if synapse goes down, what happens?

[kfatehi]

does the client somehow deal with that for us, keeping messages in a queue until reconnect? or do we need to do this?

[AndrewJDR]

Btw, this actually happened to me yesterday. There was some kind of transient error happening with my home server (probably a cloudflare thing actually), and I got this when the hangouts bridge was attempting to send a 3rd party's message to a matrix room:

handleThirdPartyRoomMessage error { [M_UNKNOWN: Internal server error]
  errcode: 'M_UNKNOWN',
  name: 'M_UNKNOWN',
  message: 'Internal server error',
  data: { errcode: 'M_UNKNOWN', error: 'Internal server error' },
  httpStatus: 500 }

Ideally, the bridge would have put this in a queue for retrying. So there's definitely a need for something here...

[erdnaxeli]

We should stop using the client.sendMessage method (and client.sendImageMessage), and instead write our own method sendMessage(client, message) that takes care of putting the message on a queue if the HS is down.

Then we need a way to known when the HS goes up again to flush the queue. We must take care of flushing the queue before sending any message to not mess the messages order.

In the other way (M->T), we can redact too old messages and warn the user on the status room that its message was not sent.