Hi, Our fetchers are parsing and we sometimes see the parsers stall and eventually hang. This happens consistently during some mappers. When they are restarted by Hadoop they're guaranteed to stall at roughly the same point, the parsers start to timeout first and then the fetcher refuses to print its usual progress.
I did some stack traces and all threads are blocked and most involve the parser. I can't seem to understand the issue here so any pointers are much appreciated. We have a useless cluster right now :) Thanks!