Date: Mon, 27 Dec 1999 12:31:24 +1030 From: Greg Lehey <grog@lemis.com> To: Gary Palmer <gjp@in-addr.com> Cc: current@freebsd.org Subject: Re: ``how to'' for Vinum and mirroring after the volume was created? Message-ID: <19991227123124.U1316@freebie.lemis.com> In-Reply-To: <41777.946259657@noop.colo.erols.net> References: <19991227120444.S1316@freebie.lemis.com> <41777.946259657@noop.colo.erols.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sunday, 26 December 1999 at 20:54:17 -0500, Gary Palmer wrote:
> Greg Lehey wrote in message ID
> <19991227120444.S1316@freebie.lemis.com>:
>> Did you forget the 'setupstate' keyword?
>
> How would that apply? The 2nd disk was not consistant as it was added
> after the data was copied onto the first and the old disks removed.
Ah. I misunderstood. Yes, in this case you have to revive.
>> [reading from the plex being revived]
>> How do you determine that?
>
> Looking at the stats vinum provided. Unfortunately they're gone from
> my scrollback buffer, but the 2nd plex was being read from, which is
> not something I'd expect from what should have been a straight copy
> from the first to the second. Also, if I remember right, the read
> data from the first plex was going up more slowly than the written
> data to the 2nd plex, which indicates to me that it may indeed have
> been round-robining the reads, even though the 2nd plex wasn't
> consistant.
It occurs to me that it is, in fact, correct to read from a volume
being revived. The data is consistent up to the point where the
revive has progressed, so the system can read from this area. Check
the function checksdstate() in vinumstate.c:
if (diskaddr > (sd->revived
+ sd->plexoffset
+ (sd->revive_blocksize >> DEV_BSHIFT))) /* we're beyond the end */
return REQUEST_DOWN;
else if (diskend > (sd->revived + sd->plexoffset)) { /* we finish beyond the end */
if (writeop) {
rq->flags |= XFR_REVIVECONFLICT; /* note a potential conflict */
rq->sdno = sd->sdno; /* and which sd last caused it */
} else
return REQUEST_DOWN;
}
>>> The second question concerns the flag fields. Is there any way to
>>> change flags on the fly? e.g. change the volume from being round
>>> robin to prefer plex?
>
>> That's one thing I haven't done yet. It's trivial, but I haven't
>> thought of a syntax for the command. Suggestions?
>
> I think (at least to my mind) the best way to abstract it would be as
> readflags/setflags commands, and do it so that it would work for flags
> on the volume, plex, subdisk or disk level. I don't know how many
> flags you really want to expose to the end-user, but I can imagine as
> the product develops, more flags will be added. e.g. veritas has
> user-settable comments fields.
>
> e.g.
>
> vinum setflag vol01 readpol=prefer preferplex=plex01
>
> (that is a tricky one, as you really are setting two flags I would
> imagine, not just one)
Hmm. This looks too VERITAS-like for my liking. In addition, this
isn't a flag. Internally it's represented as a plex index, with -1
meaning "round robin".
I've tried to keep away from this sort of stuff. I was thinking of
something more like:
vinum prefer volfoo plexbar
or
vinum prefer volfoo -1
The second example is supposed to set round robin. I don't like that
syntax either. Maybe the second example should be just
vinum prefer volfoo.
Greg
--
Finger grog@lemis.com for PGP public key
See complete headers for address and phone numbers
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19991227123124.U1316>