Re: Is database read activity primary sequential or random?
Date: Fri, 17 Apr 2015 08:34:13 -0600
Message-ID: <CAJzM94D+q-gRh46=MDwPVFOSYqV3u7X6rLV251QxKeAXK8c_pw_at_mail.gmail.com>
Jeremy really nailed it with the "...better not to improvise..." I had concerns back in November at the beginning of this migration project, but they were brushed aside. I wasn't an SE and didn't understand how storage worked. After all, I'm only a DBA. Funny thing, or maybe not, is that this storage was new to every member of the SE team and *they* didn't understand how it worked either. They didn't read the white papers EMC and Oracle provided to us, nor did they follow my/Technologent's recommendations. They decided they knew better and if we had problems later, we would work through them. Now there's a mad rush to get the redo, flashback, and controlfiles for all the migrated databases onto the SSD like I requested last year.
I was able to find a developer who understood basically how our three most critical applications were designed to work. Using that as a starting point, I'll do some research on current execution plans and such. I am not giving my approval to migrate these last 2 databases until we have a better understanding of the configuration needed. Fortunately, they will need my approval before it happens.
I've fought with the "you're just a DBA" attitude for several years now. It's tiring and frustrating, but not unexpected. I got some unexpected support from the VP of IT and the CTO this week, so that helped.
On Fri, Apr 17, 2015 at 6:32 AM, Jeremy Schneider < jeremy.schneider_at_ardentperf.com> wrote:
> On Fri, Apr 17, 2015 at 1:04 AM, MARK BRINSMEAD
> <mark.brinsmead_at_gmail.com> wrote:
> >
> > But let's not confuse the issue. I am pretty sure that the storage
> > administrators in this case are interested in what the database is
> trying to
> > do, not what the underlying physical disks actually are doing. They are
> > probably trying to decide what data to assign to which tier of storage.
>
> see now my assumption would have been the opposite - they are
> configuring disks, so they would be interested to know the actual I/O
> patterns will be so they can configure appropriately. in particular
> it smells to me like someone is toying with aggressive read-ahead...
> which is rarely a good idea for database volumes.
>
> >> On 04/16/2015 01:20 PM, MARK BRINSMEAD wrote:
> >>
> >> Take care in answering this question. And maybe spend some time with
> your
> >> storage administrators to make sure you (and they) understand why they
> are
> >> asking it.
>
> so yeah, i think that hits the nail on the head. :)
>
> is this the same SE team that previously ignored your and
> Technologent's specific requests about how to config the SAN for the
> database? if you can find out what specific SAN features they are
> considering - which is driving this question - then you could research
> the SAN feature yourself and possibly even ask back on the list if
> anyone else has experience with that specific SAN configuration (if
> it's appropriate to post the question publicly).
>
> also -- if you're on EMC storage then there should be a guide out
> there that tells you how to config your specific storage for an oracle
> database. EMC usually has those available (most storage vendors do).
> your SE team should just follow the guide. it's better not to
> improvise unless you've got a really good understanding of why you're
> going off the beaten path. if they have questions about how to
> interpret the guide, then have them send you a copy of whatever EMC
> doc they're using and you can help them understand the
> database-specific parts.
>
> --
> http://about.me/jeremy_schneider
>
-- Sandy GHX -- http://www.freelists.org/webpage/oracle-lReceived on Fri Apr 17 2015 - 16:34:13 CEST