tombentley commented on a change in pull request #9676: URL: https://github.com/apache/kafka/pull/9676#discussion_r545144134
########## File path: core/src/main/scala/kafka/log/Log.scala ########## @@ -1556,6 +1572,8 @@ class Log(@volatile private var _dir: File, done = fetchDataInfo != null || segmentEntry == null } + checkForLogDirFailure() Review comment: Yeah, I wondered about putting it in `maybeHandleIOException` too, and I've made this change. There are a few places where changing the order in which we do `maybeHandleIOException` with respect to `lock synchronized` would seem to avoid some possible races with no side effects other than increasing the duration the lock is held (for the `try` and the volatile read). I'm thinking specifically `maybeIncrementLogStartOffset`, `roll`, `delete`, `truncateFullyAndStartAt` and perhaps even `truncateTo`. But given that there will always exist some races maybe it's not worth it, WDYT? ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org