We are seeing errors on the /metrics end-point suggesting that the same metric is attempted to being exported more than once during the same collect.
I suspect this is because the ttlcache.Range() function doesn't grab a mutex for the entire loop, and that the cleanup or other modication is causing the same values to be emitted more than once during the same iteration.
This change gets a copy of the map then iterates over that, and hopefully stops that error condition from manifesting itself.
(We'll keep an eye on for a while to see if this re-occurs, but wanted to open the PR early in case the previous change is breaking anyone else)
We are seeing errors on the /metrics end-point suggesting that the same metric is attempted to being exported more than once during the same collect.
I suspect this is because the ttlcache.Range() function doesn't grab a mutex for the entire loop, and that the cleanup or other modication is causing the same values to be emitted more than once during the same iteration.
This change gets a copy of the map then iterates over that, and hopefully stops that error condition from manifesting itself.
(We'll keep an eye on for a while to see if this re-occurs, but wanted to open the PR early in case the previous change is breaking anyone else)