Sun Cluster + QFS or Sun Cluster + Veritas



  • 1. How to get the exact trap code from an mdb ::findstack call frames
    Hi, We have two UltraSparc III single CPU systems running Solaris 8 and 9, that become non-responsive over time while running our app. The crash dump analysis using mdb consistently reveals that the same thread in the RT scheduling class is the thread running at the time of our sync. This is the threads ::findstack. > 2a1007cafb1 $C 10 000002a1007cb051 preempt+0x88(30086c47ba0, 30086c571c0, 20, 1, 2a10007dd20, 3000329db88) 000002a1007cb101 trap_rtt+0x74(30086c571c0, 1, 30086c571c0, 30005d7e000, 2a1007cbba0, 30003de3ba0) 000002a1007cb1b1 trap+0x16f0(30003f6df00, 2a1007cbaf0, 2a10004bd20, 10000, 2a1007cbba0, 0) 000002a1007cb2f1 user_rtt+0x20(1, 62, 7a801bd0, 5190c, 0, 0) We may be in a cycle of continuous traps. Does anyone know how to determine the trap code from the above stack trace? Thanks! Allen
  • 2. SB1000 connected to a TV Monitor input (SVGS 800 x 600)
    Hi: I have Solaris 10 standard running on a SunBlade 1000 server, and this server's SVGA output is connected to my TV SVGA input (Proscan PS3600). Two questions: (1) I can see ASCII output in tty mode (before dt/X starts), but its too big and wraps around the screen. So, for example, a "ps -ef | more" scrolls too far and also it wraps around the screen because the font is not correctly set. I tried 'stty rows|cols' as well as changing the equivalent values at the OBP level, both with no luck. Does anyone know how to adjust this? Im used to the serverless world. (2) DT/X is not able to start, I think for the same reason since, when I try to start it, the TV clicks a few times in trying to sync up, then gives up and goes back to text mode - and I get a message indicating that I cannot start X on DISPLAY :0. Any ideas here? I thought xdmconfig(1M) would help but Im not sure if that comes on SPARC distributions (I can't find it on my system and I loaded the Entire Distribution less OEM support). Thanks in advance, Noelle
  • 3. Perl's DBD::mysql will 'make' but won't 'make test' on Solaris 10
    Hey all, I've posted this previously on a mysql group to no avail, so I thought I might ask and see if anyone here has any ideas since it seems to be having problems with a Solaris supplied library ( I am running solaris 10, with the solaris cc compiler. the module compiles fine. however, the first test of 'make test' fails as below. I've cut most of the bloated output, but if anyone thinks it would help I could certainly post it. If this has nothing to do with Solaris let me know, and accept my apologies. TIA, -jp Running make test PERL_DL_NONLAZY=1 /usr/bin/perl "-MExtUtils::Command::MM" "-e" "test_harness(0, 'blib/lib', 'blib/arch')" t/*.t t/00base.............install_driver(mysql) failed: Can't load '/export/home/root/.cpan/build/DBD-mysql-3.0002/blib/arch/auto/DBD/mysql/' for module DBD ::mysql: perl: fatal: DF_1_NOOPEN tagged object may not be dlopen()'ed at /usr/perl5/5.8.4/lib/sun4-solaris-64int/ line 230. at (eval 1) line 3
  • 4. bash v3 from Solaris 10 does not read bashrc
    Hello Just trying out Solaris 10 and I've noticed that my .bashrc file is not read when I start a bash shell. Under Solaris 9 this works as expected. Is this a bug in Solaris10? Any ideas? Thanks.

Sun Cluster + QFS or Sun Cluster + Veritas

Postby Sanjay » Wed, 01 Mar 2006 22:52:46 GMT

Right now I've got some 3.5 Veritas Clusters running Oracle8
that are due for an upgrade.

I'm playing around with the idea of getting rid of all of
our Veritas software and replacing it with Sun software
(since cluster is now free).

So it would be:

Solaris 10 (01/06 HW release, latest patches)
Sun Volume Manager
Sun Cluster 3.1

I've used Veritas for so long now & am weary of mixing it in
with Sun Cluster.  Ideally - I'd like to just go with one vendor
for my clustering.

Is MPXIO comparable to DMP?  Are people actually running Sun volume 
manager under production Oracle DB's?  I've used SVM for mirroring
boot drives on web servers and app servers but never in the database

I've read that QFS can do the volume management all on it's own -
is that what people are typically using with Sun Clusters?

