Operating System - Tru64 Unix
1748165 Members
3821 Online
108758 Solutions
New Discussion юеВ

Re: Tru64 5.1 B Domain Missing

 
SOLVED
Go to solution
Karsten J. Hilton
New Member

Tru64 5.1 B Domain Missing

All,
I am currently working on an ES40 running Tru64 5.1 B. The issue I am having is that we have oracle installed and we have 4 directories. /u01 /u02 /u03 and /u04. 3 of these directories have been set up as domains such as oradata02 for the /u02 and on and on. The problem I am having is that recently /u04, or oradata04 began to disappear and then reappear. However, now in the past two weeks it has disappeared and I am unable to get it to mount. On reboots I get an error 5 panic message, and when I try to go through sysman and create the domain it tells me it can not get the partition size. Anyway, I am attaching an error to this and if anyone could help it would be greatly appreciated.

Thanks in advance,
Karsten J. Hilton
10 REPLIES 10
jim owens_1
Valued Contributor
Solution

Re: Tru64 5.1 B Domain Missing

Your attachment shows the result of an earlier domain panic. You need to post the first messages that occur when it attempts to mount the domain.

You should also examine the binary error log to see if a device error is being reported.
Hein van den Heuvel
Honored Contributor

Re: Tru64 5.1 B Domain Missing

Sure seems like the disk for the domain is down: /dev/disk/dsk7c
Considering it has come and gone you might want to re-seat it. Pop out, push in, try again.

You may want to pre-try the raw device with:
dd if=/dev/rdisk/dsk7c bs=8k count=1000 of=/dev/null

You might want or need to use the '/sbin/advfs/verify' and/or
/sbin/advfs/salvage to bring the AdvFS data to life.

But we have a communication problem here:
"directories have been set up as domains"

The above line does not compute


Domains live on one or more devices.
Only 1 domain per device

Domains contain one of more filesets.

Directories are used as mountpoint for filesets, not domains.

Issue the command:

ls -l /dev/fdmns/*

That should show you the mapping of domains and devices.

Next use showfsets on those domains.

If you continue to have problems with this, then please re-state your question with the above naming in mind.

Specifically it would be interesting to know whether u04 is the only fileset in a domain on dsk7 or not.

hth,
Hein.
Karsten J. Hilton
New Member

Re: Tru64 5.1 B Domain Missing

Sorry about the domain confusion. Perhaps this may make more sense, and I should mention that I am new to Tru64. My background is mostly Solaris. Anyway, the way we have the system setup is that we have HSZ70 system with 6 raidsets. dsk7c is set up as raidset raid4, unit D104 with four 18 gb disks. The oradata04 was setup as a domain for dsk7c which happens to be for the /u04 directory which is on the raid4 raidset. It is that raidset or group of disks that seem to not be able to work right.

Thanks,

Karsten
Hein van den Heuvel
Honored Contributor

Re: Tru64 5.1 B Domain Missing

>> Sorry about the domain confusion.
No problem. You are trying. But the domain vs fileset situation is still not clear yet.
An other source of info for this is:
# grep /etc/fstab u0

>> have HSZ70 system with 6 raidsets.

Again an communication issue.
A 'raid set' is too generic.
I suspect it is a raid 5, but you need to confirm that for yourself: Raid-0, Raid-0+1 or Raid 5?

Anywhay.. you want to 'visit' that HSZ through it serial line interface or using hszterm over the scsi bus to find out what it thinks about the devices. What is up, what is down? HSZ> SHOW UNIT ... SHOW DISK...

Maybe that dsk7 re-appeared to Tru64 as an other disk after some fat fingering?
Chech with #hwmgr scan scsi followed by hwmgr show scsi. If need be triple check using the WWID's

>> It is that raidset or group of disks that seem to not be able to work right.

Right, so ask the HSZ for details.
It should know more than Tru64 can know.

Hein.


Karsten J. Hilton
New Member

Re: Tru64 5.1 B Domain Missing

i guess the best way to put this is that /u04 is a partition on the server, and now the partition has disappeared. The device it coralates to is /dev/disk/dsk7c.

Thanks,

Karsten
Hein van den Heuvel
Honored Contributor

Re: Tru64 5.1 B Domain Missing

Well, You may not be equiped for this task.
Call support, internal or external.
re-read teh replies.
Consult the man pages.
Remember, This is why they pay you the big bucks!

Cheers,
Hein.


Vladimir Fabecic
Honored Contributor

Re: Tru64 5.1 B Domain Missing

What does HSZ say?
There must be some problem with storage.
And post output of:
# disklabel -r dsk7
Some other things:
/u04 is mount point for file system
/dev/disk/dsk7c is disk device file
oradata04 is advfs domain name

Regards
In vino veritas, in VMS cluster
Joseph P. Smith
Regular Advisor

Re: Tru64 5.1 B Domain Missing

Good Morning Karsten,

Perhaps you have already done this or been directed. From your error file:
Type of operation: Read
Error: 5 (see /usr/include/errno.h)

So,
$ grep 5 /usr/include/errno.h
#define EIO 5 /* I/O error */

You probably have a hardware error on the physical device as "presented" to the system by the controller. Whether you or someone else has configured JBOD, RAID 1 or RAID 5 device on the controller is another level of discovery. Perhaps the controller has, as someone else suggested, lost contact and the physical disk(s) is/are O.K.

The cause of the read error may allow you not to have to restore the device from backup tapes, but I would guess it's unlikely. Here's hoping you have done level 0 backups in the recent past.

Looking to recovering Oracle, consult with your DBA as the interrelationships among files may cause Oracle not to start properly with out-of-sync databases, config files, tablespaces that are stored on other devices. The entire Oracle config may have to be restored across all disks to a point in time where all of the files are coherent.

Best of luck. ,,, Joe S.
Karsten J. Hilton
New Member

Re: Tru64 5.1 B Domain Missing

To all that have helped try to troubleshoot this issue I appreciate your input, time, and help concerning this matter. The issue after continuous troubleshooting was that the partitions were all labeled wrong. I had to go in and zero out the disk labels and redo them to get the correct disk sizes to be read. After correcting this issue we have the server up and running back in order. Again, to all, thank you very much for your help.

Karsten