RE: intel clusters in a box

  • From: "Marquez, Chris" <CMarquez@xxxxxxxx>
  • To: "Don Granaman" <granaman@xxxxxxx>, "oracle-l" <oracle-l@xxxxxxxxxxxxx>
  • Date: Fri, 21 Jan 2005 16:11:58 -0500

Wow, it's a small Oracle world out there.
Don the "OraSaurus" gave me some great info while on the suse-oracle
list back in 2001!
We discussed the (DBA) realities of 9i RAC and 7/8i OPS...GC_*
parameter...the good old OPS days.
Not many people know this but "cache fusion" was actually introduced 66%
in 8i!

Anyway, to what Jared said;
>> Suse should be given serious consideration.
>> It is in many respects superior to RedHat.
I agree. In general SuSE commitment to Oracle is well documented, but
not we marketed.
I was not long ago that RedHat, had and IPO and their own database and
gave little consideration to Oracle.
SuSE and Oracle is great!

But the reality is that times change and people/companies partner.
It clear *to me* that Oracle, Dell, RedHat all love each other.
And it seems lesser known that Oracle SuSE, IBM or Oracle SuSE, Compaq
like each other.
And in the world of "production" RAC this "partnering" is critical!

*HOWEVER*, I'm sad to report to Don that his experience of years past is
similar reality of today.
While "Oracle, Dell, RedHat" is much more stable today, when you have
problems *and* you have support, sometimes you are just more frustrated.
In my recent TAR, Oracle invoked the "JEC", Oracle/Dell "Joint
Escalation Center"!  But all the JEC seems to do was open a tick with
Dell for us!
In the end it was not and Oracle issue at all, but a hardware controller
issue, which Dell Support did NOT diagnose NOR explain.  Our SA grinded
it out himself.

What is my point;...
While support is expensive, critical, and often sucks (sorry), we found
there is an added benefit to certified supported configs.
It means that you are likely to find others "in the same boat".  And
late in the day when the DBA and SA are starring at a broken "RAC", and
Oracle and Dell are asking for RDA's, you can jump on lists like this,
or start hitting Google and help yourself! Because other people have
configs and problems like you!

So when going RAC, go certified and go with the masses!

PS Surprisingly we found that Dell was of very little help in defining
a "OracleRAC-Dell-RH Certified Config"!

If anyone is interested this is my config...and there is almost nothing
in this listed that we did have to "over come".
2 X Dell 2650
2 X PERC 4/DC (Dual Channel) RAID Controller
1 X PowerVault 220 (13 X 73GB 10k SCSI HD)
4 X Broadcom BCM5700 10/100/1000BASE-T Ethernet LAN controller
1 X Network (Interconnect) Switch Linksys
Oracle EE 9205
Oracle Cluster Manager 9205
Oracle OCFS-Oracle Cluster FileSystem 1.0.12-PROD1 _or_ Oracle
OCFS-Oracle Cluster FileSystem 1.0.13-PROD1
Red Hat Enterprise Linux ES release 3 (Taroon)

PPS We just moved our prod db back on these severs after fixing the
battle described above.

Chris Marquez
Oracle DBA
HEYMONitor(tm) -
"Oracle Monitoring & Alerting Solution"

-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Don Granaman
Sent: Friday, January 21, 2005 7:45 AM
To: stellr@xxxxxxxxxx; oracle-l
Subject: Re: intel clusters in a box

Actually, the Dell/EMC:Clariion)/RedHat Oracle-certified 9i RAC cluster
platform has been around for a while.  We've had such a critter for
almost two years.  Its OK now, but there were growing pains.  Linux is
not Solaris, or AIX, or ...  In hindsight, its humorous.  At the time,
it was exasperating.

It didn't show up in testing (albeit, not at quite full functionality
and load concurrently), but shortly after we went production, we started
having some very serious issues - instance crashes, 10,000
session/connection failures per hour, nodes losing their LUNs (very
messy), sudden and mysterious node crashes, even a few memorable
whole-cluster crashes.  It turned out to be a buggy driver
intermittently corrupting memory and trashing whatever happened to be
there.  It happened on either or both nodes, at random, sometimes lying
dormant for a week or two everywhere. This lasted for most of Q1 2004.
Oracle support had never heard of their (then at least) oft-espoused
"one stop" support for Oracle and Linux (and, especially, for RAC on
Linux).  I sent the URL ( don't have
it handy right now though.) for the (then) year+ old press release to
them before they even knew that they (supposedly) supported the Linux OS
at all.  They then suggested that we call RedHat or Dell.  We did - and
some weeks later *finally* got a diagnosis and a fixed driver (2nd

The next call to Oracle support, on an unrelated and relatively minor
issue, lead to a preliminary diagnosis that we had a "tainted kernel".
We would have to restore everything to its original certified and
thoroughly crash-prone state to be supported.  We refused and closed the

Eventually though, they forgave us.

-Don Granaman

----- Original Message -----=20
From: "Ray Stell" <stellr@xxxxxxxxxx>
To: "oracle-l" <oracle-l@xxxxxxxxxxxxx>
Sent: Thursday, January 20, 2005 1:52 PM
Subject: Re: intel clusters in a box

> No bites on this.  I emailed a dell rep.  He suggested I start reading

> at
> Seems redhat/dell/emc are working together to produce such offerings.
> On Tue, Jan 18, 2005 at 11:14:36AM -0500, Ray Stell wrote:
> >
> > I've been building my own intel/linux clusters for testing.  I'm=20
> > wondering if you have tested any commerical offerings for the whole=20
> > 9 yards; processers, FC, and storage.
> >
> >
> >
> >
> >
> > "More commonly, however, you hear that the expression comes from the

> > capacity of ready-mix concrete trucks. Concrete trucks supposedly=20
> > contain nine cubic yards when fully loaded."
> =
> Ray Stell       stellr@xxxxxx        (540) 231-4109        28^D
> --
> //


Other related posts: