Netapp dumps using Amanda

2002-07-10 Thread Ashwin Kotian

Hello there,

I'm trying to backup our Netapp filer using Amanda. I'm using the
modified dump program for Netapp obtained from the following article:

http://www.xs4all.nl/~johnpc/amanda-netapp.html

but still everytime the amanda script runs it is unable to write to tape 
& the AMANDA report gives the following summary:

FAILURE AND STRANGE DUMP SUMMARY:  tapeserver /dev/netapp/comstock lev 0 
FAILED [no backup size line]  tapeserver /dev/netapp/etc lev 0 FAILED 
[no backup size line]

The disklist has the following entry for the netapp:

tapeserver /dev/netapp/etcnetapp
tapeserver /dev/netapp/home   netapp

where "netapp" is a dumptype defined as follows:

define dumptype netapp {
   globalcomment "no compress, priority high"
   compress none
   priority high
   record no
   index }

I can see that Amanda is being able to execute the dump program on the 
filer ( I can see dump files being temporarily written to the holding 
directory ) but for some reason it is not being able to write to tape. 
However there is no tape error. And when I look at the source code in 
sendbackup.c, the error "no backup size line" is a result of condition 
dump_size == -1

What could be the cause of this error ? Is it bcoz Amanda is not being 
able to match certain patterns from the output of the Netapp dump? 
Please do let me know with whatever feedback or suggestions you might have.

Thanks,

Ashwin.




RE: Netapp dumps using Amanda

2002-07-11 Thread Poul Petersen

If you change that dumptype to "index no" does the backup work? I
can't remember if what you are seeing is the exact failure we had, but the
cause was that with indexing turned on, the restore command was closing it's
input and causing the whole pipe to fail. That is, sendbackup executes
something like:

dump {options}f - /{netapp}/{path} | tee {dumpfile} | restore -tvf - 2>&1 |
sed {stuff} > {index file}

Which because of the wrapper around dump really looks like:

rsh {netapp} dump {options}f - {path} | tee...

Since dump sends the file list before the files, the restore process
(with "t") is done long before dump if the filesystem is sufficiently large.
For some reason, when the restore command blocks input, the filer dump
command aborts and causes the dump to fail. If this is the problem you are
having, you should be able to duplicate a failure simply by doing something
like:

rsh {netapp} dump 0f - {path} | restore -tvf - 2>&1 > /dev/null

We fixed it by writing a wrapper around restore that detects when
the restore command downstreams blocks and then continues to accept input
(the actual dump contents) but simply throws it away. 

-poul

> -Original Message-
> From: Ashwin Kotian [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, July 10, 2002 5:27 PM
> To: [EMAIL PROTECTED]
> Subject: Netapp dumps using Amanda
> 
> 
> Hello there,
> 
> I'm trying to backup our Netapp filer using Amanda. I'm using the
> modified dump program for Netapp obtained from the following article:
> 
> http://www.xs4all.nl/~johnpc/amanda-netapp.html
> 
> but still everytime the amanda script runs it is unable to 
> write to tape 
> & the AMANDA report gives the following summary:
> 
> FAILURE AND STRANGE DUMP SUMMARY:  tapeserver 
> /dev/netapp/comstock lev 0 
> FAILED [no backup size line]  tapeserver /dev/netapp/etc lev 0 FAILED 
> [no backup size line]
> 
> The disklist has the following entry for the netapp:
> 
> tapeserver /dev/netapp/etcnetapp
> tapeserver /dev/netapp/home   netapp
> 
> where "netapp" is a dumptype defined as follows:
> 
> define dumptype netapp {
>globalcomment "no compress, priority high"
>compress none
>priority high
>record no
>index }
> 
> I can see that Amanda is being able to execute the dump 
> program on the 
> filer ( I can see dump files being temporarily written to the holding 
> directory ) but for some reason it is not being able to write 
> to tape. 
> However there is no tape error. And when I look at the source code in 
> sendbackup.c, the error "no backup size line" is a result of 
> condition 
> dump_size == -1
> 
> What could be the cause of this error ? Is it bcoz Amanda is 
> not being 
> able to match certain patterns from the output of the Netapp dump? 
> Please do let me know with whatever feedback or suggestions 
> you might have.
> 
> Thanks,
> 
> Ashwin.
> 



Re: Netapp dumps using Amanda

2002-07-11 Thread Ashwin Kotian

Hi Paul,

I think the reason we disabled indexing was because with indexing 
enabled, amanda was trying to use the "u" option & try to update the 
dumpdates file & was failing therein. Amanda wouldn't even do a dump & 
it complained with it a "cannot update dumpdates file" error.

Regards,

Ashwin.