I'm reading the FAQs and white papers right now but would love to
hear from some people with real world experience with the stuff.


Re: Sun Cluster + QFS or Sun Cluster + Veritas

Postby dol » Thu, 02 Mar 2006 00:41:33 GMT

In article <OAYMf.183130$ XXXX@XXXXX.COM >,



Yes. It is in use by customers. If you want to go to QFS as a cluster 
backend file system, it is only supported with SVM and not Veritas.

Since QFS cannot do mirroring it is typically used on top of
SVM mirrored metadevices.

If you plan to use Oracle RAC you need the special Oracle RAC SVM Edition.
This has now been replaced/renamed by/to Sun Cluster Advanced Edition 
for Oracle RAC. (internally known as OBAN).



Fredrik Lundholm   
dol @

Re: Sun Cluster + QFS or Sun Cluster + Veritas

Postby Sanjay » Thu, 02 Mar 2006 06:12:55 GMT


I noticed they had not certified 10g on their website yet.

Is the general consensus now that if you run Sun Cluster you go with Sun
for VM/MpXIO & QFS?

Or do a lot of people still mix and match w/Veritas.


Re: Sun Cluster + QFS or Sun Cluster + Veritas

Postby Robert Milkowski » Thu, 02 Mar 2006 13:10:44 GMT

I use SVM+SC+Oracle for some years - works perfect.

                                                Robert Milkowski

Re: Sun Cluster + QFS or Sun Cluster + Veritas

Postby SST » Thu, 02 Mar 2006 21:18:24 GMT

We currently are running two Sun Clusters, one version 3.0 with Oracle
OPS on Solaris 8 and a second with Cluster 3.1 and Oracle RAC on
Solaris 9. Both clusters use Veritas VM for storage management but use
the MPxIO drivers for multipathing.

Veritas used to make the argument that the DMP multipath driver offered
better load balancing. From what I have read on MPxIO (been out of
Solaris world for a while), round-robin is the default (only??) load
balancing model. Truthfully, not sure if it makes much difference in
95% of the cases.

If using OPS/RAC, you need your DBAs to bring down Oracle instances for
node or cluster shutdown. Instance control not included as part of a
package/resource group in this configuration. Just means more
coordination and people involved during restarts.

Installation and maintenance of pkgs/resource groups are a bit more
tedious in SC. You need to balance that against the clusters various
capabilities for diagnosing and handling faults. For instance, in HP
Service Guard, I could probably set up a custom package with storage
and networking in a day. Based on limited experience with Sun Cluster,
it might take two.

The clusters have been solid. No unplanned outages for quite some time.
DBAs and end-users are happy with performance.

Re: Sun Cluster + QFS or Sun Cluster + Veritas

Postby Sanjay » Fri, 03 Mar 2006 22:27:03 GMT

On top of UFS?  Or on top of QFS?  I know UFS supports journaling and
has come a long way but I wasn't aware that a lot of people are 
running production Oracle instances on top of UFS.


Re: Sun Cluster + QFS or Sun Cluster + Veritas

Postby Sanjay » Fri, 03 Mar 2006 22:30:25 GMT

That's an iteresting mix ;-)  I was thinking of going with QFS for
my RAC implementation.  I've used Veritas Storage Foundation for
RAC but management is trying to get away from Symantec/Veritas
for "business reasons".

Not in my case.  Availability is definitely a concern, but performance
isn't as important - the stuff just doesn't get many queries.

But if I did RAC inside of Sun Cluster I should be able to bring down
the instances thru the clutsering software, right?

How else would SC manage fail-over then?

Thank you everyone for your feedback.


Re: Sun Cluster + QFS or Sun Cluster + Veritas

Postby Robert Milkowski » Sat, 04 Mar 2006 14:07:24 GMT

First for some years I used raw-devices on top on SVM.
Last time I switched to SVM+UFS(forcedirectio).

                                                Robert Milkowski

Re: Sun Cluster + QFS or Sun Cluster + Veritas

Postby SST » Tue, 28 Mar 2006 21:40:46 GMT


Just wanted to follow up. Sorry I did not do it sooner.

In regards to the multi-pathing solution, when I said " not sure if it
makes a difference" I was speaking about the load balancing policy.
Multipathing is indeed the important part, but how the load balancing
is done across those paths is not that big a deal in most cases.

In regards to OPS, I am pretty sure OPS requires raw devices. Not sure
about RAC.

In OPS on Sun Cluster, you do not control the instances through the
cluster. The instances must me brought up and down by the DBAs after
the cluster has started. The cluster only controls the cluster
framework, which includes the  distributed lock manager, but not the
actual instances.


