I am trying this a 3rd time, W/O the atachments, and after resubscribing to
the list :-(

What a morning.

In any case the debug files can be found at http://beachcave.net/amanda

And I ma sorry if theis is a repeat, but I never saw the other 2 atempsts
come back from the list.



I am starting a new thread on this, as the old one was getting to be hard
to find things in, and I have not worked on this for a couple of days.

To revies, I am trying to get OpenBSD clients working with anything
resembling a modern version of Amanda. The version of OpenBSD I am using is
4.5, which is the current version. I have a 2.6.1 master machine, running
on Ubuntu Linux. I was sucesfully backing up a number of older versions of
OpenBSD machines, with various older versions of Amanda. Amanda 2.5.0p1
with bsd auth works on these new OpenBSD machines. Amanda 2.5.2p1 (which I
believe I need to support the user freindly recovery programs) does not
work,

Amcjeck runs fine, and the backups start, but > 50% of them fail part way
through the backup. I have tried going to bsdtcp auth with no improvement.
I have truned of client compression with no improvemnt.

Someone on the list told me that it appeared that the failure was occuring
in the stream_accept() function call. which is defined in
common-src/stream.c. I add some additional debuging for last nights run, as
follows:

         } else if(nfound == 0) {
                         dbprintf(plural(_("stream_accept: timeout after %d\n"),
                         _("stream_accept: timeout after %d\n"),
                                                                                
        timeout),
                                         timeout);
+               dbprintf(plural(_("stream_accept: server_socket =  %d\n"),
+                       _("stream_accept: server_socket = %d\n"),
+                                       server_socket),
+                                                                               
         server_socket);

But, it did not trigger. Amstaus reports:

pblab:wd0a                    1        18m finished (23:04:26)
pblab:wd0d                    1         0m finished (23:04:32)
pblab:wd0e                    0         1m finished (23:04:22), PARTIAL
pblab:wd0f                    1         5m finished (23:04:12), PARTIAL

Here is what's left over in the dumpdisk area 

-rw------- 1 amanda amanda 1452032 2009-08-26 22:59 pblab.wd0e.0
-rw------- 1 amanda amanda 5560320 2009-08-26 22:59 pblab.wd0f.1
-rw------- 1 amanda amanda  205411 2009-08-26 22:52 pdns2.wd0a.0
-rw------- 1 amanda amanda   88462 2009-08-26 22:53 pdns2.wd0e.0
-

Here are the size estimates:

wd0a 0 SIZE 98393
wd0a 1 SIZE 19100  
wd0f 0 SIZE 9507019
wd0f 1 SIZE 257082
wd0d 0 SIZE 180

I am atching the amandad debug files from last nights run.

I have compiled amanda on the OpenBSD machine with denguing symbols, so
that I run it under the debuger. Can anyone sugest what to do next to try
to localize the problem?

Thanks for all the help on this!!

PS, can someone breifly describe the control flow of sending a backup, so
that I cna figure out where to put more diagnostics in the appropriaye
code? I assume amandad is where that needs to go? Or does it need to go in
sendbackup? Or both?

BTW, here are teh DLE's for this amchine:

pblab wd0a stan-test 1 eth0
pblab wd0f stan-test 1 eth0
pblab wd0d stan-test 1 eth0
pblab wd0e stan-test 1 eth0

And here is the dumptype df:

define dumptype stan-test {
    global
        comment "Non-root partitions on reasonably fast machines"
        compress none
        maxdumps 4
        priority medium
#    auth "bsdtcp"
}

The only active lines in global are:

    index yes
    record yes

Any sugestions whatsover as to how to proceed with debuging this will be
greatly appreciated!!

Thanks, again.
-- 
One of the main causes of the fall of the roman empire was that, lacking
zero, they had no way to indicate successful termination of their C
programs.