I''ve run across a slightly unusual - at least to me - situation where
my load average bumps up to 200-300, as shown by prstat or uptime, but %cpu
utilized on the system ranges from 30 to 50%. When I check the run occupancy,
via dtrace, I see a very wide range of queued processes on the various
cpu''s. Anyone got an idea on how to check on two things:
1) how to determine why some cpu''s have 0 to 1 threads wait on that
cpu''s queue and other cpu''s have 30 to 70 processes on their
queue.
2) how to determine what the threads are locked against that prevents them from
running.
These threads are all from a Sybase IQ database which is one process and 2000 to
4000 lwp underneath.
Thanks
This message posted from opensolaris.org