Similar Threads:

1.Clustering Sun: Sun cluster or Veritas?

Any of you have experience with Sun Cluster software and/or Veritas

I have worked with IBMS HACMP but none of the other solutions.  Even
that was fairly simple active/passive clusters, nothing too fancy.

Any pros/cons to running either Suns cluster software or Veritas' on
Sun systems? Probably either V880s or V1280s.  I have been looking at
the docs. on Suns site about Sun Cluster 3.1 but of course that is
just technical info, always best to hear from others with real
experience too :)


2.Best Practices "Migrating Veritas VCS Cluster to SUN Cluster"


Do you have any Best Practicies or Blueprints Paper?

Thanks for your support


3.Veritas Cluster versus SUN Cluster

hi all

we would like to migrate Veritas Cluster 3.5 with VxVM, VxFS to Sun Cluster
3.0. has anyone a cockbook or a migration ?

thanks in advance

4.QFS /Sun Cluster 3.1

I am in the process of choosing between RAW devices or a QFS (Sun
Cluster 3.1/Solaris 9) to support a busy online transactional Oracle
RAC 10g application.

I want to know of any QFS drawbacks or performance issues.  Again, this
will be a very busy online transactional system supporting in excess of
3000 concurrent users.

I would like to hear from anyone with a similar setup.

Nabil Courdy

5.Problem during Veritas VM upgrade on a Sun Cluster 3.0


I have been upgrading Veritas Volume Manager 3.2 to 3.5 on a Sun Cluster 3.0
node. This is what I did:

- boot the node into single-user nocluster
- install patch 110648-33  :  should enable SC 3.0 to support VxVM 3.5
- reconfiguration boot
- back to single user nocluster mode, performed the VxVM upgrade according
to the manual method described in the "VERITAS Volume Manager 3.5
Installation Guide" pages 58 through 61.
- after another reconfiguration boot, I get this ginormous amount of
messages on the console:

Jul 13 15:44:31 clnode1 SUNWscucm.ucmm_reconf: cvm requests reconfiguration
in step cmmstep3
Jul 13 15:44:31 clnode1 Cluster.OPS.UCMMD: prog <ucmm_reconf> failed on step
<cmmstep3> retcode <205>
Jul 13 15:44:32 clnode1 ID[vxclust]: starting return time: 07/13
15:44:32.515:  seq # 80
Jul 13 15:44:32 clnode1 ID[vxclust]: ending step return time: 07/13
Jul 13 15:44:39 clnode1 ID[vxclust]: starting step1 time: 07/13
15:44:39.516:  seq # 81
Jul 13 15:44:39 clnode1 ID[vxclust]: merging joiners and leavers
Jul 13 15:44:39 clnode1 ID[vxclust]: members 3 joiners 3 leavers 0
Jul 13 15:44:42 clnode1 ID[vxclust]: starting step2 time: 07/13
15:44:42.514:  seq # 81
Jul 13 15:44:42 clnode1 ID[vxclust]: calculating master time: 07/13
15:44:42.515: , nnodes = 2
Jul 13 15:44:42 clnode1 ID[vxclust]: port number is 5585
Jul 13 15:44:42 clnode1 ID[vxclust]: smallest member time: 07/13
Jul 13 15:44:43 clnode1 ID[vxclust]: link 1 connected 7
Jul 13 15:44:43 clnode1 ID[vxclust]: link 1, votes 1
Jul 13 15:44:43 clnode1 ID[vxclust]: CVM:MASTER=1 SELF=0
Jul 13 15:46:12 clnode1 ID[vxclust]: timeout time: 07/13 15:46:12.582:  step
Jul 13 15:46:12 clnode1 ID[vxclust]: t/o in master step - retrying
Jul 13 15:46:12 clnode1 SUNWscucm.ucmm_reconf: cvm requests reconfiguration
in step cmmstep3
Jul 13 15:46:12 clnode1 Cluster.OPS.UCMMD: prog <ucmm_reconf> failed on step
<cmmstep3> retcode <205>

And so on, ad nauseam.

Does anyone have any clue at all as to what the heck is going on?

thank you in advance


6. SUN Cluster mount Veritas Volume as raw-device

7. Sun Cluster 3 vs.Veritas VCS 3.5

8. scconf crash: registering a Veritas local disk on Sun Cluster (localonly=true)

Return to unix


Who is online

Users browsing this forum: No registered users and 10 guest