On 2006-05-29 13:38, Michael D Schleif wrote:

* On 2006:05:25:08:25:07-0500 I, Michael D Schleif <mds AT helices DOT org>, 
scribed:

Something has changed in amanda.

I have been running amanda on this lan for several years.  For the most
part, DLE's have been constant for at least six months.  I have six
linux servers, all running debian.  Regarding amanda-server, my records
show that I upgraded amanda to version:

    2.4.5 on 16JUN05

Everything was backing up, and restoring, to my satisfaction, until last
week.  At that time, two servers (brono & jord) were terribly old,
regarding kernel and debian os.  So, I upgraded via aptitude, which also
upgraded amanda-client to version:

    2.5.0

Since that time, many -- but, NOT all -- DLE's on brono and jord are
FAIL'ing, e.g.:

    brono  /var  lev 0  FAILED [missing result for /var in brono response]
    jord   /var  lev 0  FAILED [missing result for /var in jord response]

Yes, both of these servers have many DLE's; but, as stated above, this
HAS been working without incident at the older version.  Numbers of
DLE's:

    brono  137
    jord   219

<snip />

Bdale Garbee published to debian repository version 2.5.0p2.

I have tried this on brono and jord, and this does NOT resolve the
problem.

I now have this on ALL of my boxen, except brono and jord, which I have
downgraded to 2.4.4p3-3.

Last night was my first completely successful backup in more than one
week!

I have received several private emails explaining the situation.  I do
understand those issues.  However, amanda DOES succeed in my situation
in versions _prior_ to v2.5.x -- and it FAILS in ALL v2.5.x ;<

In Amanda 2.4.x there is upper limit of 64K on the size of the request
packets using the UDP.
That resulted in errors once you got above the system limit.
The system limit for UDP packets is on some systems only 8K or so, but
most (all?) can be increased to 64K.  Larger than 64K is not possible

In Amanda 2.5.0 there is code that breaks up the request in multiple
chunks.  However, that code is only implemented in the server side.
The client code has not yet any provisions to re-assemble the multiple
requests (or even to detect that the packet was divided!).  I guess
that doing this in a backward compatible way was not evident (multiple
possibilities exist, some using feature bits, to extend the protocol,
another way is to change the server code even more).
Now, it just happens (why I don't know) that 2.5.0 breaks up those
chunks in 32Kbytes, while 64Kbytes would have be good enough too  (maybe
because the answer needs to fit in a 1 packet too -- which it currently
does, but maybe the implementor forsaw more info in the reply?).

It fixes it until you hit the 64Kbyte limit, at which time, 2.5.1 or
2.5.2 will have removed that limit, we hope.

Setting the limit to 64K instead of 32K is perfectly fine here.
But it does not solve the fundamental problem that was already present
in Amanda 2.4.x either.

This I do NOT understand.

Or I'm confused about what exactly you do not understand...

What am I missing?

Or me?

How will I know when a new version corrects this problem?

Watching the ChangeLog and NEWS file...


--
Paul Bijnens, xplanation Technology Services        Tel  +32 16 397.511
Technologielaan 21 bus 2, B-3001 Leuven, BELGIUM    Fax  +32 16 397.512
http://www.xplanation.com/          email:  Paul.Bijnens AT xplanation DOT com
***********************************************************************
* I think I've got the hang of it now:  exit, ^D, ^C, ^\, ^Z, ^Q, ^^, *
* F6, quit, ZZ, :q, :q!, M-Z, ^X^C, logoff, logout, close, bye, /bye, *
* stop, end, F3, ~., ^]c, +++ ATH, disconnect, halt,  abort,  hangup, *
* PF4, F20, ^X^X, :D::D, KJOB, F14-f-e, F8-e,  kill -1 $$,  shutdown, *
* init 0, kill -9 1, Alt-F4, Ctrl-Alt-Del, AltGr-NumLock, Stop-A, ... *
* ...  "Are you sure?"  ...   YES   ...   Phew ...   I'm out          *
***********************************************************************