Re: SCAN address

From: onedbguru <onedbguru_at_yahoo.com>
Date: Wed, 10 Aug 2011 16:02:04 -0700 (PDT)
Message-ID: <cf0453c0-ef90-4619-9765-871176a4d0dd_at_t5g2000yqj.googlegroups.com>



On Aug 7, 2:16 pm, Mladen Gogala <gogala.mla..._at_gmail.com> wrote:
> On Sun, 07 Aug 2011 08:38:26 +0000, Mladen Gogala wrote:
> > First, RDBMS based http and OEM will not work without SCAN address on a
> > RAC db. Second, with the requirement to put SCAN into DNS, DBA can no
> > longer install RAC without any help, the person who maintains the DNS
> > server, usually not DBA, will need to configure it.
>
> > I am not aware that anybody was requesting SCAN address, it's a
> > convenience, an endpoint for programmers to tie their services to, not
> > something that would help the user. Making a mess of the installation
> > and and making the SCAN address mandatory is infuriating, especially
> > when having in mind that this is done so it's safer to use cheap
> > Elbonian programmers. That is the reason for adding an additional level
> > of complexity to otherwise already convoluted, buggy and complex RAC
> > installation. In 10g,  RAC installation became a form of black art. In
> > 11g Oracle Corp. added more complexity to already convoluted and complex
> > environment, with even less real documentation. The real question is
> > whether this complexity has paid off? What did RAC 10g and RAC 11g
> > provide that RAC 9i couldn't have provided? The answer is ASM. The real
> > purpose of ASM is to block the competition. Competitors can and will use
> > OCFS but competitors cannot use ASM. Thus the complexity. Now, we got
> > SCAN address requirement, which is infuriating. I didn't even mention
> > the idiotic multicast flop, which belongs to the same category.
>
> I tried redirecting it with DBMS_XDB.SETLISTENERENDPOINT to VIP hostname,
> but to no avail. It still didn't register with the listener. My
> conclusion is that SCAN is being rammed down our throats. I will hack
> around it.
>
> --http://mgogala.byethost5.com

you do know that you don't NEED SCAN, right? you can still start the default listeners, the same way as before. SCAN is more analogous to a cluster-alias (remember DEC/VMS clustering?? ) As for the install of SCAN, you can use a single IP address in the /etc/hosts file to get it installed and add the DNS later. (been there, done that on many installations - it just ain't that hard).

Just curious, why would one use OCFS if they are not using Oracle? No other database can have an active-active cluster database - it is more than just access to the database files. There are locking mechanisms to prevent one db node from over-writing what another db node just wrote at nearly the same time.

I have been using ASM for quite some time now and on some very sizeable databases. I was a dba in a fairly well-known telco and used ASM on a 380TB data warehouse on a multi-node SUN 6900 cluster. The management of this sort of beast without ASM and ASSM, OMFand LMT, would be nearly impossible to keep one's sanity. We even moved the ASM devices from one storage vendor to another online while adding 1TB/ day. Try that with any other database or file system. I have seen many attempts using SAN storage tools - mostly failed (search for the recent story on Amazon's cloud fiasco when they tried to move to new storage online - their some x000 clients were without their data for more than 36 hours while they recovered.)

as for the "infuriating" , have you not heard of DBA2.0. If you are not a DBA 2.0, you need to work in that direction. DBA2.0 is a SYS/ NET/SAN/DB Admin all rolled into one... Not to mention that being a DBA2.0, you really start to understand the "system" not just our little world of databases. Received on Wed Aug 10 2011 - 18:02:04 CDT

Original text of this message