Re: vinum raid5: newfs throws an error
Greg 'groggy' Lehey writes: There was once an error in the stripe size calculations that meant that there were holes in the plexes. Maybe it's still there (old Vinum is not being maintained). But you should have seen that in the console messages at create time. Vinum reports the disk sizes as 17500MB (da1) and 17359MB (da2, da3). The raid5 volume and plex have a size of 33GB. This looks like the kind of scenario where that could happen. Try this: 1. First, find a better stripe size. It shouldn't be a power of 2, but it should be a multiple of 16 kB. I'd recommend 496 kB. This won't fix the problem, but it's something you should do anyway 2. Calculate the length of an exact number of stripes, and create the subdisks in that length. Try again and see what happens. 3. Use gvinum instead of vinum and try both ways. Ok, I decreased the stripe size to 496, regardless of whether it has anything to do with my problem. Next I set the subdisk length to 17359m on all disks, and things started to work ok. No more newfs errors here. Before doing this I also had a brief encounter with gvinum. There is no manpage in 5.3BETA7, so I assumed it groks the same config files as vinum. However, this did not do me any good as it simply rebooted the box. I guess gvinum works better in RELEASE. Thanks a lot for your help. Markus -- Markus Hoenicka [EMAIL PROTECTED] (Spam-protected email: replace the quadrupeds with mhoenicka) http://www.mhoenicka.de ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: vinum raid5: newfs throws an error
On Monday, 6 December 2004 at 23:44:59 +0100, Markus Hoenicka wrote: Greg 'groggy' Lehey writes: There was once an error in the stripe size calculations that meant that there were holes in the plexes. Maybe it's still there (old Vinum is not being maintained). But you should have seen that in the console messages at create time. Vinum reports the disk sizes as 17500MB (da1) and 17359MB (da2, da3). The raid5 volume and plex have a size of 33GB. This looks like the kind of scenario where that could happen. Try this: 1. First, find a better stripe size. It shouldn't be a power of 2, but it should be a multiple of 16 kB. I'd recommend 496 kB. This won't fix the problem, but it's something you should do anyway 2. Calculate the length of an exact number of stripes, and create the subdisks in that length. Try again and see what happens. 3. Use gvinum instead of vinum and try both ways. Ok, I decreased the stripe size to 496, regardless of whether it has anything to do with my problem. Next I set the subdisk length to 17359m on all disks, and things started to work ok. No more newfs errors here. OK, looks like this was the hole in the plex issue. I thought that was gone. Greg -- When replying to this message, please copy the original recipients. If you don't, I may ignore the reply or reply to the original recipients. For more information, see http://www.lemis.com/questions.html See complete headers for address and phone numbers. pgpn7akcOjP72.pgp Description: PGP signature
vinum raid5: newfs throws an error
Hi all, now that I can use the full capacity of my disks, I'm stuck again. I'm trying to set up a raid5 from three SCSI disks (I know that a serious raid5 should use five disks or more, but I have to make do with three at the moment). The configuration is as follows: drive ibma device /dev/da1s1e drive ibmb device /dev/da2s1e drive ibmc device /dev/da3s1e volume raid5 setupstate plex org raid5 512k sd length 0m drive ibma sd length 0m drive ibmb sd length 0m drive ibmc This works ok. Then I run vinum init to initialize the drives. Trying to create a filesystem on this construct results in the error message: newfs: wtfs: 65536 bytes at sector 71130688: Input/output error Is that trying to tell me that my calculation of the group size is incorrect? Does it have to do anything with the fact that the three disks have slightly different capacities? Vinum reports the disk sizes as 17500MB (da1) and 17359MB (da2, da3). The raid5 volume and plex have a size of 33GB. BTW creating a concatenated volume on the same disks works ok, newfs does not throw an error here. Any help is appreciated. Markus -- Markus Hoenicka [EMAIL PROTECTED] (Spam-protected email: replace the quadrupeds with mhoenicka) http://www.mhoenicka.de ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: vinum raid5: newfs throws an error
On Monday, 6 December 2004 at 3:05:31 +0100, Markus Hoenicka wrote: Hi all, now that I can use the full capacity of my disks, I'm stuck again. I'm trying to set up a raid5 from three SCSI disks (I know that a serious raid5 should use five disks or more, but I have to make do with three at the moment). The configuration is as follows: drive ibma device /dev/da1s1e drive ibmb device /dev/da2s1e drive ibmc device /dev/da3s1e volume raid5 setupstate plex org raid5 512k sd length 0m drive ibma sd length 0m drive ibmb sd length 0m drive ibmc This works ok. Then I run vinum init to initialize the drives. Trying to create a filesystem on this construct results in the error message: newfs: wtfs: 65536 bytes at sector 71130688: Input/output error Is that trying to tell me that my calculation of the group size is incorrect? Does it have to do anything with the fact that the three disks have slightly different capacities? There was once an error in the stripe size calculations that meant that there were holes in the plexes. Maybe it's still there (old Vinum is not being maintained). But you should have seen that in the console messages at create time. Vinum reports the disk sizes as 17500MB (da1) and 17359MB (da2, da3). The raid5 volume and plex have a size of 33GB. This looks like the kind of scenario where that could happen. Try this: 1. First, find a better stripe size. It shouldn't be a power of 2, but it should be a multiple of 16 kB. I'd recommend 496 kB. This won't fix the problem, but it's something you should do anyway 2. Calculate the length of an exact number of stripes, and create the subdisks in that length. Try again and see what happens. 3. Use gvinum instead of vinum and try both ways. Greg -- When replying to this message, please copy the original recipients. If you don't, I may ignore the reply or reply to the original recipients. For more information, see http://www.lemis.com/questions.html See complete headers for address and phone numbers. pgpRINJrCFKRb.pgp Description: PGP signature
