Re: [suse-oracle] Question regarding ASM on SLES 11

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Vladimir Barac [05.09.2011 10:09]:
> Huh, strange.
> 
> We initially had reboots due to OCFS2 being under heavy load - that was in early SLES 10 SP2 days.
> 
> With kernel updates that came later on, issues were gone.
> 
> I believe that what also helped us a lot were updated drivers for HBAs, network cards, BIOS and so on. 

I used OCFS2 (1.2) with SLES9 up to SP4. Now with SLES11 I use OCFS2 1.4
(or 1.5.0, if modinfo tells the truth). I use open-vm-tools (as
recommended by SAP), they include the drivers necessary for VMWare, as
far as these are not integrated by the kernel... all security patches
(e.g of the kernel) are applied on a regular basis, but it may take half
a year or longer until the next reboot...

The first issues were because of an incorrect parametrization of the RAC
nodes, no options set for the HBAs, for example, so that they tried
auto-failover while multipathing tried, too, what caused a bad crash of
everything. Then a lot of troble came from too small timeout values, and
the last good thing was the use of hugepages.

The last disturbance was caused by a flick on the ESX hosts' connection
(a blade center) with the RAID - it caused some strange inodes on the
backup volume and therefore a load > 150 until I noticed and powered off
the VMs. It was on a sunday morning at about 3:30, not a time I watch
the welfare of my systems normally ;-)

Regards,
Werner

> 
> 
> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf Of Werner Flamme
> Sent: Monday, September 05, 2011 10:42 AM
> To: suse-oracle ML
> Subject: Re: [suse-oracle] Question regarding ASM on SLES 11
> 
> Vladimir Barac [05.09.2011 09:00]:
>> What kind of problems with OCFS2?
> 
> Extreme high load (>600) when a bad sector is found on a volume (causing
> the nodes to timeout every attempt to log in), sporadic node reboots
> (especially the RAC nodes, not the SAP nodes), interconnect problems in
> the OCFS2 network (the net admins never allowed a switch problem as a
> reason).
> 
> When a bad sector is found, you'll never find out where it is from. At
> least we don't.
> 
> Maybe it's because we use ASM as well as OCFS2, we don't know.
> Everything went much more stable when the db admins discoverd huge pages
> and the value of higher timeouts...
> 
> I use OCFS2 (1.2) for my SAP nodes for a longer time now, and I had
> hardly any error. Maybe because it were only 3 nodes (all in one rack),
> and now we have about 40 (VMWare and physical, different racks,
> different switches).
> 
> It is hard to debug when you have other tasks to do in parallel :-\
> 
> Regards,
> Werner
> 
>> -----Original Message-----
>> From: [email protected] [mailto:[email protected]] On Behalf Of Werner Flamme
>> Sent: Monday, September 05, 2011 9:59 AM
>> To: suse-oracle ML
>> Subject: Re: [suse-oracle] Question regarding ASM on SLES 11
> 
>> Vladimir Barac [05.09.2011 07:47]:
>>> Good morning
> 
>>> Why bothering with ASM?
> 
>>> If on linux, use OCFS2. It really works fine.
> 
>>> ASM adds a bit of complexity, and removes flexibility.
> 
>>> SAP support is right, ASM is under Oracle's support. 
> 
>>> We are running OCFS2, as I mentioned before, for our SAPs here. Heavy load, really heavy load and no issues so far. 
> 
>> Vladimir,
> 
>> we use OCFS2 right now. But it is highly discouraged by Oracle, to use
>> OCFS2 and ASM in parallel. And on the RAC, there are about 20 non-SAP
>> databases running on ASM, so...
> 
>> Besides, every now and then we have problems with OCFS2 (some of them on
>> the ocfs2 mailing list ;-)), so our db team wants to switch to ASM. And
>> since SAP allows ASM, I'm out of arguments.
> 
>> I'd prefer SAP db anyway ;-)
> 
>> Regards,
>> Werner
> 
> 
>>> -----Original Message-----
>>> From: [email protected] [mailto:[email protected]] On Behalf Of Werner Flamme
>>> Sent: Friday, September 02, 2011 1:37 PM
>>> To: suse-oracle ML
>>> Subject: [suse-oracle] Question regarding ASM on SLES 11
> 
>>> Hi everyone,
> 
>>> we want to update our Oracle RAC 11.1 (2 nodes, running with Oracle
>>> Linux 5.x) to Oracle RAC 11.2.0.2.
> 
>>> Afterwards, we plan to migrate our SAP databases to ASM. Since we have
>>> no experience with ASM, there are some questions.
> 
>>> 1) when our DB runs inside ASM on the RAC nodes, do I have to install
>>> ASM on the SAP servers (SLES 11 SP1) too? If yes, there are the SLES
>>> packages oracleasm and oracleasm-kmp-$flavor, are these enough?
> 
>>> 2) Can I symlink a file outside ASM to a file inside ASM?
> 
>>> 3) Is it correct that we can't use ASM and OCFS2 simultaneously? At
>>> least, not on the RAC nodes?
> 
>>> Fucking SAP support says that these questions are not theirs to answer.
>>> Can someone please help me?
> 
>>> Regards,
>>> Werner
> 
> 
>> _______________________________________________
>> suse-oracle mailing list
>> [email protected]
>> http://listx.novell.com/mailman/listinfo/suse-oracle
> 
>> ______________________________________________________________________
>> This email has been scanned by the External Email Security System.
>> ______________________________________________________________________
> 
> _______________________________________________
> suse-oracle mailing list
> [email protected]
> http://listx.novell.com/mailman/listinfo/suse-oracle
> 
> ______________________________________________________________________
> This email has been scanned by the External Email Security System.
> ______________________________________________________________________
> 
_______________________________________________
suse-oracle mailing list
[email protected]
http://listx.novell.com/mailman/listinfo/suse-oracle

______________________________________________________________________
This email has been scanned by the External Email Security System.
______________________________________________________________________

- -- 

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.16 (GNU/Linux)
Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org/

iEYEARECAAYFAk5kjMQACgkQk33Krq8b42OMoACeO1lVgvOEDASpx3xjsUas5QCc
pmAAnRH3u8B6epzietPebLGD+bQhpcFf
=069C
-----END PGP SIGNATURE-----
_______________________________________________
suse-oracle mailing list
[email protected]
http://listx.novell.com/mailman/listinfo/suse-oracle