LWP leak in agent12c?
I use Hobbit (now called Xymon) to monitor our Oracle DB servers - Solaris 10 upd 8. One of the metrics Hobbit watches is CPU usage based on prstat output. Our primary production DB is on a V490 with 4 dual core CPUs (8 cores total). I typically receive alerts two to four times a day when the "CPU" usages exceeds 4 (half the number of cores). Recently I have seen alerts (with include the output of the prstat) showing a single process using up to 26%. Example:
PID USERNAME SIZE RSS STATE PRI NICE TIME CPU PROCESS/NLWP
24034 oracle 325M 254M sleep 59 0 114:23:55 26% java/234
7298 oracle 1242M 798M sleep 48 2 0:02:28 5.6% oracle/14
0