Closed aine-etke closed 1 year ago
Related #493
I've posted some related ideas to improve error behaviour instead of disabling it outright to chat, quoted here for convenience:
alternatively to showing the rss error on first timeout(?) and increasing the timeout from 10s to 30s, it could only show it on consecutive poll timeouts? so if it times out once then it fails silently, but if $interval minutes later it times out again, then it sends an error to matrix. the argument being that feeds should not be so time sensitive since you probably don't expect them to be absolutely realtime
Maybe the bridge could consolidate multiple RSS timeout errors into one message instead of repeating it for potentially every feed a room is subscribed to?
I'd prefer to explicitly disable it. We use rss feeds for project news and we don't need any error reporting in that specific room
It would be neat to have more error resilience on one hand, but also error reporting only to admin rooms for example on the other.
Yeah, suspect you are both right. We should reduce the amount of noise from hookshot in the rooms, but also try to fix the timeouts.
The error history is now in the RSS widget, so perhaps it's not that important to announce it explicitly as messages. At the same time, it is useful to know when a feed breaks completely to check if the URL has changed, or something.
Timeouts themselves will be almost always intermittent though, so perhaps we can just silence them – or require them to fail for a longer period of time to matter?
I'd like to have timeout error disabled by default. In current form it just spams room if there is a random glitch. If the feed URL can't be reached, it could be displayed in the widget.
If the feed URL can't be reached, it could be displayed in the widget.
i get that y'all don't want to get spammed. personally I bumped the timeout from 10s to 30s and it's fine apart from very few exceptions.
if you disable errors in chat, how will you notice that a feed broke? do you check the widget every now and then manually? or do you just not care and accept that feeds may silently disappear/break?
I would suggest to change the threshold for the errors. A single timeout doesn't mean anything. If a feed couldn't be fetched for a week, this is a legitimate concern. But being unsuccessful once? Nope, please don't, a matrix room is not a (de-)centralised log system. This only leads to alert fatigue.
Hiya, sorry about the noise. I had thought this did only alert on several failures although I think that logic may not be as smooth as we hoped. I'll put in something to ensure it doesn't make so much noise. Please hang tight.
This appears to have been resolved with a very black and white decision users now have to make Is there another more nuanced option planned?
Please, give ability to disable errors like
in the matrix rooms