Poul Petersen wrote:

>   If you change that dumptype to "index no" does the backup work? I
>can't remember if what you are seeing is the exact failure we had, but the
>cause was that with indexing turned on, the restore command was closing it's
>input and causing the whole pipe to fail. That is, sendbackup executes
>something like:
>
>dump {options}f - /{netapp}/{path} | tee {dumpfile} | restore -tvf - 2>&1 |
>sed {stuff} > {index file}
>
>   Which because of the wrapper around dump really looks like:
>
>rsh {netapp} dump {options}f - {path} | tee...
>
>   Since dump sends the file list before the files, the restore process
>(with "t") is done long before dump if the filesystem is sufficiently large.
>For some reason, when the restore command blocks input, the filer dump
>command aborts and causes the dump to fail. If this is the problem you are
>having, you should be able to duplicate a failure simply by doing something
>like:
>
>   rsh {netapp} dump 0f - {path} | restore -tvf - 2>&1 > /dev/null
>
>   We fixed it by writing a wrapper around restore that detects when
>the restore command downstreams blocks and then continues to accept input
>(the actual dump contents) but simply throws it away. 
>
>-poul
>
>>-Original Message-
>>From: Ashwin Kotian [mailto:[EMAIL PROTECTED]]
>>Sent: Wednesday, July 10, 2002 5:27 PM
>>To: [EMAIL PROTECTED]
>>Subject: Netapp dumps using Amanda
>>
>>
>>Hello there,
>>
>>I'm trying to backup our Netapp filer using Amanda. I'm using the
>>modified dump program for Netapp obtained from the following article:
>>
>>http://www.xs4all.nl/~johnpc/amanda-netapp.html
>>
>>but still everytime the amanda script runs it is unable to 
>>write to tape 
>>& the AMANDA report gives the following summary:
>>
>>FAILURE AND STRANGE DUMP SUMMARY:  tapeserver 
>>/dev/netapp/comstock lev 0 
>>FAILED [no backup size line]  tapeserver /dev/netapp/etc lev 0 FAILED 
>>[no backup size line]
>>
>>The disklist has the following entry for the netapp:
>>
>>tapeserver /dev/netapp/etcnetapp
>>tapeserver /dev/netapp/home   netapp
>>
>>where "netapp" is a dumptype defined as follows:
>>
>>define dumptype netapp {
>>   globalcomment "no compress, priority high"
>>   compress none
>>   priority high
>>   record no
>>   index }
>>
>>I can see that Amanda is being able to execute the dump 
>>program on the 
>>filer ( I can see dump files being temporarily written to the holding 
>>directory ) but for some reason it is not being able to write 
>>to tape. 
>>However there is no tape error. And when I look at the source code in 
>>sendbackup.c, the error "no backup size line" is a result of 
>>condition 
>>dump_size == -1
>>
>>What could be the cause of this error ? Is it bcoz Amanda is 
>>not being 
>>able to match certain patterns from the output of the Netapp dump? 
>>Please do let me know with whatever feedback or suggestions 
>>you might have.
>>
>>Thanks,
>>
>>Ashwin.
>>


-- 
Ashwin Kotian
Systems Administrator
Hitachi Storage Software Inc.
3031 Tisch Way Suite# 200
San Jose, CA-95128
Tel.No: 408-553-6693






Re: Netapp dumps using Amanda

2002-07-11 Thread Jon LaBadie

On Thu, Jul 11, 2002 at 01:56:04PM -0700, Ashwin Kotian wrote:
> Hi Paul,
> 
> I think the reason we disabled indexing was because with indexing 
> enabled, amanda was trying to use the "u" option & try to update the 
> dumpdates file & was failing therein. Amanda wouldn't even do a dump & 
> it complained with it a "cannot update dumpdates file" error.

Shouldn't that be the "record" parameter, not the index param?
You can also compile amanda to use its own "amandates" file instead
of the system dumpdates file.


-- 
Jon H. LaBadie  [EMAIL PROTECTED]
 JG Computing
 4455 Province Line Road(609) 252-0159
 Princeton, NJ  08540-4322  (609) 683-7220 (fax)



Re: Netapp dumps using Amanda

2002-07-12 Thread Ashwin Kotian

Yes you are right. I'm sorry we have "indexing" enabled. It is the 
record that we have disabled. Sorry bout that part.

Ashwin.

Jon LaBadie wrote:

>On Thu, Jul 11, 2002 at 01:56:04PM -0700, Ashwin Kotian wrote:
>
>>Hi Paul,
>>
>>I think the reason we disabled indexing was because with indexing 
>>enabled, amanda was trying to use the "u" option & try to update the 
>>dumpdates file & was failing therein. Amanda wouldn't even do a dump & 
>>it complained with it a "cannot update dumpdates file" error.
>>
>
>Shouldn't that be the "record" parameter, not the index param?
>You can also compile amanda to use its own "amandates" file instead
>of the system dumpdates file.
>
>