RE: ARCn
Date: Thu, 1 Aug 2013 16:28:54 -0400
Message-ID: <00c501ce8ef5$beb04d80$3c10e880$_at_rsiz.com>
/religious war warning
I also don't know exactly what the code does. One way to finesse this with *some* volume managers is to use external redundancy to have sufficient plexes (aka images, mirrors) in a RAID sense on the storage that you can dispense with having multiple members (that's the religious war part). Some volume managers allow you to specify your read preference so you get what you're asking for from the underlying media instead of from Oracle's code.
/warning off.
Without extensive testing (which I'm unlikely to do) or the code in hand (which is unlikely to happen in lieu of a bonkers offer that recruits me to Oracle) I would not argue with Hans' guess.
If anyone really knows, please chime in.
footnote: Jared has some really good (meaning only funny after the resolution) war stories about someone less than competent deleting one member of a redo group. My experience is that when someone is that idiotic they usually delete all the members. That is the nature of the "religious" argument. I can see both sides of the argument: Sometimes the wrong person gets the keys to the kingdom.
mwf
-----Original Message-----
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org]
On Behalf Of Ricard Martínez
Sent: Thursday, August 01, 2013 3:24 PM
To: fuzzy.graybeard_at_gmail.com
Cc: oracle-l_at_freelists.org
Subject: Re: ARCn
Hi
We have 2 members in 2 different disk, and one is faster than the other, so
im wondering, if it follows a rule (e.g: always one), i can use fast disk
for member 1. The other reason is curiosity :) I know it wont be a big
difference really, but i just start thinking/reading about it and i realise
i didnt know the answer for sure.
You are completely right about the *one* now that im reading it again, and im agree that the logical way of programing is starting with group1 member 1 and go on, but maybe there is more info/docs about it that i didnt know, and that can confirm the way it does it.
Thanks
On Thu, Aug 1, 2013 at 7:49 PM, Hans Forbrich <fuzzy.graybeard_at_gmail.com>wrote:
> On 01/08/2013 12:12 PM, Ricard Martínez wrote:
> > Hi
> > I have a doubt about the archive process. One redo log group have at
> least
> > 2 members. When the ARCn process reads it to create the archive log,
> > did
> it
> > always reads member 1? or its random?
> Curious why this is important to you. (There are legit reasons, just
> wondering.)
> > Cause here seems that is always member 1:
> >
> http://docs.oracle.com/cd/E11882_01/server.112/e25789/physical.htm#i10
> 06259 That doc says " the database is in ARCHIVELOG mode, and if
> automatic archiving is enabled, then the archiver process (ARCn) will
> archive
> *one* of these files. If A_LOG1 is corrupted, then the process can
> archive B_LOG1."
>
> The way I read that comment, "One of these files" does not imply "File 1"
>
> Then again, unless there is an I/O issue, why would it matter which
> one it reads? I personally would program this to say "start by
> reading member one, continue with member one of the next group unless
> we end up switching due to corruption, at which time we switch to
> member 2 and stick with that series unless we are forced to switch again."
> --
> http://www.freelists.org/webpage/oracle-l
> > >
-- http://www.freelists.org/webpage/oracle-l -- http://www.freelists.org/webpage/oracle-lReceived on Thu Aug 01 2013 - 22:28:54 CEST