I don’t recall exactly, but it’s more like days rather than hours. At some point the instances will mark you as down, and then stop trying to federate with you, so there’s a hard limit but it’s fairly generous and not especially aggressive.
Lemmy does exponential backoff with something like this: retry_in = min(1.25^retry_count, 24*3600)
Basically, the timer for the next retry will grow exponentially until it reaches 1 day, then it continues to retry every day.
By that time, about 5 days will have gone by going through the previous retry failures.
If will eventually give up and mark as dead but I don’t remember the threshold for that.
Yeah. There’s a retry queue, which does expire after a certain time period, but for a short outage that’s how it’d work.
Do you know about how long it is?
I don’t recall exactly, but it’s more like days rather than hours. At some point the instances will mark you as down, and then stop trying to federate with you, so there’s a hard limit but it’s fairly generous and not especially aggressive.
I found the PR for the queue, and it mentions retries but doesn’t seem to mention exact timing, at least to my quick read. ( https://github.com/LemmyNet/lemmy/pull/3605 )
Thanks for further details and that PR! Days sounds great. I usually try to fix it within 12 hours or so
Lemmy does exponential backoff with something like this:
retry_in = min(1.25^retry_count, 24*3600)Basically, the timer for the next retry will grow exponentially until it reaches 1 day, then it continues to retry every day.
By that time, about 5 days will have gone by going through the previous retry failures.
If will eventually give up and mark as dead but I don’t remember the threshold for that.
Thanks for the function! Great to know :)