[prev in list] [next in list] [prev in thread] [next in thread]
List: evms-devel
Subject: RE: [Evms-devel] [Data Corrupted Bug?] When degraded RAID1
From: <alvin.tseng () zyxel ! com ! tw>
Date: 2005-05-31 13:44:06
Message-ID: 591CFFAEAA8FB741B0467284FA582F374DE6C3 () zytwbe01 ! ZyXEL ! com
[Download RAW message or body]
Hi Kevin,
I've tried your suggested way, that is perfectly worked. Right now I can replace the \
original disks of my raid 1 with lager disks by EVMSN. (Just like you said, partition \
-> selected the created segment to repair -> quit, then it work.)
Is it possible the bug come from evms cli?
Regards,
Alvin
-----Original Message-----
From: Kevin Corry [mailto:kevcorry@us.ibm.com]
Sent: 2005/5/28 [星期六] 下午 09:37
To: evms-devel@lists.sourceforge.net
Cc: Alvin Tseng -曾一哲
Subject: Re: [Evms-devel] [Data Corrupted Bug?] When degraded RAID1
Hi Alvin,
On Fri May 27 2005 6:28 am, Alvin Tseng wrote:
> When I tried to repair the RAID 1 device, I found EVMS will report my md
> device are corrupted at such situation.
>
> Main Object: Replace RAID 1 members from older and smaller disks (2 Maxtor
> 120G SATA) to newer and larger disks (2 Hitachi 160G SATA)
>
> Initial environment: md/md0 - [sdc5, sdd5], md/md1 - [sdc6, sdd6], md/md2 -
> [sdc7, sdd7], md/md3 - [sdc8, sdd8] sdc, sdd: Maxtor 120G SATA
> sdc5 ~ sdc7, sdd5 ~ sdd7: 32MB size partition
> sdc8, sdd8: 114.4GB partition
> All created by evms 2.3.4 on a patched 2.4.27 kernel (dm patched).
>
> Test 1:
> 1. replace old sdd (120G) with new sdd (160G), then boot.
> 2. partition new sdd into sdd5, sdd6, sdd7 and sdd8 (all same size as old
> sdd)
Are you using EVMS to do this partitioning?
> 3. repair md0 with sdd5, md1 with sdd6, md2 with sdd7 and md3 with sdd8,
> done. After resync finished, shut it down.
> 4. replace old sdc (120G) with new sdc (160G), then boot.
> 5. enter evmsn, it tell the md0, md1, md2 and md3 are degraded, but all
> containers, regions and volumes are there top of the new sdd (160G).
> 6. partition new sdc into sdc5, sdc6, sdc7 and sdc8 (all same size as old
> sdc)
> 7. enter evmsn, it didn't show any md stuffs, all md data and objects above
> md are missing.
Again, are you using EVMS to do the partitioning? I'm guessing not. You said
you rebooted with the new sdc and EVMS initially saw all the MD regions. Then
you said you partitioned sdc and entered evmsn again, which is when all the
MD regions seemed to disappear.
When you first replace sdc and evmsn shows you the degraded MD regions, simply
stay in evmsn to do the partitioning for sdc. Choose "Actions -> Add ->
Segment Manager" and assign the DOS Segment Manager to sdc. Then in the
"Segments" panel you can select the freespace segment, press enter, choose
"Create Segment" from the pop-up menu, and select the desired size and
options. When you've created all the necessary segments, you can then go to
the "Regions" panel and repair each MD region with the appropriate segment.
Save the changes and check /proc/mdstat to see if the MD regions are syncing.
> (In evmsn page "1.available objects", it has sdb, sdd, sdc5, sdc6, sdc7,
> sdd8)
> (In evmsn page "5.disk segments", it has sdc_mbr, sdc_ebr0,
> sdc_freespace1, sdc5, sdc_ebr1, sdc6, sdc_ebr2, sdc7, sdc_ebr3, sdc8,
> sdc_freespace2)
> (no more sdd segments information!!)
