ON_ERROR_RETRY_EXPONENTIAL_BACKOFF overflows into 0s delay #618
Comments
|
Ouch, thanks for reporting this and appreciate the analysis. Your understanding of the bug seems right to me and makes it an easy fix -- fairly nasty bug. In the meantime I'd recommend using a fixed interval backoff e.g. 10 minutes. I'll get this patched in the next release. Done in #619 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I've configured a hook with CONDITION_SNAPSHOT_START that pings for repo availability and has Error behaviour: ON_ERROR_RETRY_EXPONENTIAL_BACKOFF
Basically this works fine: when the hook fails for the first couple of times, retries occur with a backoff delay exponentially rising up to 1 hour, as expected.
However, after a certain amount of retries, the backoff delay suddenly drops to 0 seconds, causing constant retries.
I have no GO experince, but suppose this is caused by an overflow here:
backrest/internal/hook/hook.go
Line 139 in a1e3a70
Duration is a 64bit type if I googled correctly, so the argument won't fit after the 65th attempt. Should limit the attempt variable to stop growing when 1 hour is reached.
Btw, thanks for this nice piece of software, makes using restic almost a no-brainer :)
The text was updated successfully, but these errors were encountered: