Long shutdown times

[driskell]

Long shutdown times

[driskell]

Hi,

It's due to the internal queue waiting to flush. Anything that's already inside publisher needs to be sent first, thus the slight delay. If logstash is busy it will take longer. If the connection is severed it will not shutdown until it reconnects. These are known issues at the moment but I deemed them a low risk for most cases, and even if it gets killed thats just old behaviour - in most cases though it will shutdown cleanly.

I'll be improving it soon. So that if the publisher has a contiguous queue it just discards it. If it has a partially ack'd queue (non-contiguous) it will save it to a persistence file, reload on startup and resume send. So that shutdown is always as quick as possible. The only wait will then be for currently outstanding ACK - but the moment they timeout (approx 15s default) it will shutdown.

If I don't get around to the above improvements soon I'll document it on the README and CHANGELOG - probably as I should've done!

Jason

[driskell]

Not seen this too much and I think it's not a simple thing to resolve. Any resolution would involve "trashing" state as it were, so in a way I wonder if letting it be killed is fine - at least then it's a signal of a problem somewhere rather than looking like a normal exit. Sorry for long response on this. Just got around to some time to tinker and update things.