Re: Running out of buffers?

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Fri, 27 Apr 2018 12:36:31 +0200
    From:        Martin Husemann <martin%duskware.de@localhost>
    Message-ID:  <20180427103631.GH21562%mail.duskware.de@localhost>

  | Syslogd could catch it and log a better error message,

How?   All it can ever discover is that some datagram (on its
unix domain socket) lost a message.   Where it came from,
what it was about (even how many messages were dropped)
is unavailable, what else would syslog be expected to do?

What's more, what is the reader of the log in which the error
from syslog is reported supposed to do?

"syslog lost a message - probably some debug spam from a mailer
or something (those are the ones that tend to log most) - who cares?"

If that's to be the response, and in may cases, that is what it
wil be, what's the point of even writing that message?

  | but it IS an
  | important error and should not be silently ignored, IMHO.

Where it is important (or might be) is at the sender  of the
message - if the sender discovers that the message could
not be buffered, then that sender can wait a bit, and send
again later (or later send a different message indicating how
many messages from it were lost, or something) - at least
that way the received gets to more information than (something
was lost).

This is kind of like missed phone calls, back in the days before
caller-ID, answering machines, ...   Your neighbour/roommate
tells you "your phone rang".   Great - what do I do with that info?
But if the caller calls back, and says "I tried to call you earlier..."
then you know what happened, regardless of whether anyone
told you they heard your phone ring.

A message that says you missed something, but no clue what,
is close to useless

kre