Disappearing/Intermittent Index files

[StarGuppy]

In early April a "value added" solutions provider installed Networker 6.0 with
35 clients and a EXB-220LVD-1xMammoth2LVD 20-tape tape robot+drive.

Legato runs on a dual PII 333 Win NT 4.0 SP5 machine and backs up
SGI, Solaris, HP, RHLinux, WinNT. There are 4 NIC's on the machine
though we backup 6 subnets, one NIC is an antique 10Mb/s. We intend
to replace two of the NICs with a quad 100/10 NIC card soon.

We have 25 clients configured for incremental backups every day of the
week at 23:59 except Friday when we run a full backup.

The system was installed April 9th,10th and worked fine until April 18th when
we lost a client machine to a hacker.

The corrupt client was then removed from the save group for the next
backup and then !successfully! recovered and added back to the backup list.

Since then, when one examines the backup index to determine what backups are
available, there are only the last night's backup and April 18th's.

All the intermediate backup indexes disappear.

But not for all machines. The backup server itself seems to have all the
backups available in its index, as does the machine that was recovered from backup.
A few other machines seem to have a good set of index files... 6/25 total.

On April 30th we lost a disk and recovered another machine from backup. There
are now backup indexes for that machine upto April 18th and since May 1st...
(even though we recovered from the full backup on the 27th).

We started running nsrck on all machines to rebuild index files once or twice
a week. I understand this reads all the tapes and regenerates an index file,
I would guess it might add all files on tapes to a single index file and thereby
effectively erase old ones... but the new index files we make disappear after
the next backup runs...

Our "solution provider" (cough) told us to run nsrck for all clients then nsrim -X
and then nsrck -F. Sure enough nsrck rebuilt all the indexes though it took
5 hours then we did nsrim and nsrck. AND THEY ALL DISAPPEARED...sorry
for the capitals...but...

In response to several inquiries, Legato has told us to upgrade WinNT to SP6a
and upgrade Legato to 6.0.2. This only tells me they have no bloody idea what the
problem is.

I apologize for this not containing more, er, backup/ground info...but I started here
May 1st and missed the 3 days technical training that were provided to the sysops
who can't seem to solve the problem on our side.

 

[ag100]

I've never seen this problem that you're describing.. BUT.. First, know that you should still recover your data, if you had to. Run: mminfo -m on the backup server.

That should list all volumes in the media database. Each save set on the volume should be recoverbale using a save set recover, or other means.

Also, unless you specify otherwise, the client indexes get backed up with the client, so you should be able to recover the indexes using the nsrck -L 7 command if necessary.

That's all I can think of for now.. Good Luck!

 

[nicola]

Hi,
this is a known media db bug for NW 6.0.
The 6.0.1 version from Legato came after few weeks from 6.0, and resolved.

Your data is still on the tapes, but media db lost the track of it.

Upgrade to 6.01 or 6.02 and scanner -i on the tapes you have the data on will make the recover possible via Nwrecover.

Bye

Nicola

 

[ag100]

Hey -

I've seen different symptoms with this bug..

Like nicola said, if you can upgrade, do it.. If you can't, get a utility for NetWorker called mdbck.. This should fix that error too..

Good Luck!

 

[un1xh3ad]

We still have the same problem , we have a call open to Sun.
If we ever solve it , i will post the results here.

Neil

 

[StarGuppy]

It turned out that the "value-added" re-seller had explicitly told us to misconfigure our
system.

Somewhere there is a parameter that tells the system how long
you want the index files to "remember", this was set to zero at the explicit instructions of the "expert" who provided on-site training at the cost of $5000 or something. On top of this the value-added re-seller didn't notify legato we had software support from them so when I called them I had to threaten, bluster and shout and
write my member of parliament (a vague concept to the Californian at the help desk).

Needless to say when the legato support tech pointed me to the problem in
under 90 seconds I was rather upset.

 

[mkkeane]

Sorry you have had such a difficult time with support on this issue. Though it sounds like you found a problem with browse and rentention times, and this ligitimately would cause the symptoms you are experiencing, Nicola et al are correct in that 6.0 version of NetWorker was recalled due to fatal errors in the databases. This requires that you upgrade to any version other that 6.0 (6.0.2,6.0.3, 6.1, 6.1.1, 6.2). Please read the release suppliment of 6.01 to get more details on this.

Glad you found resolve.

Matthew