Closed arhimondr closed 5 years ago
In case all hosts are down, retrying the same hosts over and over again creates an infinite loop. That creates a lot of CPU pressure, as there is no delay between retrying in case of a ConnectionFailedException.
ConnectionFailedException
CC: @electrum @dain
If we kept a count of the failed connect attempts per address, we could add a simple backoff per address.
Resolved by https://github.com/airlift/drift/pull/101
In case all hosts are down, retrying the same hosts over and over again creates an infinite loop. That creates a lot of CPU pressure, as there is no delay between retrying in case of a
ConnectionFailedException
.