Re: 10g RAC without vendor clusterware
- From: Koen Van Langenhove <Koen.Van_Langenhove@xxxxxxxxxxx>
- To: Alexander Gorbachev <gorbyx@xxxxxxxxx>
- Date: Fri, 04 Feb 2005 11:45:41 +0100
Hi Chris,
I see what you mean and I agree, but unless you're running solely on a
HW RAID system, you will eventually have to patch your volume manager
too, which often requires a reboot since quite a few of them are
implemented as kernel modules. In our environment we have always
experienced that clustered systems have relatively low uptime scores
because a) it's too hard and complex for service people to learn and
manage cluster, vm and database and b) you have to patch/upgrade 3
products, leading to large dowtimes. So it looks to me like there's a
lot of room for improvement to that by integrating the 3 of them and
having only one supplier to rely on for support. I can't wait for Oracle
11 with the integrated OS :-)
Further, since patches are applied to the software, (not the database)
you do have to restart the other instances anyway. I'm not convinced
that the extra restart, maybe 2 to 3 minutes, is going to be the big
difference during a scheduled downtime.
All in all I think most of us can think of the inconveniences and
drawbacks, what I would like to know is how stable the new features are
running with the current patchlevel and if it's at all possible to live
with such a system in production, for whatever reason.
Thanks.
Alexander Gorbachev wrote:
>Hi Koen,
>
>See below...
>
>On Thu, 03 Feb 2005 12:09:45 +0100, Koen Van Langenhove
><Koen.Van_Langenhove@xxxxxxxxxxx> wrote:
>
>
>>Hi Alex,
>>looks like you're a little pessimistic about the ASM instance. Would
>>you care telling us why ? It's just another instance, why would be it be
>>less reliable than the 'real' instances ?
>>
>>
>Exactly. It's probably not less reliable but I don't see why it would
>be MORE reliable and that what I would demand. My perception is that
>clusterware is normally more reliable than Oracle database.
>
>
>
>>BTW it's certainly not the definitive test, but I just killed the ASM
>>instance of the second node, and CRS managed to restart ASM and the
>>other instances nicely.
>>
>>
>The key word is RESTARTED.
>Imagine that you need to install a new patchset for your ASM instance
>or one-off patch. All you DBs will have to be down during that time.
>we can't afford it.
>
>
>
--
Regards,
Koen
Siemens ICM/ICN
IC MN D IS
Phone : +32 14 25 3000
Email : Koen.Van_Langenhove@xxxxxxxxxxx
- - - - - - -
Unix IS user-friendly .., it's just quite picky about its friends.
--
//www.freelists.org/webpage/oracle-l
Other related posts:
