|
truss -fdD is good. You can check the truss output first. Identify the big timestamp jumps. If there's none, then it's hard to say where the bottleneck is. It could simply be that the tnslsnr process is stressed in general. If the process virtual memory is indeed > 2G (see 老熊's message #24), that's almost certainly a problem. My listeners on various servers have at most 140 M. This one is 11gR2:
$ ps -ef | grep [t]ns
oracle 26945 1 0 Aug08 ? 00:00:19 /u01/app/grid/bin/tnslsnr LISTENER -inherit
$ ps -o vsz 26945
VSZ
142736
Keep checking yours to see if the size always goes up. If it does, search for "listener memory leak" on MOS.
Yong Huang |
|