Possible deadlock during fetching part (only when I use a non-production feature "zero-copy replication") #37423
Labels
experimental feature
Bug in the feature that should not be used in production
There is a possible deadlock in
Fetcher::fetchPart
when fail to do fetching with zero-copy.As the stack trace shows, there is a recursive call to
fetchPart
when fail to do fetch with zero-copy. Each call requires obtain an http session from a session pool which has a maximum size limit. If there are many threads callingfetchPart
simultaneously, the pool may be exhausted on the first call. And all threads will block on the second, eventually leading to a deadlock.The text was updated successfully, but these errors were encountered: