I don't know a lot of the internals of TAR, but I find much of this curious:


- Why is the --listed-incremental option being used for a level 0? Shouldn't
it be left off to do a full dump? 

- I've heard on the list many times that GnuTAR only understands one level
of incremental. Why is it trying 4 and 5? (I'm guessing this is an AMANDA
convention, happening because the last backup was 4, so planner tries 4 and
5.)

- Your exclude file is reported as missing each time. Have you placed a copy
of your exclude file on the client, server, or both? (IIRC, it should be on
the client.)

- I seem to remember that for incrementals to work with GnuTAR, there is a
file that keeps track of the stuff on the disk. A permissions problem with
that file could cause it to always return full results if it never has a
chance to record what was backed up last time. 

> -----Original Message-----
> From: Brashers, Bart -- MFG, Inc. [mailto:Bart.Brashers AT mfgenv DOT com]
> Sent: Friday, August 23, 2002 3:11 PM
> To: Amanda Users (E-mail)
> Subject: RE: Questions about bumpsize, frustration
> 
> 
> 
> Thanks for your responses, here is a combined answer:
> 
> > Is it possible that it needs an update to dump? 
> 
> I use tar (GNU tar) 1.13.19, not dump.
> 
> > Is this happening on all of your machines or just a few?  
> 
> I'm running amanda on the head node of my cluster, so all the 
> disks are
> local.  I only back up 1 of my slave nodes (since the others 
> are identical)
> and then only parts (/var/log, /usr/spool/PBS, etc.).  The 
> problem only
> happens on this one machine (itself), but there's not much on 
> any other
> machine to back up.
> 
> > What version of AMANDA?
> 
> Version is amanda-2.4.2p2.
> 
> > Do you set record to yes in your dumptype?
> 
> Yes, I use "record yes".
> 
> > What's the output of `amadmin <config> disklist beow 
> /home/proj/bp-9725`?
> 
> line 49:
>     host beowulf:
>         interface default
>     disk /home/proj/bp-9725:
>         program "GNUTAR"
>         exclude list "/usr/local/lib/amanda/exclude.gtar"
>         priority 1
>         dumpcycle 14
>         maxdumps 1
>         strategy STANDARD
>         compress NONE
>         auth BSD
>         kencrypt NO
>         holdingdisk YES
>         record YES
>         index YES
>         skip-incr NO
>         skip-full NO
> 
> > What does the /tmp/amanda/sendsize.debug file on beow look like? 
> 
> From last night's, 8/22:
> 
> -------------
> calculating for amname '/home/proj/bp-9725', dirname 
> '/home/proj/bp-9725'
> sendsize: getting size via gnutar for /home/proj/bp-9725 level 0
> sendsize: missing exclude list file 
> "/usr/local/lib/amanda/exclude.gtar"
> discarded
> sendsize: spawning /usr/libexec/runtar in pipeline
> sendsize: argument list: /bin/gtar --create --file /dev/null 
> --directory
> /home/proj/bp-9725 --one-file-system --listed
> -incremental /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_0.new
> --sparse --ignore-failed-read --totals .
> Total bytes written: 15848878080 (15GB, 15GB/s)
> .....
> sendsize: getting size via gnutar for /home/proj/bp-9725 level 4
> sendsize: missing exclude list file 
> "/usr/local/lib/amanda/exclude.gtar"
> discarded
> sendsize: spawning /usr/libexec/runtar in pipeline
> sendsize: argument list: /bin/gtar --create --file /dev/null 
> --directory
> /home/proj/bp-9725 --one-file-system --listed
> -incremental /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_4.new
> --sparse --ignore-failed-read --totals .
> Total bytes written: 15848867840 (15GB, ?B/s)
> .....
> sendsize: getting size via gnutar for /home/proj/bp-9725 level 5
> sendsize: missing exclude list file 
> "/usr/local/lib/amanda/exclude.gtar"
> discarded
> sendsize: spawning /usr/libexec/runtar in pipeline
> sendsize: argument list: /bin/gtar --create --file /dev/null 
> --directory
> /home/proj/bp-9725 --one-file-system --listed
> -incremental /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_5.new
> --sparse --ignore-failed-read --totals .
> Total bytes written: 15848867840 (15GB, ?B/s)
> -----------------
> 
> So yes, each of level 0, 4 and 5 return about the same size 
> from sendsize.
> There was about 680M changed last night, so maybe I should 
> look at 8/20
> instead.  It's just about the same story.  The oldest surviving
> sendsize.debug is from 8/19.  Is there some place I can 
> change a setting to
> keep more of these temp files?  I can't find it in amanda.conf.
> 
> [root amanda/Daily]% \ls -1
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725*
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_0
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_1
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_2
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_3
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_4
> 
> [root amanda/Daily]% head
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_4
> 1029991548
> 2081 19611660 ./contemp.cpuf/post
> 2081 19611665 ./contemp.cpuf/vis/plots
> 2081 20004866 ./calpuff/calmet
> 2081 19988482 ./calpuff
> 2081 19529746 ./calpuff/bc.post
> 2081 19939332 ./calpuff/calmet/upa
> 2081 16367618 ./calpuff/dep.con
> 2081 19955715 ./calpuff/calmet/sfc
> 2081 20135938 ./contemp.cpuf
> [root amanda/Daily]% head
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_3
> 1029509897
> 2081 19611660 ./contemp.cpuf/post
> 2081 19611665 ./contemp.cpuf/vis/plots
> 2081 20004866 ./calpuff/calmet
> 2081 19988482 ./calpuff
> 2081 19529746 ./calpuff/bc.post
> 2081 19939332 ./calpuff/calmet/upa
> 2081 16367618 ./calpuff/dep.con
> 2081 19955715 ./calpuff/calmet/sfc
> 2081 20135938 ./contemp.cpuf
> [root amanda/Daily]% head
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_2
> 1029428874
> 2081 19611660 ./contemp.cpuf/post
> 2081 19611665 ./contemp.cpuf/vis/plots
> 2081 20004866 ./calpuff/calmet
> 2081 19988482 ./calpuff
> 2081 19529746 ./calpuff/bc.post
> 2081 19939332 ./calpuff/calmet/upa
> 2081 16367618 ./calpuff/dep.con
> 2081 19955715 ./calpuff/calmet/sfc
> 2081 20135938 ./contemp.cpuf
> [root amanda/Daily]% head
> /var/amanda/gnutar-lists/beowulf_home_proj_bp-9725_1
> 1029343540
> 2081 19611660 ./contemp.cpuf/post
> 2081 19611665 ./contemp.cpuf/vis/plots
> 2081 20004866 ./calpuff/calmet
> 2081 19988482 ./calpuff
> 2081 19529746 ./calpuff/bc.post
> 2081 19939332 ./calpuff/calmet/upa
> 2081 16367618 ./calpuff/dep.con
> 2081 19955715 ./calpuff/calmet/sfc
> 2081 20135938 ./contemp.cpuf
> 
> This just tells me amanda has always been confused, through 
> all the levels
> of backup.
> 
> Hopefully, something in the above sticks out like a sore thumb, and a
> lightbulb will go off above someone's head.  I sure don't see 
> why sendsize
> is so wrong.
> 
> Bart
> ---
> Bart Brashers                         MFG Inc.
> Air Quality Meteorologist             19203 36th Ave W Suite 101
> bart.brashers AT mfgenv DOT com            Lynnwood WA 98036-5707
> http://www.mfgenv.com                 425.921.4000 Fax: 425.921.4040
>