Re: [suse-oracle] 11gR1 RAC on SLES

I dont know any problems with 11g install on sles10 sp2, but you MUST UPDATE system from sp2 to online-update (kernel 
39.3 is recommended or higher) - sles10 sp2 had the bug in aio.


  ----- Original Message ----- 
  From: Peter Hawkins
  To: [email protected]
  Sent: Wednesday, August 19, 2009 6:29 AM
  Subject: Re: [suse-oracle] 11gR1 RAC on SLES


  Hi,

  We're running 11g on SLES 10 SP2. We were initially on SP1, but did an upgrade to SP2.

  The install has a few gotchas, but I think oracle have fixed a few of the issues (I had most problems with the 
clusterware layer).

  I hit a few more problems with upgrading from 11.1.0.6.0 to 11.1.0.7.0, but the system still runs okay (although the 
services have got a bit fried).

  FYI, we're running two instances joined to an EMC SAN and are using ASMlib (means you have to synchronise the asmlib 
driver and kernel version if you want an oracle supported setup).

  Regards

  Peter Hawkins
  Senior Oracle DBA

  Homeserve
  Sunnyside Road North
  Weston-super-Mare
  BS23 3QY
  Tel: +44 1934 637108
  Mob: +44 7748 100122
  Fax: +44 1934 637188



  2009/8/19 Ciro Iriarte <[email protected]>

    2009/8/17 Alexei_Roudnev <[email protected]>:

    > I have automated installation for sles10 sp2 / oracle rac 10.2.0.4
    >
    > Works fine but require a few unusial steps (not because of SuSE but because
    > of RAC):
    >
    > 1) Upgrade to the kernel *39.3 (at least). original sles10 sp2 release had
    > broken aio.
    >
    > 2) I am nlt sure about 11R but 10.2.0.1 had crs not cpompatible and had a
    > bug in ocr fortmat. Solution - just copy files from 10.2.0.4. 11G don't have
    > this
    > problem, more likely.
    >
    > 3) Starting with 11.1.0.7 and 10.2.0.4, oracle switched from hangcheck-timer
    > to home-grown (!@#$) oprocd. By default, this beast wil kill the system
    > after any minor delay (I find out that it kiils in pretty easy). So you must
    > set up diagtimer 13 to make sure that oprocd runs with the margin 10 seconds
    > and not 1/2 second.
    > You can find a lot on the internet.
    >
    > 4) The whole CRS is a mess of bugs, glitches and so on. Timeouts influence
    > each other, oprocd had strange design and kills the system long before it is
    > necessary for the fencing. If you worn on Vmware, you must disable oprocd
    > totally; to make it al more funny, they change the variable DISABLE_OPROCD
    > meaning so that now 'true' means 'run it' and nkt 'skip it' and vice versa
    > (so I just 'chmod a-x oprocd' and enabled old hangon timer on VMware).
    >
    > In short, it is a lot of fun. I managed to install RAC in 3 hours from
    > scratch, but it is 40 steps script (*whcih do it). Good point is that, not
    > sure about 11.1.0.7, but 10.2.0.4 + CRS CPU patch + 88 database patches - do
    > work well (after all).
    >


    It would be nice to see papers with this kind of setups
    (current/latest Oracle on SLES, RAC and/or ASM),  for example, people
    from my office don't like to "experiment",  they demand
    "working/certified solutions" :s

    Having a paper to entertain them would be a good entry point for
    "experimenting" :P

    Regards,

    --
    Ciro Iriarte
    http://cyruspy.wordpress.com
    --

    _______________________________________________
    suse-oracle mailing list
    [email protected]
    http://listx.novell.com/mailman/listinfo/suse-oracle





------------------------------------------------------------------------------


  _______________________________________________
  suse-oracle mailing list
  [email protected]
  http://listx.novell.com/mailman/listinfo/suse-oracle

_______________________________________________
suse-oracle mailing list
[email protected]
http://listx.novell.com/mailman/listinfo/suse-oracle