitoring software was discussed: Nagios, Cacti, Zabbix, Ganglia, Diamond, Riemann, Graphite, etc.
This seems to be a consequence of a diverse, fragmented landscape of monitoring needs. But there is a lot of overlap in needs. If we look at system monitoring (ignoring services and applications) the core requirement is simple: get metrics out of machines and into a datastore for analysis.
There's a scope issue here too. Some people were referring to metrics collection only, others seemed to be thinking about broader collection (& logging) and analysis.
There are a bunch of alternative dashboards for graphite.
It's often informative to do some indsutry introspection with smart people away from the usual Internet echo chambers. There were a bunch of interesting observations on what's changed since the last Scale Summit (late 2012).
I've cherry-picked a few points:
Attempting to mildly troll the room by claiming Python 2 would be dead by the next Scale Summit was (unsurprisingly) met with disagreement.
Suggestions for getting the best out of new team members:
None of these are too surprising, but it's always good to hear experiences of what has worked and what didn't.