Re: [logs] Re: syslogd / some analysis

From: Marcus J. Ranum (mjrat_private)
Date: Wed Jan 30 2002 - 06:53:48 PST

  • Next message: Nick Vargish: "Re: [logs] syslog/mysql/webGUI."

    Rich Salz wrote:
    >> presumably some output queue was getting overrun and messages
    >> were silently discarded - my sendto(...) code never returned an error,
    >> by the way...
    >
    >UDP packets aren't acked, so if the server couldn't keep up and was
    >dropping them, that's the behaviour you'd expect.
    
    I measured this back in the early 90's and I bet it hasn't changed since
    then. The problem isn't that the server can't keep up, it's that the kernel
    discards the packets before they even get to the input cycle of the
    server. There have been huge changes in the various BSD kernels but I
    bet it's still the same thing.
    
    What I observed was that if I sent 10,000 UDP syslogs from machine 'A'
    to server 'B' and had machine 'C' running tcpdump to count them I only
    saw machine 'A' transmit a few hundred. The client side code on 'A' (as you
    say) didn't register an error because UDP isn't required to. What was
    happening was there was a queue max depth on the interface output queue
    that was getting overrun and the kernel was just chucking the packets.
    TCP retransmits took care of making sure things worked for TCP(with a
    big thruput hit) but the UDP got silently lost. My informal testing last night
    seems to show that the same thing still happens. This is _correct_ behavior
    for a UDP application so it's not surprising.
    
    mjr.
    ---
    Marcus J. Ranum          Chief Technology Officer, NFR Security, Inc.
    Work:                           http://www.nfr.com
    Personal:                      http://www.ranum.com
    
    
    ---------------------------------------------------------------------
    To unsubscribe, e-mail: loganalysis-unsubscribeat_private
    For additional commands, e-mail: loganalysis-helpat_private
    



    This archive was generated by hypermail 2b30 : Wed Jan 30 2002 - 07:06:27 PST