Hi Martin. Your plan sounds good, but why not to use special tools such monitoring tasks? Like monit[1], which could not only tell you that CouchDB instance is not available by HTTP API, but also does his process even alive, notify about any problems and restart service if so.
[1] http://mmonit.com/monit/ -- ,,,^..^,,, On Sun, Jul 8, 2012 at 5:51 PM, Martin Hewitt <[email protected]> wrote: > Hi all, > > I'm working up a monitoring system to track our database servers' statuses > and I was just wondering if anyone had any hints or tips of how best to > detect a server going down or becoming unreachable? > > My current plan is to subscribe to the _changes feed for a system database, > say _users, with a heartbeat of a second or so, and mark the server "at risk" > if the connection drops, or the heartbeat doesn't arrive, and mark it as > "down" if the _changes connection can't be re-established or if a second > heartbeat in a row is missed. > > Martin
