Linode needs to reboot all their servers for a security upgrade which they tell us means up to 2 hours per server (we have one for the webserver and one for the database). We’re going to take the excuse to upgrade our servers before then. In theory we can set up Mongo’s database replication and do this with zero downtime. In practice we failed to get that to work last time we tried so we’re probably going to just take the hit on downtime while the database copies over to the new server. We think that means half an hour.
Whenever we have downtime, we live-tweet what’s going on via @beemstat.
Thanks everyone! Sorry for any hassle this entails (it would actually be really valuable to hear about such, so we can gauge for example how much effort it’s worth putting into db replica sets and such to get downtime near or literally to zero, or if it’s only likely to be a big hassle if it were multiple hours, etc).