DTrace doesn't require root, but some dtrace-specific privs, but root should give these to you.... But yeah, you can just use a pstack look as Riyaj already wroge (note that pstack doestn't see into OS kernel-space) and you can use my os_explain script with this one-liner for basic profiling :-) solaris01$ *for i in {1..100} ; do pstack 6774 | os_explain -a ; done | sort -r | uniq -c* 100 ???????? 100 main 100 opimai_real 100 sou2o 100 opidrv 100 opiodr 100 opiino 100 opitsk 100 ttcpip 100 opiodr 100 kpoal8 100 SELECT FETCH: 100 GROUP BY SORT: Fetch 100 COUNT: Fetch 100 NESTED LOOP JOIN: Fetch 100 TABLE ACCESS: Fetch 100 kdsttgr 97 kdstf0100101km 3 expeal 11 kdstsrp0km 23 kdst_fetch 28 expeal 10 COUNT: Rop 1 sskgslcas 1 slmxnoop 2 qeaeCn1 19 ktrget 1 ksu_dispatch_tac 1 kcbrls 23 expepr 2 evareo 5 yield 2 ktrgcm 1 kews_timeout 11 kcbgtcr 17 evareo 1 times 7 lmebco 1 ktrexf 1 kcbzgs 3 kcbs_simulate 2 yield 1 kssadf_stage 1 kcbsacc 1 kssadf_stage_numa On Thu, Jun 7, 2012 at 11:02 PM, Ls Cheng <exriscer@xxxxxxxxx> wrote: > Hi > Yes probably changing the way this process is run we can get the objective > but seems like there are more processes suffering same issue so now we are > looking into finding the root cause. > > I have applied a couple of patches which deals with performance degredation > and memory leak in 10.2.0.5 without success, tried some truss without > success neither > > dbms_profiler gives funny results, wonder if it's working > > Next tests will be running 10046 for 1000 and 2000 customers and probably > mix with Tanel dtrace script however I doubt I can use it because last time > I ran dtrace root privileges was requiered > > Thanks > > > -- > //www.freelists.org/webpage/oracle-l > > > -- //www.freelists.org/webpage/oracle-l