RE: VKTM ? Anyone?
Date: Tue, 16 Sep 2008 15:48:11 +0200
I totally agree / came to the same insight and thanks for the extra info. I hadn't noticed "VTKM" before and I agree that this could have been the unlucky victum.
While thinking about it today, my guess is that it somehow/maybe hooks into, supports "DBTime" et.all. It really sounds like an extention to / base for ASH/AWR/ADDM/etc. I once (Miracle Open World 2007) heard Graham (Wood) talk about the problems with the "wallclock" on "refined as UNIX" aka Linux systems. So I guess if I would use _disable_vktm uthen probably Graham will be unhappy
It good to know I have options to tweek, but for now I will go for solutions that general DBA practices will support and/or otherwise find a workaround that will grap less PGA.
OS, platform, version = Oracle Enterprise Linux 5 base release.
uname -a: Linux srv01-18-102 2.6.18-8.el5 #1 SMP Tue Jun 5 23:11:13 EDT 2007 i686 i686 i386 GNU/Linux
I will have a look into syslog. It could be interesting to see what it traced.
From: Tanel Poder [mailto:tanel.poder.003_at_mail.ee]
Sent: dinsdag 16 september 2008 12:59
To: Marco Gralike; oracle-l_at_freelists.org Subject: RE: VKTM ? Anyone?
If the xml loading process managed to use up all free memory in the system (and eventually failed with ORA-4030) then other processes can't do further allocations either. VKTM might have been the unlucky victim in this case.
I consider setting some sort of virtual memory limit with ulimit a good practice. I normally set it some 10x higher amount than I think the process ever needs (so that processes wouldn't fail due occasional spikes in memory usage), but rogue processes and memory leaks will not be able to eat up all memory.
On which OS, platform and version you are?
Linux kernel allows virtual memory overallocation ( if
/proc/sys/vm/overcommit_memory > 0 ), but in that case it starts killing
processes if it runs out of memory when processes start touching those allocated pages. If the memory overcommit is enabled, then again VKTM might have been unlucky victim again. Syslog should have some lines about it in such case i think...
Anyway, if the root cause for the issues wasn't memory overallocation but just VKTM crashing, there's a _disable_vktm parameter you can ask support about ;-)
-- Regards, Tanel Poder http://blog.tanelpoder.com <http://blog.tanelpoder.com/> http://n.otepad.com <http://n.otepad.com/> - n.ote this! ________________________________ From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Marco Gralike Sent: Tuesday, September 16, 2008 15:59 To: oracle-l_at_freelists.org Subject: VKTM ? Anyone? Apparently I missed something somewhere... Errors in fileReceived on Tue Sep 16 2008 - 08:48:11 CDT
ORA-56703: VKTM process died unexpectedly Mon Sep 15 22:16:47 2008 Error occured while spawning process q001; error = 56703 PMON (ospid: 16980): terminating the instance due to error 56703 Mon Sep 15 22:16:53 2008 Termination issued to instance processes. Waiting for the processes to exit Instance termination failed to kill one or more processes Mon Sep 15 22:17:15 2008 Instance terminated by PMON, pid = 16980 [oracle_at_srv trace]$ oerr ora 56703 56703, 00000, "VKTM process died unexpectedly" // *Cause: An explicit kill or internal error caused the death of VKTM // background process. // *Action: Restart the instance. Apparently that this is serious enough, my instance crashed and to my surprise without something like a ORA-0600 (maybe because "it is handled..."), although I can imagine that this process, apparently with dependencies with PMON, will cause the other important backgound processes to shut down (if not already the case). For those who are interested this happened after trying to insert a lot off big xml files into a binary xml xmltype, table which caused a lot of ORA-04030, that caused job queue processes to die, that caused server (shared server) processes to die, that signaled pmon to terminate the instance. So my question... Does anyone know what a "VKTM" background process is...? -- http://www.freelists.org/webpage/oracle-l