What steps will reproduce the problem?
Use the fire hose adapter with multiple ingestion nodes, after some time the
system will hang. By looking at the stack traces, you will find that the lock
and log managers are in a deadlock state. I would say the lock manager should
not be talking with the log manager at all. This is probably one of the
consequences of issue 621, and probably can be go away once that issue fixed.
I also noticed that sometimes in other NCs stack traces, some of the ingestion
threads are waiting because the memory component was not flushed. After initial
investigation, I found out that the active transaction count (that is being
decremented by the log manager) was off by one. This also needs to be
investigated.
Original issue reported on code.google.com by salsuba...@gmail.com on 25 Aug 2013 at 7:30
Original issue reported on code.google.com by
salsuba...@gmail.com
on 25 Aug 2013 at 7:30