> (In evmsn page "6.logical disks", it has sdb, sdc sdd)
> (In evmsn page "7. Plug-ins", bulah bulah bulah...)
> (no more other pages)
>
> It's so weird, all I do is partition my new disk, but the another disk's
> evms objects disappear?
>
> OK, I tried 2nd test.
> Test 2:
> 1. replace old sdc (120G) with new sdc (160G), then boot.
> 2. enter evmsn, it tell the md0, md1, md2 and md3 are degraded, but all
> containers, regions and volumes are there top of the old sdd (120G).
> 3. partition new sdc into sdc5, sdc6, sdc7 and sdc8 (all same size as old
> sdc)
> 4. Still failed to repair, evmsn told me no more md objects on this system.
> The reports of evmsn are same as 7 of Test 1.
>
> I don't have idea what's wrong with that. I have kept the logs, but they
> are too large (more than 500KB each), if you want the log (by debug level),
> mail me, and I will mail it back to you.
Yes, please send me any/all of the logs (gzipped please, and only to me, not
the list), especially the ones where EVMS no longer finds the MD regions.
Also let me know if my suggestion above works.
--
Kevin Corry
kevcorry@us.ibm.com
http://www.ibm.com/linux/
http://evms.sourceforge.net/
[Attachment #3 (text/html)]
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=big5">
<META NAME="Generator" CONTENT="MS Exchange Server version 6.5.7226.0">
<TITLE>RE: [Evms-devel] [Data Corrupted Bug?] When degraded RAID1</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->
<P><FONT SIZE=2>Hi Kevin,<BR>
<BR>
I've tried your suggested way, that is perfectly worked. Right now I can replace the \
original disks of my raid 1 with lager disks by EVMSN. (Just like you said, partition \
-> selected the created segment to repair -> quit, then it work.)<BR> <BR>
Is it possible the bug come from evms cli?<BR>
<BR>
Regards,<BR>
<BR>
Alvin<BR>
<BR>
<BR>
-----Original Message-----<BR>
From: Kevin Corry [<A \
HREF="mailto:kevcorry@us.ibm.com">mailto:kevcorry@us.ibm.com</A>]<BR>
Sent: 2005/5/28 [星期六] 下午 09:37<BR>
To: evms-devel@lists.sourceforge.net<BR>
Cc: Alvin Tseng -曾一哲<BR>
Subject: Re: [Evms-devel] [Data Corrupted Bug?] When degraded RAID1<BR>
<BR>
Hi Alvin,<BR>
<BR>
On Fri May 27 2005 6:28 am, Alvin Tseng wrote:<BR>
> When I tried to repair the RAID 1 device, I found EVMS will report my md<BR>
> device are corrupted at such situation.<BR>
><BR>
> Main Object: Replace RAID 1 members from older and smaller disks (2 Maxtor<BR>
> 120G SATA) to newer and larger disks (2 Hitachi 160G SATA)<BR>
><BR>
> Initial environment: md/md0 - [sdc5, sdd5], md/md1 - [sdc6, sdd6], md/md2 -<BR>
> [sdc7, sdd7], md/md3 - [sdc8, sdd8] sdc, sdd: Maxtor 120G SATA<BR>
> sdc5 ~ sdc7, sdd5 ~ sdd7: 32MB size partition<BR>
> sdc8, sdd8: 114.4GB partition<BR>
> All created by evms 2.3.4 on a patched 2.4.27 kernel (dm patched).<BR>
><BR>
> Test 1:<BR>
> 1. replace old sdd (120G) with new sdd (160G), then boot.<BR>
> 2. partition new sdd into sdd5, sdd6, sdd7 and sdd8 (all same size as old<BR>
> sdd)<BR>
<BR>
Are you using EVMS to do this partitioning?<BR>
<BR>
> 3. repair md0 with sdd5, md1 with sdd6, md2 with sdd7 and md3 with sdd8,<BR>
> done. After resync finished, shut it down.<BR>
> 4. replace old sdc (120G) with new sdc (160G), then boot.<BR>
> 5. enter evmsn, it tell the md0, md1, md2 and md3 are degraded, but all<BR>
> containers, regions and volumes are there top of the new sdd (160G).<BR>
