On Tue, Dec 06, 2005 at 12:06:19PM -0500, Jean-Francois Malouin wrote:
> Hi,
> 
> I've got this amanda client (server and client at v2.4.5) with a 1.5TB
> raid splitted/chunked in ~70 DLEs (using gtar) and some of them don't
> make it to tape as the planner timeouts. Now, would someone be kind
> enough to jolt my memory as to what is the relationship between UDP
> packet size and the number/size of DLEs entries along with their
> exclude list? The FAQ mentions something about this (UDP packet size
> of 64Kb and "results missing") but that might not be quite up to date. 
> 
> This raid partition has been splitted using quite a few 'exclude
> append "./blah"' in the disklist and I wonder if it would be better to
> have them in a file on the client side rather than explicitely in the
> disklist file per se. Would that make any difference at all regarding
> what get stuffed in the UDP packets at the estimate phase -- making
> them smaller -- so that I wouldn't hit that hard limit?
> 

>From a Dec. 2004 reply by Paul Bijenes:

||
|| The problem seems to be in the reply packet.
|| 
|| I've already seen problems with a UDP-packet overflow, but that's
|| unlikely.  That problem happened with older versions where the UDP
|| size was only 8Kbyte or so. Currently it is 64K, but it could be
|| limited by the OS too, of course.  The reply packet is usually larger
|| than the request packet, because it contains 1 to 3 lines for each
|| DLE (level 0, current level, current plus 1).
||

-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)