Re: mdadm, partitioned array

2007-10-12 Thread Fab
estone.ca> writes: > Running some bonnie++ benchmarks gave extra slow write results on > /dev/md_d0p1. > but the same tests on /dev/md_d0p2 are normal. Read results are approx > the same on both partitions, but only writes are half as fast on p1 > then p2. Same problem. But I wasn't as bad as y

Re: mdadm, partitioned array

2007-10-10 Thread martin f krafft
also sprach David Brodbeck <[EMAIL PROTECTED]> [2007.10.10.1731 +0100]: > Although I've seen more people (including myself) shoot themselves in the > foot with lvreduce than with any other utility. After I did it I > immediately realized it was my own fault, but I still felt kind of > like this:

Re: mdadm, partitioned array

2007-10-10 Thread David Brodbeck
On Oct 9, 2007, at 6:32 PM, martin f krafft wrote: LVM is fine. I don't see a reason not to use it. Although I've seen more people (including myself) shoot themselves in the foot with lvreduce than with any other utility. After I did it I immediately realized it was my own fault, but I st

Re: mdadm, partitioned array

2007-10-09 Thread martin f krafft
also sprach Douglas A. Tutty <[EMAIL PROTECTED]> [2007.10.10.0102 +0100]: > I didn't see any _partition_ option on Etch's installer. If you want > partitions of a raid device, it looks like you need to mark it for use > with LVM. There isn't one. Partitionable arrays are not implemented in partma

Re: mdadm, partitioned array

2007-10-09 Thread Douglas A. Tutty
On Tue, Oct 09, 2007 at 07:59:35PM +0100, martin f krafft wrote: > also sprach [EMAIL PROTECTED] <[EMAIL PROTECTED]> [2007.10.09.0611 +0100]: > also sprach Douglas A. Tutty <[EMAIL PROTECTED]> [2007.10.09.1718 +0100]: > > LVM is perfect. > > You wish. :) OK, nothing in any OS is perfect and Lin

Re: mdadm, partitioned array

2007-10-09 Thread michael
Quoting martin f krafft <[EMAIL PROTECTED]>: also sprach [EMAIL PROTECTED] <[EMAIL PROTECTED]> [2007.10.09.0611 +0100]: So, as far as I can tell, I only see the slow down on the first partition of a partitioned single raid 5 array. (Didn't test raid 1 or raid 10) Can you try a third partition

Re: mdadm, partitioned array

2007-10-09 Thread martin f krafft
also sprach [EMAIL PROTECTED] <[EMAIL PROTECTED]> [2007.10.09.0611 +0100]: > So, as far as I can tell, I only see the slow down on the first > partition of a partitioned single raid 5 array. (Didn't test raid > 1 or raid 10) Can you try a third partition. > Interesting results. Weird indeed. You

Re: mdadm, partitioned array

2007-10-09 Thread Douglas A. Tutty
On Tue, Oct 09, 2007 at 08:40:45AM -0700, [EMAIL PROTECTED] wrote: > Quoting Martin Marcher <[EMAIL PROTECTED]>: > >2007/10/9, [EMAIL PROTECTED] <[EMAIL PROTECTED]>: > >Out of interested: Why are you partitioning your RAID and not using LVM? > > Good question. > I guess mostly because I'm not too

Re: mdadm, partitioned array

2007-10-09 Thread michael
Quoting Martin Marcher <[EMAIL PROTECTED]>: Hi, 2007/10/9, [EMAIL PROTECTED] <[EMAIL PROTECTED]>: Adding some more info: Interesting results. I wonder if anyone else has a partitioned raid 5 array and have normal speed results? I do. Well sort of. I don't partition RAIDs but I put LVM on to

Re: mdadm, partitioned array

2007-10-09 Thread Martin Marcher
Hi, 2007/10/9, [EMAIL PROTECTED] <[EMAIL PROTECTED]>: > Adding some more info: > Interesting results. > I wonder if anyone else has a partitioned raid 5 array and have normal > speed results? I do. Well sort of. I don't partition RAIDs but I put LVM on top of it. I couldn't tell any difference b

Re: mdadm, partitioned array

2007-10-08 Thread michael
Adding some more info: I killed my raid 5 and created a raid 0 instead. After creating the single raid 0 array, I partitioned it into 2 parititions. Formatted and mounted both and ran the same speed tests. Both tests turned out normal. Then, I killed the raid 0 and instead created 2 raid 5

Re: mdadm, partitioned array

2007-10-08 Thread michael
Quoting [EMAIL PROTECTED]: host:~# cat /proc/mdstat Personalities : [raid1] [raid10] [raid6] [raid5] [raid4] md_d0 : active raid5 sdd3[3] sdc3[2] sdb3[1] sda3[0] 14650944 blocks level 5, 64k chunk, algorithm 2 [4/4] [] md1 : active raid10 sda2[0] sdd2[3] sdc2[2] sdb2[1] 976729

Re: mdadm, partitioned array

2007-10-08 Thread michael
Quoting "Douglas A. Tutty" <[EMAIL PROTECTED]>: Well, its raid5 which is slow to begin with. Have you verified with mdadm that the array is complete and not running in degraded mode (which would slow it down further)? Sorry, forgot to mention that I have check mdstat to make sure the arra

Re: mdadm, partitioned array

2007-10-08 Thread michael
Quoting "Douglas A. Tutty" <[EMAIL PROTECTED]>: Well, its raid5 which is slow to begin with. Have you verified with mdadm that the array is complete and not running in degraded mode (which would slow it down further)? Tell us about sd[a-f]. What are they and how are they connected to the com

Re: mdadm, partitioned array

2007-10-08 Thread Douglas A. Tutty
On Mon, Oct 08, 2007 at 04:42:51PM -0700, [EMAIL PROTECTED] wrote: > > I created raid 5 array via the following: > > # mdadm -C /dev/md_d0 -l5 -n 6 /dev/sd[a-f]3 --auto=p > > Upon reboot, I cfdisk'd the md_d0 into md_d0p1 and md_d0p2 > Formatted each partition and mounted one to /mnt and the oth

mdadm, partitioned array

2007-10-08 Thread michael
Hi, I created raid 5 array via the following: # mdadm -C /dev/md_d0 -l5 -n 6 /dev/sd[a-f]3 --auto=p Upon reboot, I cfdisk'd the md_d0 into md_d0p1 and md_d0p2 Formatted each partition and mounted one to /mnt and the other to /mnt2. Running some bonnie++ benchmarks gave extra slow write results

mdadm, partitioned array

2007-10-08 Thread michael
Hi, I created raid 5 array via the following: # mdadm -C /dev/md_d0 -l5 -n 6 /dev/sd[a-f]3 --auto=p Upon reboot, I cfdisk'd the md_d0 into md_d0p1 and md_d0p2 Formatted each partition and mounted one to /mnt and the other to /mnt2. Running some bonnie++ benchmarks gave extra slow write results