Wed, 10 Feb 1993 02:48:22 +0100
|
The new DISTRIBUTE has been running on SEARN for about 48h now, and while
the difference in wall-clock time is not very impressive on SEARN (which
is mostly I/O bound), the CPU consumption is down to about the same level
as LMail, which is to say about 3 times less than before.
When watching the console logs for "abnormal" behaviour, you should
disregard any job coming from LISTEARN. I wasted over 1h verifying that
all the "crazy" DISTRIBUTE paths I was observing were really what
LISTSERV had been ordered to do by the originating LISTEARN, rather than
some obscure bug. Here is an example:
Date: Sun, 7 Feb 93 23:52:47 IST
From: ListEARN List Processor (1.3) <[log in to unmask]>
Subject: Output of your job "ERIC"
To: [log in to unmask]
Path information (excerpt):
RICEVM1 : Direct send
UBVM : RICEVM1 UBVM
USCVM : RICEVM1 USCVM
SEARN : Direct send
UTORONTO: SEARN RICEVM1 PSUVM MCGILL1 UTORONTO
RUTVM1 : FRMOP11 RICEVM1 PUCC RUTVM1
When acting as a DISTRIBUTE relay, LISTSERV merely implements the paths
calculated by the originating server. If the routing for RICEVM1 varies
depending on the final destination, that is what LISTSERV will do and
this is not a bug. If in doubt, check the old console logs to see if the
routing has changed (bear in mind that small changes are possible due to
new subscribers and so on).
Eric
|
|
|