vinum and newfs
What impact do disk block and fragment sizes have on a vinum volume? I've been benchmarking an array of three drives in striped and raid5 configurations with various stripe sizes. I've noticed that I get better results in just about every instance by passing -b 16384 -f 2048 to newfs. This doesn't make sense to me as those are the defaults for newfs if they are not specified, but looking at the disklabel after a newfs, it shows 8192/1024. Should these options really make a performance difference, and if so, how? Thanks, Mike ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: vinum and newfs
On Fri, 18 Jul 2003 10:26, Mike Maltese wrote: What impact do disk block and fragment sizes have on a vinum volume? I've been benchmarking an array of three drives in striped and raid5 configurations with various stripe sizes. I've noticed that I get better results in just about every instance by passing -b 16384 -f 2048 to newfs. This doesn't make sense to me as those are the defaults for newfs if they are not specified, but looking at the disklabel after a newfs, it shows 8192/1024. Should these options really make a performance difference, and if so, how? Thanks, Mike I have had similar experience, getting 8192/1024 when using newfs on a vinum volume. Obviously 16384/2048 is not the default in this case, in spite of the newfs man pages. In a classical file system I believe these numbers are taken from the disklabel and it is realy the disklabel that supplies these defaults for the partitions. For vinum the individual volumes do not have a corresponding disklabel partition. -- All guess work so don't take it too seriously. Malcolm ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
HELP: vinum and newfs on 5.0-RC1
Okay, I had a vinum RAID5 set working on 4.7. Since I didn't have any data on it yet and saw today's announcement of 5.0-RC1, I thought I'd give 5.0-RC1 a try on the box in question. Vinum sees the RAID5 set I created just fine, so I decided to use newfs to create a UFS2 filesystem on the volume. Call me an idiot, but I can't seem to figure out how to do this despite searching various FreeBSD mailing lists. It appears 5.0 no longer supports the -v option, so I assume vinum support is now built-in. Here's what I'm trying: # newfs -O 2 -U /dev/vinum/raid5vol newfs: /dev/vinum/raid5vol: 'e' partition is unavailable That's funny. After reading around a bit, I wondered if perhaps some sort of disk label is now required on the vinum volume. However, no matter how I try using disklabel, disklabel complains about my attempts. Here's my vinum setup: drive d0 device /dev/ad0s1f drive d1 device /dev/ad1s1f drive d2 device /dev/ad2s1f drive d3 device /dev/ad3s1f volume raid5vol plex name p5 org raid5 512s vol raid5vol sd name sd0 drive d0 plex p5 len 232798208s driveoffset 265s plexoffset 0s sd name sd1 drive d1 plex p5 len 232798208s driveoffset 265s plexoffset 512s sd name sd2 drive d2 plex p5 len 232798208s driveoffset 265s plexoffset 1024s sd name sd3 drive d3 plex p5 len 232798208s driveoffset 265s plexoffset 1536s DEVFS shows: /dev/vinum: control controld plex: p5 sd: sd0 sd1 sd2 sd3 raid5vol Disklabel /dev/vinum/raid5vol shows me: type: Vinum disk: vinum label: radi flags: bytes/sector: 512 sectors/track: 698394624 tracks/cylinder: 1 sectors/cylinder: 698394624 cylinders: 1 sectors/unit: 689394624 rpm: 14400 interleave: 1 trackskew: 0 cylinderskew: 0 headswitch: 0 # milliseconds track-to-track seek: 0 # milliseconds drivedata: 0 3 Partitions: #size offsetfstype [fsize bsize bps/cpg] a: 69839462404.2BSD 1024 8192 0 # (Cyl. 0 - 0) b: 6983946240 swap # (Cyl. 0 - 0) c: 6983946240unused0 0# (Cyl. 0 - 0) I tried editing it, setting a: to unused, size 0, removing b:, and creating e: just like a: as above. Of course disklabel complained about a zero size, so I completely removed a:, then disklabel complained about e: being out of range a-c, so I renamed e: as a: and it seemed happy. At least until I double checked the data and a subsequent disklabel call showed that all my changes had been silently discarded and the above showd up anew. The vinum label command also appears useless, happily executing but changing nothing. Now for the questions: How does one create a new filesystem (UFS2 in particular) on a vinum volume in 5.0? Is some sort of label required? Help! Thanks! Aaron out. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-questions in the body of the message
Re: HELP: vinum and newfs on 5.0-RC1
I have also been having a few problems with vinum on 5.0-RC1. I've been having a hard time making a striped or raid5 vinum with more than 4 disk's. On the 5th+ disk's vinum will see and says device is not configured. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-questions in the body of the message
Re: HELP: vinum and newfs on 5.0-RC1
On Monday, 9 December 2002 at 20:23:44 -0500, David Rhodus wrote: I have also been having a few problems with vinum on 5.0-RC1. I've been having a hard time making a striped or raid5 vinum with more than 4 disk's. On the 5th+ disk's vinum will see and says device is not configured. That's a very different issue. I'm still investigating the one to which you replied, but in your case I'd like to see the config file and the contents of /dev. Greg -- When replying to this message, please copy the original recipients. If you don't, I may ignore the reply or reply to the original recipients. For more information, see http://www.lemis.com/questions.html See complete headers for address and phone numbers To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-questions in the body of the message