RAC Services

[Jeff Bisch]

Hello,

  I am have installed a two node Oracle RAC, 11gR1 (11.1.0.7.0), 64bit on Windows 2008.  My question is regarding which services should be set to AUTOMATIC, and which services should be set to MANUAL.

Currently, the services on both nodes are set to the following:

ORACLEOBJECTSERVICE - AUTOMATIC

ORACLEVSSWRITER - MANUAL

ORACLECLUSTERVOLUMESERVICE - AUTOMATIC

ORACLECRS - MANUAL

ORACLECSS - MANUAL

ORACLEEVM - MANUAL

ORACLESERVICE+ASM1 - MANUAL

ORACLESERVICE<DATABASE> - MANUAL

ORACLELISTENER1 - MANUAL

  My issue with the services being set this way, is that RAC is supposed to be high availability, and with these settings, if a node were to fail, for whatever reason, the services would not restart.

  I was told that the services need to be this way so that if there is a failure caused by a corruption of some sort, that it would prevent the corruption being propagated to the other node.

  I think that this configuration defeats the purposes of using RAC - I am right, or wrong?

  So, back to my original question, in an environment described above, which services should be set to automatic, and which services should be set to manual? I have not found any clear definition in Oracle's literature.

Thanks in advance for your help,

  jeff

[ddf]

We run a 3-node RAC on Linux and none of the instances are set to
autostart. RAC is designed so that if one node fails the remaining
nodes pick up the slack while the out-of-service node is repaired and
restarted. This does prevent any corruption from propagating through
the system.


David Fitzjarrell

[Jeff Bisch]

Thanks for your response!

 

I have one further concern, and I know that this would be extremely rare -

 

What if your data center has a power hit, and it causes all of the nodes in the cluster to reboot at the same time?  If none of them are set to auto start, then you are left with a system that is down until you can get someone to connect and bring the nodes back on line.

 

Given that our systems are 24/7, I am just trying to be prepared for the worst case scenario.

 

Thanks again,

    jeff

--
You received this message because you are subscribed to the Google Groups "ORACLE_DBA_EXPERTS" group.
To post to this group, send email to oracle_db...@googlegroups.com.
To unsubscribe from this group, send email to oracle_dba_experts+unsub...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/oracle_dba_experts?hl=en.

[jgar the jorrible]

On Oct 17, 9:02 am, Jeff Bisch <jbis...@yahoo.com> wrote:
> Thanks for your response!
>  
> I have one further concern, and I know that this would be extremely rare -
>  
> What if your data center has a power hit, and it causes all of the nodes in the cluster to reboot at the same time?  If none of them are set to auto start, then you are left with a system that is down until you can get someone to connect and bring the nodes back on line.
>  
> Given that our systems are 24/7, I am just trying to be prepared for the worst case scenario.

Theoretically, you should have redundant power, and that should never
happen.

Of course, we have an ups, and a phone line set up to call when the
ups gets set off so someone can come in and deal with it before the
ups dies. So some utility was working in our parking lot, cut the
phone lines, then cut the power, and it was off for hours longer than
the ups could handle, and nobody knew.

I worked in another place that had a big new Detroit Allison generator
for the building. The generator would turn on to test itself once a
week. It was supposed to not be connected to live power when it did
that. Oh well. It set off the alarm on the UPS, but since nothing
seemed to be wrong, my boss eventually taped over the silence alarm
button. So we had two unsynchronized 3-phase feeds out of phase with
each other, with the generator frequency varying from 58 to 61 Hz, all
feeding into the same building wires.

It's common to not autostart, you definitely don't want some bogus
controller or disk to go into a loop starting Oracle up halfway. I
have autostart, but the downsides are considered. I don't need RAC.

The take-away is, Murphy is more devious than anything you can come up
with. I was at one secure mil site where the computers were behind
huge blast doors, you had to sign in with the guard, and they still
had some electrician come in and start working on the big machine,
turning it off and on until Oracle was corrupted.

>  
> Thanks again,
>     jeff

jg
--
@home.com is bogus.
http://www.itp.net/10261-oracle-takes-the-moral-high-ground-with-sap-

[ddf]

> @home.com is bogus.http://www.itp.net/10261-oracle-takes-the-moral-high-ground-with-sap-- Hide quoted text -
>
> - Show quoted text -

Besides our 3-node production RAC we also have a 3-node standby RAC at
the ready, in a geographically separate data center, so that if,
perchance, the main data center loses power beyond the limit of the
UPS batteries we cam fail over in a reasonable amount of time to the
standby.

That's never had to happen (knock on wood) but we're prepared.

Refering to Joel's comment about the Detroit Allison generator, I
worked for a major telecom for a couple of years on contract who also
had a brand new Detroit Allison diesel generator that was tested
monthly for proper operation. Golly gee whiz, the tests ALWAYS worked
without a hitch. Unfortunately that same generator apparently knew
the difference between 'test' and 'omg we have no municipal power' and
failed to start when the power grid went dark. Needless to say the
computer systems ran as long as the UPS batteries had power, then shut
down (fortunately the UPS monitoring software performed graceful
shutdowns). There was no alarm on the Detroit Allison so it took
customer complaints to get someone to the facility and assess what had
actually happened. The company recovered before dawn the next day,
and mechanics and service personnel were clustered around the
generator determining why this beast failed to run only a week after a
successful test.

That one was attributed to Mrs. Murphy.


David Fitzjarrell

[jgar the jorrible]

On Oct 17, 3:26 pm, ddf <orat...@msn.com> wrote:

>
> That one was attributed to Mrs. Murphy.

Good one David! Prolly the old starter battery, eh?

We'll only being seeing more and stranger power infrastructure issues,
as we add more distributed and less predictable power from "green"
generation. They're even talking about getting rid of frequency
control in the US.
http://wattsupwiththat.com/2011/06/25/it-hertz-when-you-do-that-power-grid-to-stop-regulating-60-hz-frequency/

Yet another DVR died on Sunday with hours of unwatched favorite
shows. And "copyright restrictions" prevents backups or propagation
to the new DVR. Sigh.

jg
--
@home.com is bogus.

http://www.eweek.com/c/a/Enterprise-Applications/Oracles-Top-10-Java-Moves-of-2011-277356/