As soon as EBS Release 12 became available, I started with the installation, eager to find out what the differences were compared to 11i.
I had had the opportunity to attend some trainings sessions and presentations about the new technology features, and, well, you know how things go when you are a techy… You cannot wait until you get the opportunity to get your hands on it..
I was planning to install it in a Multi-node installation, on two nodes, having RAC in mind for the future (two database servers and two application servers, all on 2 nodes).
Since I do not have sufficient hardware at my direct disposal, I decided to give it a try on VMware Workstation 5.5. I used this environment earlier when I implemented Oracle EBS 11.i on 9i/RAC and 10g/RAC and ASM/RAW/CFS. This is my test environment to get experienced with the technology.
So, I removed my existing 11.i environment and started all over again with Release 12 🙂
The first impression: What a huge chunk of software that is! phew… In order to install a Vision Demo environment, I was required to free up no less than 175GB of diskspace…
Then the installation finally started. I was using GPFS to install the shared APPL_TOP, and put the database on local storage, to migrate it to ASM as soon as the installation was done.
However, after the installation of the first node, DB and AS (which took surprisingly “short”, by the way;-), I ended up with an Apache server that failed to start.
The log files showed me that it wanted to create a file which it could not, for no aparent reason:
Apache logfile (adapcctl.txt) showed me the following:
Error
–> Process (index=1,uid=5,pid=4264)
failed to start a managed process after the maximum retry limit
Log:
/apps/inst/apps/PROD_db01/logs/ora/10.1.3/opmn/HTTP_Server~1
The HTTP_Server~1 logfile showed me the follwing information
Ouch! ap_mm_create(1048576, “/apps/inst/apps/PROD_db01/logs/ora/10.1.3/Apache/mm.3658”) failed
Error: MM: mm:core: failed to memory map memory file (Invalid argument): OS: No such file or directory
The “Error:” was a result of the failure in creating the file. That was clear enough. However, the question that remained (and still remains…) unanswered is: Why is it impossible to create this file? The directory is there, the ownership and permissions are set correctly (by rapidinstall), and still it is impossible to create a file? Ok, 1G in size, I guess from the log, but still, that should not be a problem, should it?
After 4 weeks of research, I decided to give up RedHat Enterprise Linux 3 AS, and give Oracle Enterprise Linux a try. I didn’t feel like configuring GPFS this time (no experience yet, and the pressure is gaining…), so I gave OCFS2 a try. This comes with the OEL distro.
I configured 2 volumegroups, each with one logical volume. One for the staging area, the other to hold the Shared APPL_TOP. I configured OCFS2 (There are some nice articles describing the setup of OCFS available on various websites), and started the installation.
To my big surprise I found out that I bumped into exactly the same error message as before…!?!?
I upgraded to (actually completely reinstalled) Oracle Enterprise Linux, moved away from GPFS, took OCFS2 instead, and still it was failing on me. With exactly the same error message…
I decided to take one more look into the system, and thought of one last trick. I didn’t consider it to work, but still, one can never be certain… After this, I would have given up the idea of at least having a shared APPL_TOP….
My absolute last trick saved me at last…
Desperate as I was, I deleted /apps/inst/apps/prod_db01/logs/ora/10.1.3/Apache and put a symbolic link into its place, pointing to /oracle/Apache, which was on local filesystem.
Tadaa! That worked. Now my Apache Server started without problems.
I will now try to find out whether this reproduces on actual hardware. I will keep you updated on this.
Hi, I am trying to install R12.1.1 on external HDD. I bumped into this error
/Transcend/d01/inst/apps/VIS_orappl/ora/10.1.3/Apache/Apache/bin/apachectl startssl: execing httpd
Ouch! ap_mm_create(1048576, “/Transcend/d01/inst/apps/VIS_orappl/pids/10.1.3/Apache/logs/mm.32297”) failed
Error: MM: mm:core: failed to memory map memory file (No such device): OS: No such file or directory
You mentioned to create a symbolic link to /oracle/Apache. But there is not /oracle/Apache in my HDD. And how do i create a symbolic link
pls help
Please take a look at http://oraclever.blogspot.com/2009/05/e-business-suite-release-1211-it-is.html for a better solution to the issue here.
The key is to locate your Instance Home ($INST_TOP) on a local file system. It doesn’t need to be on a shared file system, even when your Application Tier File System is shared.
Timothy, Not quite. GPFS is also available for Linux, and officially only supported on Linux on xSeries hardware, though it works very well on other hardware as well (I have run it on my Dell laptop with VMware RHEL and Oracle Enterprise Linux Guests).
GPFS is for AIX – GFS is for linux
Timothy,
Not sure whether you will be monitoring this thread still, but you could always consider GPFS by IBM. I am not entirely sure whether it is supported on non-IBM hardware, but you can always try to find out. GPFS, in my humble opnion, is a very solid cluster filesystem.
I got this to work – I do have a question though – on metalink it states that Apache 10g is not supported with ocfs2 – so my question would be this – if I show up at a customer site and they want a shared application tier on linux – what should I use – besides NFS
Timothy,
Thanks for your reply. 11.5.10.2 doesn’t show this behaviour, because it is not running Oracle Application Server 10G 10.1.3. It is using the older version still (1.0.2.2.2 – 9i). Therefore you will not have this behaviour with this version.
Arnoud
I ran into the exact same issue. But never did come up with a fix. Here is something interesting about this ocfs2 shared application file system – if you do it with 11.5.10 CU2 – you will not run into the Apache error – but for R12 you do. I used NFS because I could not figure out a work around. I may try this again today with your solution.
Naqi,
I am almost convinced that this hadn’t to do with Vision or Fresh install. The problems were in the Apache server, and as far as I am aware, there is no difference for the Apache server when it comes to comparing Vision against Fresh.
Regards,
Arnoud
We did an install recently, different from yours. R12 on Red Hat AS 4 – test instance. We had no room for vision. This was a single node installation. Apart for a slow install (low of resources), the installation completed without any errors. So not sure if this could be something specific to vision.