Re: Problem with ORACLE Parallel Server

These figures are definitely NOT okay.
An average time to acquire a DFS lock handle of 18.22 seconds (the figures are given in hundredths) is indicative of a problem. (For example, when parsing a simple select at one node Oracle must take out a non-PCM lock against cached dictionary information at the other node - this step would result in a wait for DFS lock handle.)

However - an average of 18.22 seconds is very misleading, since it is probably the result of the occasional long waits that you are seeing. The problem is that you can't get in to pin down the exact wait when the wait starts to happen.

I suggest you use the orambx facility (if that is still around in this version of the VMS port) to send a '10046' trace event at level 8 (for wait) at a couple of sessions, then look at their trace files for the DFS Lock Handle waits after a big wait has occurred. The P1 value (in hex) will tell you the lock type, and mode that the process is waiting on.

--

Jonathan Lewis
Yet another Oracle-related web site:  http://www.jlcomp.demon.co.uk

Günter Gratzer wrote in message <8mjso6$2cl$1_at_rohrpostix.uta4you.at>...


>Hi Jonathan,

>thanks for the quick answer!

>It's not a 'dynamic resource allocated' and 'dynamic lock allocated'

>problem. I had this before, so I adjusted the lm_ress and lm_locks

>parameter. The limits are not reached if I look in v$resource_limit.

>Concerning DFS events I just found 'DFS lock handle' the values are:

>Total_waits 40221

>Total_timeouts 37116

>Time_waited 733034

>Average_wait 18,22

>I'm not sure if this values are ok or not. I will look to the documentation

>tomorrow if I find something about this event.

>

>See you

>Günter

>

>Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk> schrieb in im Newsbeitrag:

>965570595.27100.1.nnrp-08.9e984b29_at_news.demon.co.uk...

>>

>> Always look in v$system_event first.

>>

>> In this case, events like '%DFS%' may

>> give you a clue.  There may be some

>> problem with IDLM locks.  Also look in

>> the alert.log for lines like:

>>     dynamic resource allocated

>>     dynamic lock allocated

>>

>>

>>

>> --

>>

>> Jonathan Lewis

>> Yet another Oracle-related web site:  http://www.jlcomp.demon.co.uk

>>

>> Günter Gratzer wrote in message <8mjje3$quv$1_at_rohrpostix.uta4you.at>...

>> >Hi,

>> >I have installed ORACLE Parallelserver (8.0.5 with about 20 patches) on

 VMS


>> >(V7.2-1). There are two machines in the VMS-Cluster. If I start only one

>> >instance, ORACLE is running with more or less good performance. If I

 start


>> >the second instance on the other machine, the database hangs every few

>> >minutes for a while. Hangs mean I can't access the Database even with

>> >SQLPLUS. There is no problem on the VMS if I look with MONITOR SYSTEM.

 No


>> >ORACLE process is using CPU, no pagefaults no IOs. There is enough free

>> >memory left. After a while I can access the Database again.

>> >I didn't get any good and useful hints from ORACLE until now. They just

 told


>> >me it's a tuning problem, but they didn't told me how I can find out

 what's


>> >the problem and where I should tune.

>> >

>> >Maybe somebody had a similar problem and can give me some hints.

>> >

>> >Thanks!

>> >Günter

>> >

>> >

>>

>>

>

>