My server was down[0] for a few hours once and so was my XMPP account. After it was back up, I got a small VPS from a friend to act as a backup/fallback/redundancy-whatever. ejabberd offers that out-of-the-box, meaning that I now have two ‘nodes’ hosting my XMPP domains – if one of them goes down, I either don’t notice at all (if I was connected to the other one) or just have to tell Pidgin to reconnect. s2s connections will similarly just reconnect/don’t notice.
I liked Prosody better, though, and the arcane failure modes[1] of ejabberd still annoy me. So if you get that into Prosody (or did, I didn’t check in the last year or so), I will definitely consider switching back.
[0] Apparently Strato thinks you’re being attacked if a large stream of UDP packets from the 26C3 network hits your server. And I was just using my VPN…
[1] I recently moved that VPS and failed to adapt iptables rules on the other host. Instead of telling me that it couldn’t connect (at maximum log verbosity), ejabberd just did nothing/crashed. I can’t exclude an oversight on my part, but the failure to log anything was rather annoying.
Ah, clustering. Understood, and entirely justified.
Prosody clustering is on the roadmap, but it's a tricky problem to solve properly (and I don't think anyone has to date). We've even had people switch from ejabberd clusters to Prosody, as ejabberd's clustering wasn't working out for them anyway (it seems to be designed more for spreading load than increasing availability).
Our view is that we wanted to get the foundations of the server right before we began tackling such complex problems. Nevertheless I look forward to welcoming you back some time soon :)
If ejabberd starts up, I found it to work rather nicely, so it is not a big problem to me.
But I wish you the best of luck with Prosody, as I said, I really liked it at the time and I trust it only got better :) – regarding the ‘soon’, 2015 is likely an optimistic estimate given Debian release schedules etc…
My server was down[0] for a few hours once and so was my XMPP account. After it was back up, I got a small VPS from a friend to act as a backup/fallback/redundancy-whatever. ejabberd offers that out-of-the-box, meaning that I now have two ‘nodes’ hosting my XMPP domains – if one of them goes down, I either don’t notice at all (if I was connected to the other one) or just have to tell Pidgin to reconnect. s2s connections will similarly just reconnect/don’t notice.
I liked Prosody better, though, and the arcane failure modes[1] of ejabberd still annoy me. So if you get that into Prosody (or did, I didn’t check in the last year or so), I will definitely consider switching back.
[0] Apparently Strato thinks you’re being attacked if a large stream of UDP packets from the 26C3 network hits your server. And I was just using my VPN…
[1] I recently moved that VPS and failed to adapt iptables rules on the other host. Instead of telling me that it couldn’t connect (at maximum log verbosity), ejabberd just did nothing/crashed. I can’t exclude an oversight on my part, but the failure to log anything was rather annoying.