From mboxrd@z Thu Jan 1 00:00:00 1970 To: 9fans@9fans.net Subject: Re: [9fans] Mirrorarenas in p9p From: "Russ Cox" Date: Thu, 31 Jul 2008 12:06:52 -0400 In-Reply-To: <073120081453.4828.4891D1D2000C4C1E000012DC22243429029B0A02D2089B9A019C04040A0DBF9B9D0E9A9B9C040D@att.net> MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit Message-Id: <20080731160501.EA1CF1E8C3B@holo.morphisms.net> Topicbox-Message-UUID: f88e7608-ead3-11e9-9d60-3106f5b1d025 > I'm running a p9p venti server under Linux using several > (small by current standards) physical drives as my arena > pool. I got a 500GB external USB drive to use as a backup > of the arenas. But here's the gotcha. The number of blocks > per cylinder on the 500GB drive isn't the same as the other > smaller drives. So, I can't create a partition that's exactly > the same size as one of the smaller drives, which means that > the last arena is a different size when I format it. mirrorarenas > complains about that and considers it fatal. I modified > it so that if the destination arena is bigger than the source, > it's just a warning and not fatal and it's happily mirroring > as we speak. But I've got a couple of questions: > - Is this going to cause any problems? > - Is there a better way to handle this? It would be best to set up an arena set that matches the one you are mirroring exactly. To cope with the lack of real partitioning support in operating systems that are not Plan 9, the p9p venti lets you specify disk subpieces in the partition names. So you can say /dev/hda:0-10G to get the first ten gigabytes of /dev/hda. Or /dev/hda:0-12345M or /dev/hda:1G-123123123123M etc. Also, if you do create a Plan 9 partition and then prep it to make subpartitions (you'd have to use 9vx, copying the binaries from sources), then p9p venti can handle those partition names too: /dev/sda1:arenas assuming /dev/sda1 is a Plan 9 fdisk partition containing a subpartition named arenas. Yes, it's terrible. You use Unix long enough, you get used to that. Russ