> 6. partition new sdc into sdc5, sdc6, sdc7 and sdc8 (all same size as old<BR>
> sdc)<BR>
> 7. enter evmsn, it didn't show any md stuffs, all md data and objects above<BR>
> md are missing.<BR>
<BR>
Again, are you using EVMS to do the partitioning? I'm guessing not. You said<BR>
you rebooted with the new sdc and EVMS initially saw all the MD regions. Then<BR>
you said you partitioned sdc and entered evmsn again, which is when all the<BR>
MD regions seemed to disappear.<BR>
<BR>
When you first replace sdc and evmsn shows you the degraded MD regions, simply<BR>
stay in evmsn to do the partitioning for sdc. Choose "Actions -> Add \
-><BR> Segment Manager" and assign the DOS Segment Manager to sdc. Then in \
the<BR> "Segments" panel you can select the freespace segment, press enter, \
choose<BR> "Create Segment" from the pop-up menu, and select the desired \
size and<BR> options. When you've created all the necessary segments, you can then go \
to<BR> the "Regions" panel and repair each MD region with the appropriate \
segment.<BR> Save the changes and check /proc/mdstat to see if the MD regions are \
syncing.<BR> <BR>
> (In evmsn page "1.available objects", it has sdb, sdd, \
sdc5, sdc6, sdc7,<BR> > sdd8)<BR>
> (In evmsn page "5.disk segments", it has sdc_mbr, \
sdc_ebr0,<BR> > sdc_freespace1, sdc5, sdc_ebr1, sdc6, sdc_ebr2, \
sdc7, sdc_ebr3, sdc8,<BR> > sdc_freespace2)<BR>
> (no more sdd segments information!!)<BR>
> (In evmsn page "6.logical disks", it has sdb, sdc sdd)<BR>
> (In evmsn page "7. Plug-ins", bulah bulah bulah...)<BR>
> (no more other pages)<BR>
><BR>
> It's so weird, all I do is partition my new disk, but the another disk's<BR>
> evms objects disappear?<BR>
><BR>
> OK, I tried 2nd test.<BR>
> Test 2:<BR>
> 1. replace old sdc (120G) with new sdc (160G), then boot.<BR>
> 2. enter evmsn, it tell the md0, md1, md2 and md3 are degraded, but all<BR>
> containers, regions and volumes are there top of the old sdd (120G).<BR>
> 3. partition new sdc into sdc5, sdc6, sdc7 and sdc8 (all same size as old<BR>
> sdc)<BR>
> 4. Still failed to repair, evmsn told me no more md objects on this system.<BR>
> The reports of evmsn are same as 7 of Test 1.<BR>
><BR>
> I don't have idea what's wrong with that. I have kept the logs, but they<BR>
> are too large (more than 500KB each), if you want the log (by debug level),<BR>
> mail me, and I will mail it back to you.<BR>
<BR>
Yes, please send me any/all of the logs (gzipped please, and only to me, not<BR>
the list), especially the ones where EVMS no longer finds the MD regions.<BR>
Also let me know if my suggestion above works.<BR>
<BR>
--<BR>
Kevin Corry<BR>
kevcorry@us.ibm.com<BR>
<A HREF="http://www.ibm.com/linux/">http://www.ibm.com/linux/</A><BR>
<A HREF="http://evms.sourceforge.net/">http://evms.sourceforge.net/</A><BR>
<BR>
</FONT>
</P>
</BODY>
</HTML>
-------------------------------------------------------
This SF.Net email is sponsored by Yahoo.
Introducing Yahoo! Search Developer Network - Create apps using Yahoo!
Search APIs Find out how you can build Yahoo! directly into your own
Applications - visit http://developer.yahoo.net/?fr=offad-ysdn-ostg-q22005
_______________________________________________
Evms-devel mailing list
Evms-devel@lists.sourceforge.net
To subscribe/unsubscribe, please visit:
https://lists.sourceforge.net/lists/listinfo/evms-devel
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic