Re: Dump/restore mystery (off topic)

From: Triffid (triffid_at_nebula.net)
Date: 12/01/04


Date: Tue, 30 Nov 2004 22:31:14 -0500


Chris wrote:
> Triffid wrote:
>
>>Not a Solaris question, actually CX/SX - but the relevant newsgroups
>>have been dormant for 10+ years so I'll try my luck here. I think I need
>>a lateral thought, but none have come to me yet...
>>
>>I am responsible for some ancient Harris systems running CX/SX. Some
>>background here:
>>
>>http://tinyurl.com/5e3s4
>>
>>One of them refused to reboot after an unscheduled power outage last
>>weekend - hard disk failure. This should not have been a problem - it
>>happens with monotonous regularity - I just drop in a new disk, boot
>>from tape, and restore the latest backup.
>>
>>This time it didn't work. The restore script reported numerous instances
>>of "file not found on tape". The system was (barely) bootable after
>>restore, but many files and directories were missing. I tried several
>>backup tapes with the same result, leading me to conclude the tape drive
>>had been writing unreadable backups for some time without reporting any
>>errors.
>>
>>I tried restoring a backup from another system with identical hardware
>>and similar software configuration as I figured I could reconcile the
>>differences without too much difficulty. This worked, but I cannot
>>backup and restore the resulting system - I get the same "file not found
>>on tape" errors.
>>
>>I've narrowed the issue down somewhat:
>>
>>1. Backup tape from System B restores without error on System A
>>2. System A boots normally after restore, and appears to be a perfect
>>System B clone
>>3. Backup of System A immediately after restore completes without error
>>4. Attempt to restore System A from step 3 tape gives numerous "file not
>>found on tape" errors
>>5. Swap out the System A tape drive (6 times so far!) and repeat from
>>step 1 - same result!
>>
>>So it seems to me I either have 6 tape drives that boot and read just
>>fine, but can't write a usable backup tape - or something else is
>>causing the problem. 6 tape drives with the same problem seems highly
>>unlikely. The backup script must be identical (since it was restored
>>from System B), as must the restore script (it's on the boot partition
>>of the backup tape), so WTF is going on here?
>>
>>TIA
>>
>>Triffid
>
>
>
> Can you dump System A to tape while it's idle?
> Then try to restore from that tape.

It depends on your definition of 'idle'

All System A tests were run with network interfaces disconnected, the
only logon was root at the serial console. However, all file systems
were mounted and init level 2 processes were running.

> Are you dumping live file systems in the first place?

I understand dump should ideally be run against unmounted file systems,
however shutting down to single user for backups is often not an option,
so dumping live file systems at 'quiet' times is common practice. This
can cause missing file errors on restore, but IME the missing files are
*always* those which are expected to have had changes during the dump.

In this case restore reports missing file errors for numerous binaries
which never change. I've never seen that before, not even on systems
which were 'busy' during dump.

That said, I would very much like to try dumping unmounted file systems
on System A to see if the symptoms change - but the only miniroot
available does not include the fdump command so it's not possible to
dump / unmounted.

Thanks for your reply,

Triffid



Relevant Pages

  • Re: Dump/restore mystery (off topic)
    ... Triffid wrote: ... > from tape, and restore the latest backup. ... The restore script reported numerous instances ...
    (comp.unix.solaris)
  • Re: Dump/restore mystery (off topic)
    ... Triffid wrote: ... > from tape, and restore the latest backup. ... The restore script reported numerous instances ...
    (comp.unix.solaris)
  • Dump/restore mystery (off topic)
    ... from tape, and restore the latest backup. ... The restore script reported numerous instances ... So it seems to me I either have 6 tape drives that boot and read just ...
    (comp.unix.solaris)
  • Re: Dump/restore mystery (off topic)
    ... > from tape, and restore the latest backup. ... The restore script reported numerous instances ... > So it seems to me I either have 6 tape drives that boot and read just ...
    (comp.unix.solaris)
  • Re: Backup statergy - comments please
    ... > Sun from a CD, then back up all the file systems to tape, so ... of running out of space on a DDS4 tape would be small! ... I do an incremental backup (using a different ... dumpfs / /export/home/backup ...
    (comp.sys.sun.admin)