Client index empty

[denisfr]

Hello,

I've got a linux client in 733 which have an index size of 0 byte. I had to restore a file for this client, so I found that this index was 0 byte since several days.

Even a new full save cannot populate the index.

I've tried nsrck -L6, and nsrck -m, but I still got 0 byte.

I took a save (where index was 20 MB) and performed a nsrck -L7 -t to find these datas, but no way : after the restore NW performs a nsrck, and still 0 byte.

I've deleted/recreated the client (with deletion of /nsr/index/my_client) and performed a full save, but my index is still 0 byte. Savesets are all valids (so I can restore via saveset in case of emergency)

Any idea ?

 

[605]

If this is not a bug, which i doubt so far, then the info must fill another CFI database.

Run "nsrls" . This would run a summary of all your CFIs.
Then run a backup of a known number of files.
Finally re-check with nsrls where the number of files changes. This could give you a clue.

 

[denisfr]

Except NW, no changes for any client...
And of course, CFI still 0 byte...

I ran the save with debug flags (-vvvvv) and it confirms to me that the CFI is right (client id in the generated save job is correct):

....
04/22/08 09:48:12 savegrp: 152101l570:/home succeeded.
04/22/08 09:48:12 savegrp: 152101l570:index started
04/22/08 09:48:12 savegrp: build_ss_job: save -s san152101 -S -g GR152101=9999=S=15J -LL -f - -m san152101 -V -l full -LL -W 78 -N index:4ca679bb-00000004-4742dc89-4742dc88-10390000-370a961e /nsr/index/152101l570
* 152101l570:index /nsr/index/152101l570/db6/
* 152101l570:index /nsr/index/152101l570/
* 152101l570:index
san152101: index:152101l570 level=full, 1 KB 00:00:01 2 files
04/22/08 09:48:14 savegrp: 152101l570:index succeeded.
04/22/08 09:48:15 savegrp: nsrim run recently, skipping
...

Date and time are marked today under the .../index directory :

root@san152101 /nsr/index/152101l570 # ll
40 total
drwx--S--- 3 root sys 512 17 avr 11:23 ./
drwxr-sr-x 130 root sys 3072 17 avr 11:23 ../
drwx--S--- 2 root sys 512 22 avr 09:45 db6/
-rw-r--r-- 1 root sys 13 17 avr 11:23 .nsr
-rw-r--r-- 1 root sys 980 17 avr 11:23 README
-rw------- 1 root sys 0 17 avr 11:23 v6ck.lck

root@san152101 /nsr/index/152101l570/db6 # ll
24 total
drwx--S--- 2 root sys 512 22 avr 09:45 ./
drwx--S--- 3 root sys 512 17 avr 11:23 ../
-rw------- 1 root sys 48 22 avr 09:45 v6hdr
-rw------- 1 root sys 0 17 avr 11:23 v6hdr.lck
-rw------- 1 root sys 0 22 avr 09:45 v6journal

I think that it can drive me to a Legato call...

 

[605]

Is this "152101l570" really your client name and is "GR152101=9999=S=15J" your group ? - Then you got more than one problem ;-) - Just kidding ....

So the CFI structure is there. I am thinking now that you might write to a pool where the option "Store index entries" is unchecked. But this explanation would be too easy, correct?

 

[denisfr]

Too easy, as you said, I've controlled this first (even if I knew it was pool-based, I've searched in diagnostic mode if it was not a new client-based 'feature').

I've performed tests with two pools and for one, it saves data for almost 60 clients (among them 3 other linux in NW 733) without such problems...

The test of this morning was made on another pool than usually (but with store index checked).

For the names : I've 180-200 hosts (windows, solaris, linux, aix...) so I must have a 'rigourous codification' in naming my hosts and my NW objects to perform correct data management

 

[605]

Well, i am almost clueless. What i would suggest is that you delete the CFI directories and restart NW to recognize that to let him create new ones automatically. But that is my last suggestion.

 

[denisfr]

No way ! Still 0 byte...