[Veritas-bu] NDMP backups going 30 hours
2001-10-02 14:35:39
Subject: |
[Veritas-bu] NDMP backups going 30 hours |
From: |
Steve Kappel <steve.kappel AT veritas DOT com> (Steve Kappel) |
Date: |
Tue, 2 Oct 2001 13:35:39 -0500 (CDT) |
Jeff Kennedy wrote:
>I was under the impression that all this did was not allow browsing in
>NBU but that the data was still there, if you knew what you were after.
>So for disaster recovery it was fine, or if you knew the path to the
>data you could enter it and still restore. Is this incorrect?
I am not aware of any mechanism in NetBackup that allows a restore
to specify an object that is not in the catalog.
You could do a manual restore on the filer using the native restore
command. Of course this would require figuring out which tape
and file # it is in.
>Steve Kappel wrote:
>>
>> Jeff Kennedy wrote:
>> >Would the directive 'set HIST=N' help this situation?
>>
>> Sure, if you don't need to be able to restore parts of the backup.
>> HIST=N means no file history so all that will be in the catalog
>> is the backup directory; that is all that you could restore using
>> NetBackup.
>>
>> >Steve Kappel wrote:
>> >>
>> >> Jeff Kennedy wrote:
>> >> >But the data has stoped being transferred 10 hours earlier. I have 3
>> >> >ndmp backups that are running 24+ hours but no data has moved for over 6
>> >> >hours. The filer is doing nothing and the ndmp session is halted
>> >> >because the backup is complete, all it's waiting for is a close message
>> >> >from NBU.
>> >> >
>> >> >I checked some stats on my NBU server and the root disk is fairly
>> >> >hammered, it's reading at 3MB/sec with nothing else going on; no other
>> >> >backups, no restores. And this server is dedicated to backups, no other
>> >> >services.
>> >> >
>> >> >The images are all on an A5000 so it's not writing the images (btw, the
>> >> >A5000 is idle too). The only thing I can figure is the BPFSMAP_TMP
>> >> >directory that holds an inode map for each ndmp session. These files
>> >> >are fairly huge, 300MB's to 800MB's each. What I'm wondering is what's
>> >> >happening at this point in the backup? The inode maps are run at the
>> >> >beginning of the backup on the filer so what is happening after the
>> >> >backup is complete? Is this verification of some type?
>> >>
>> >> Node-based NDMP file history (that is messages containing "DIR" and
>> >> "NODE" info) must be collected until the end of the backup and
>> >> then processed to convert into the full path catalogs used by
>> >> NetBackup. We call this the "bpfsmap phase". It is responsible
>> >> for the temporary files in BPFSMAP_TMPDIR as well as the memory
>> >> utilization. Both resources vary by the number of objects in
>> >> the backup. See the NetBackup NDMP Admin Guide for information.
>> >>
>> >> The NDMP session cannot be finished nor the tape released as
>> >> the converted catalog needs to be written to tape.
>> >>
>> >> The upcoming 4.5 release will radically improve this.
>> >>
>> >> __________________________________________________________________________
>> >> Steve Kappel steve.kappel AT veritas DOT com
>> >> VERITAS Software (Engineering) steve AT stevekappel DOT com (Personal)
>> >
>> >--
>> >=====================
>> >Jeff Kennedy
>> >Unix Administrator
>> >AMCC
>> >jlkennedy AT amcc DOT com
>>
>> __________________________________________________________________________
>> Steve Kappel steve.kappel AT veritas DOT com
>> VERITAS Software (Engineering) steve AT stevekappel DOT com (Personal)
>
>--
>=====================
>Jeff Kennedy
>Unix Administrator
>AMCC
>jlkennedy AT amcc DOT com
__________________________________________________________________________
Steve Kappel steve.kappel AT veritas DOT com
VERITAS Software (Engineering) steve AT stevekappel DOT com (Personal)
|
|
|