User Functions
Don't have an account yet? Sign up as a New User
Who's Online
Guest Users: 9
|
| View previous topic :: View next topic |
| Author |
Message |
halcyon943 JBOD

Joined: 17 Oct 2008 Posts: 1
|
Posted: Fri Oct 17, 2008 10:44 am Post subject: XSan 1.4.2 metadata lost |
|
|
Hello,
Setup:
1 MDC connected through a Brocade FC switch to XServe RAID. 4 Windows Servers connect through FC switch and StorNext client.
Yesterday, our SAN volumes lost their label randomly. Reading through the forums I was able to relabel our volumes to their previous state using cvlabel. The volumes are now labeled correctly, however the SAN volume will not start. Our SAN, called 'SAN', has two pools, Meta and Users. The Meta is a RAID 1 on the Upper Controller (drives 1&2). Users is a RAID 5 on the lower controller (drives 9-12).
Using RAID Admin, everything is green lights. However, when I go to verify the parity data on the RAID 1, it says there is no parity data available. The RAID 5 still has its parity data.
I've tried every cvfsck command I've come across to try to rebuild the journaling/metadata. I can see 'SAN' in cvfsck. Here is cvfsck output when I run without switches:
| Quote: | ** NOTE ** Read Only Check.
File system journal will not be recovered.
The results may be inconsistent and mis-leading.
journal_check error: Journal Descriptor marker corrupt
*Warning*: The file system journal is corrupt. A new journal must be built
to recover from this error. If you choose to continue cvfsck
will verify the file system. After cvfsck verifies or repairs
the file system, cvupdatefs must be run to rebuild the journal.
Continue? [(y)es/(N)o] y
Super Block information.
FS Created On : Wed Dec 31 19:00:00 1969
Inode Version : '0.0'
File System Status : *Dirty*
Allocated Inodes : 0
Free Inodes : 0
FL Blocks : 0
Next Inode Chunk : 0x0
Creating Meta allocation check file.
Creating Users allocation check file.
Stripe Group Meta ( 0) 0x3a6f700 blocks.
Stripe Group Users ( 1) 0xaf4bc00 blocks.
Building Inode Index Database 0 (100%).
Journal extent (base-0x0 end-0x0) block 0x0 doubly allocated! |
I'm concerned with the FS created on date as well as the journal_check error. Any attempts, using -j, -C, -wv, -K to rebuild the journal have been unsuccessful.
Currently, the SAN will not mount and I cannot 'activate' it using cvadmin. In addition, I am getting this error in the volume log:
| Quote: | Configuration:
DiskTypes-5
Disks-5
StripeGroups-2
ForceStripeAlignment-1
MaxConnections-75
ThreadPoolSize-128
StripeAlignSize-256
FsBlockSize-4096
BufferCacheSize-32M
InodeCacheSize-8192
RestoreJournal-Disabled
RestoreJournalDir-None
[1017 08:07:15] 0xa000ed88 (Info) Self (hawk) IP address is 10.1.70.9 .
[1017 08:07:15.012506] 0xa000ed88 (Debug) No fsports file - port range enforceme
nt disabled.
[1017 08:07:15] 0xa000ed88 (Info) Listening on TCP socket hawk:49211
[1017 08:07:15] 0xa000ed88 (Info) Node [0] [hawk:49211] File System Manager Logi
n.
[1017 08:07:15] 0xa000ed88 (**FATAL**) PANIC: /Library/Filesystems/Xsan/bin/fsm
ASSERT failed "htons(di->di_marker_st) == DINODEMARKER_ST" file fsm.c, line 880
[1017 08:07:15] 0xa000ed88 (**FATAL**) PANIC: aborting threads now. |
1) Any ideas on how I can get the SAN mounting again? (Maybe just get the RAID 5 mounted so I can get at the data?)
2) If I blow away the RAID 1 and recreate it, will I be able to rebuild the metadata or be able to restart the SAN?
Any help would be appreciated.
Thanks
EDIT: As a side question, is it possible to verify/see the individual storage pools' data? At this point, all I really care about is getting to the data on the RAID 5. |
|
| Back to top |
|
 |
mb JBOD

Joined: 19 Oct 2008 Posts: 1
|
Posted: Sun Oct 19, 2008 12:47 pm Post subject: |
|
|
Hey,
1. if you blow away your Raid-1, you will loose all data.
2. How many LUNs are there? What was your %full before the crash?
3. How do you know you re-labeled the LUNs correctly? Where did you get the old info?
4. What has Apple support said?
5. You can't get any data back before Raid-1 is up.
Cheers, MB |
|
| Back to top |
|
 |
svavaroe partially protected

Joined: 08 Aug 2006 Posts: 8
|
Posted: Thu Sep 15, 2011 12:30 pm Post subject: |
|
|
same problem here.
Any help ? or success ? |
|
| Back to top |
|
 |
abstractrude Xsan Master

Joined: 13 Mar 2008 Posts: 864
|
Posted: Thu Sep 15, 2011 2:49 pm Post subject: |
|
|
| are you sure someone didnt format the LUNS on a windows box? |
|
| Back to top |
|
 |
ravi Xsan Master

Joined: 06 Mar 2008 Posts: 149
|
Posted: Fri Sep 16, 2011 3:43 am Post subject: Re: XSan 1.4.2 metadata lost |
|
|
| halcyon943 wrote: | Hello,
The volumes are now labeled correctly |
Including the correct sector size?
R |
|
| Back to top |
|
 |
nrausch Xsan Master

Joined: 14 Sep 2007 Posts: 202
|
Posted: Fri Sep 16, 2011 8:29 am Post subject: |
|
|
| Quote: | | However, when I go to verify the parity data on the RAID 1, it says there is no parity data available. The RAID 5 still has its parity data. |
RAID 1 is a mirror without parity, so no worries about the parity data there!
RAID 1 (mirroring without parity or striping)
It sounds like someone did format the luns from a windows box.
Switch zoning is important, especially in a mixed OS environment.
Are the switches zoned to prevent the windows clients from seeing the metadata raid1? |
|
| Back to top |
|
 |
wrstuden Xsan Master

Joined: 04 Jan 2008 Posts: 99
|
Posted: Sat Sep 17, 2011 12:33 pm Post subject: |
|
|
Yeah, sounds like your Superblock is gone. Inode version 0 and a date in 1969 sound like a wipe. You should interview the windows users (with a stick?).
It might have been good to look at the volumes in Disk Utility before you fixed the labels. If it were a simple NTFS reformat, I would have expected OS X to recognize the format. So maybe you need to chat w any Linux users… |
|
| Back to top |
|
 |
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|
|