This content has been marked as final. Show 15 replies
Last couple lines from the omsca.log
Oct 28, 2011 3:22:57 PM oracle.sysman.omsca.adapter.wls.OMSWLSAdapter adapterCleanup
FINE: Cleaning the OMS from the repository
Oct 28, 2011 3:22:58 PM oracle.sysman.omsca.adapter.wls.OMSWLSAdapter adapterCleanup
FINE: Got the Host URL as: null
Oct 28, 2011 3:22:58 PM oracle.sysman.omsca.framework.OMSGenericAdapter postDeployAndReposSetup
SEVERE: OMSCA-ERR:Post "Deploy and Repos Setup" operations failed. Check the trace file:/u01/oracle/product/middleware/oms/cfgtoollogs/omsca/omsca_20111028150320.log
Oct 28, 2011 3:22:58 PM oracle.sysman.omsca.framework.OMSCAFreshInstall execute
SEVERE: Post "Deploy and Repos Setup" operations failed.
Oct 28, 2011 3:22:58 PM oracle.sysman.omsca.framework.OMSConfigAssistantDriver main
FINE: Got resultfalse
from the EMGC_OMS1.out file:
Error occurred during initialization of VM
Could not reserve enough space for object heap
Could not create the Java virtual machine.
<Oct 28, 2011 3:22:12 PM> <FINEST> <NodeManager> <Waiting for the process to die: 7385>
<Oct 28, 2011 3:22:12 PM> <INFO> <NodeManager> <Server failed during startup so will not be restarted>
<Oct 28, 2011 3:22:12 PM> <FINEST> <NodeManager> <runMonitor returned, setting finished=true and notifying waiters>
Yeah well, there are plenty people which still try it. Entire step-by-step guides can be found on the net. Since it's not a production system (otherwise i'd have more recources available), i'm not too worried about certification and was hoping for a bit more useful hints.
64bit works, even on my 4GB of RAM. Looking at top and ps, i'd say the single java process which gets started runs out of memory as it grows pretty much to the 2GB limit.
Interesting - I'm running into the same issue on 32 bit OL 5.7 UEK.
The WebLogic EMGC_OMS1 managed server is being started with these memory parameters:
-Xms256m -Xmx1524m -XX:MaxPermSize=612M
If you look at the admin server (EMGC_ADMINSERVER) you'll see it is started with
-Xms256m -Xmx512m -XX:MaxPermSize=512m
So between those two there's 512-2GB used for just the OMS JVMs.
I suppose the question is how much memory can a process allocate under 32 bit OL? I think this is something like 1.7GB without hugemem/pages - that looks like it should be OK given the max heap size above. The 12c installer pre-reqs are 3900MB memory and probably similar for swap.
In my case I also have a 11.2 database running on the same VM - its SGA was about 0.5GB.
Think I'm going to bump up the vMem and swap and try again (though based on your experiences I'm not holding my breath!).
Actually of course I should have been adding the PermGen space on top of the java heap... the JVM process is then looking like it needs over ~2.1GB.
Anyway, I just tried starting WebLogic manually (using the command it was using - just copy from the EMGC_OMS1.out log file if you want to try) and, sure enough, the managed server won't start up. The only way you can get it to start is to drop the max heap size or max PermGen size by about 350MB!
(OMSCA deletes all the domain etc config from under gc_inst after the failure has occurred so you can't just run up the servers with node manager).
Assuming the mx and MaxPermSize parameters are fixed in the installer somewhere (rather than being set dynamically according to memory) I can't see how this can being working for anyone on Linux 32 bit!
Surely Oracle wouldn't release software on a platform that hasn't been though any regression tests and doesn't even install..?!
I've come up with a workaround to allow you to install OEM CC 12.1 on 32 bit Linux (in this example, I'm using OL 5.7 UEK).
1) Run the OEM CC installer but choose to install software only
2) Edit the file <ORACLE_HOME>/oms/sysman/omsca/scripts/wls/create_domain.py which sets the Managed Server JVM startup parameters:
- change -Xmx1524m to -Xmx1324m
- change -XX:MaxPermSize=612M to -XX:MaxPermSize=412M
If you look at the comments in the file you'll see that various values have been tweaked between 11g and 12c releases (and actually the 612 is a typo for 512MB anyway).
3) Run <ORACLE_HOME>/oms/sysman/install/ConfigureGC.sh and enter details as you would normally.
You wouldn't do this on a production system of course (raise an SR instead, but then again production is almost certainly going to be 64 bit anyway) but it seems to work OK on a test laptop.
Hope this helps,
Thanks for the update Gokhan.
Well, assuming we're both using the JDK 6 (1.6.0_24) provided in the OEM 12.1 download, then I guess it must be a difference between your Oracle Enterprise Linux R5 U6 (you were using the default Red Hat kermel presumably?) and my Oracle Linux R5 U7 UEK (default now).
It looks as though the memory parameters have been tweaked on 32 bit to make them as big as possible and this might change slightly between JVM and kernel revisions.
Yes, I used the oracle validated RPM (oracle-validated-1.1.0-14.el5) too for convenience when installing an 220.127.116.11 EE database on the same VM for the repository. It'll be interesting to hear if anyone else has have such issues with 32 bit (other than Rabbit and user11982958). Unfortunately, even on a Core i7 and SSD, the configuration stage takes a good 30-45 mins or so to fail so really you need to try starting a JVM with comparable settings beforehand.
Note I've just successfully completed a clean install of the new 18.104.22.168.0 BP1 release on 32 bit OL 5.7.
The <ORACLE_HOME>/oms/sysman/omsca/scripts/wls/create_domain.py file still had the old memory settings, which I'm pretty sure won't work on this OS. I didn't want to waste my time so set them to the lower values as I described earlier in this thread and the configure was fine.
If anyone else is installing OEM 12c on a 32 bit Linux please post your experiences here.