z/vm in production server

[Former Member]

hello all,

does anybody have experience with running PRODUCTION suse linux servers under Z/VM ?

are you satisfied with the results?

is there anything to watch out for?

thanks,

[Former Member · Accepted solution]

Hi Omer,

we have a number of customers running this scenario (SuSE Linux on System z under z/VM) in production.

See for example [http://www-01.ibm.com/software/success/cssdb.nsf/CS/STRD-83LL69?OpenDocument&Site=corp&cty=en_us].

The story behind this link also answers the question about z10 from your next append.

Please contact me directly if you need more details or want to get into direct contact with reference customers.

best regards, Harald

[brian_walker · Accepted solution]

Sounds like you're looking at an all Z installation/implementation.

We started that way with z/VM 5.2 and SLES 10. I would be interested in knowing whether or not anyone has been able to successfully follow all of the IBM redbooks and implement hipersockets, OSPF, NFS storage and SAP ASCS/SCS on z/OS, etc. We ran into several opportunities (some eventually fixed, others worked around) before we switched to Windows 2003 x64 application servers (the switch was for a completely different reason).

Other 100% Z customers we spoke with ended up NOT following some of the redbooks and did NOT implement OSPF, ran their NFS servers and ASCS/SCS on a zLinux bare-metal LPAR instead of under z/OS, etc.

If we had it to do over again, I think we'd liked to have seen the entire SAP stacking running and working together in a lab with all of the redbook instructions/recommendations implemented.

Brian

[brian_walker · Accepted solution]

The customer referenced in the link above (http://www-01.ibm.com/software/success/cssdb.nsf/CS/STRD-83LL69?OpenDocument&Site=corp&cty=en_us) is one we've spoken to before that did NOT implement all of recommendations from the IBM redbooks for SAP and HA (aka near-continuous availability). Specifically, they did not implement OSPF and they run their NFS and ASCS/SCS on a bare-metal zLinux LPAR instead of under z/OS. They also have only 1 Z10 so they have a "sysplex in a box".

Brian

[Former Member · Accepted solution]

Brian,

the customer reference that I put into my append was in response to the subject of this thread and I think it answered Omer's question.

You are right that the referenced customer did not completely implement the "Industry reference architectures" (see [|]) or the HA scenario described in the [Business Continuity for SAP on IBM System z|ftp://ftp.software.ibm.com/eserver/zseries/software/sap/Business_Continuity_for_SAP.pdf]. But on the other hand I would never expect that ALL of these recommendations fit for ALL customers.

In your previous append you were asking for an environment in which these recommendations are implemented and I can assure you that we have such an environment live in the IBM Lab in Boeblingen/Germany.

best regards, Harald

[brian_walker · Accepted solution]

Harald,

Thanks for that information. Did IBM/SAP need to do anything special to the startup dependencies in SUSE for ospfd(from Quagga/Zebra)? An issue we ran into is that the default startup order for ospfd had it starting AFTER some network services (nfs client comes to mind) and/or stopping BEFORE those same network services stopped. This would cause the Linux system to hang for a while until something timed out during startup and/or shutdown. We were running SLES10.

Also, I believe I was responding to the question regarding "anything to watch out for" and relating my experiences. Along those lines, I believe the customer in the link you provided is also still taking downtime in the early morning hours every Sunday to perform some offline reorg work in DB2 z/OS on some very large tables. Not necessarily related to running under z/VM but I still think it is worth mentioning since DB2 z/OS is the only database which can be used for SAP running under z/VM and zLinux. We have some non-DB2 z/OS SAP systems in our landscape for which the database does not require explicit runstats or reorgs (these are automated in the database itself). I've heard this functionality will be included/provided in DB2 z/OS v10 when it is released.

Brian

[Former Member · Accepted solution]

Hello Brian,

the startup dependencies issue does indeed require special attention and we have recommended a way to address it in our SAP SDN Paper [Implementing High Availability for SAP NetWeaver 7.1 Technology on System z|http://www.sdn.sap.com/irj/scn/index?rid=/library/uuid/d0935882-878a-2c10-3b80-e4def4262679&overridelayout=true]. (see the chapter "Environment Adaptations for AIX and for Linux on System z" in this document)

best regards, Harald

Edited by: Harald Duvenbeck on Apr 15, 2010 1:07 PM

[brian_walker · Accepted solution]

Harald,

Thanks for that information. I look forward to reviewing it. Perhaps it should be included in the next Business Continuity document?

It is also of interest to note that the Business Continuity document you linked to was created in October of 2008. This appears to be the first version of the document based on DB2 z/OS v9.1 and Netweaver 7.0 (at least as stated on page xiii). Unfortunately for us, we started on DB2 z/OS v9.1 and Netweaver 7.0 in July of 2007. Since this documentation did not exist at that time, it would certainly explain some of the issues we encountered. Hopefully many of these "watch outs" now have documented solutions and/or workarounds.

Regards!

Brian

[brian_walker · Accepted solution]

Harald,

I have reviewed the SDN paper you linked above. That documentation is very good! I wish it had existed (and for NW 7.00 and not 7.10) back in 2007 when we did the majority of our SAP installs on system Z. It think it would have eliminated many of the issues we encountered.

Brian