RE: Cflowd dies, cfdcollect can't reconnect until after reboot.

From: Larry Rosenman (ler@airmail.net)
Date: Tue Jan 16 2001 - 11:10:37 PST

  • Next message: Jeff Olenchek: "cflowd and inputIfIndex"

    I've also seen cfdcollect die with a SIGSEGV on a regular basis.

    Seems to be an unreasonable read length.
    (FreeBSD 4.2-STABLE)

    LER

    -----Original Message-----
    From: cflowd-owner@ipn.caida.org [mailto:cflowd-owner@ipn.caida.org]On
    Behalf Of Soren Petersen
    Sent: Tuesday, January 16, 2001 12:02 PM
    To: Martin_Nieuwelaar@infonet-europe.com
    Cc: cflowd@caida.org
    Subject: RE: Cflowd dies, cfdcollect can't reconnect until after reboot.

    Hi Martin,

    Thank you for your excellent explanation. That turned to be exactly what
    was happening here too. I'm running cflowd on a solaris 8 box. I guess
    I'll have to try and automate the restart process till the problem is
    fixed. Does anyone know when to expect a new version of cflowd that might
    address this problem?

     - Soren

    On Tue, 16 Jan 2001 Martin_Nieuwelaar@infonet-europe.com wrote:

    > Soren,
    >
    > I have experienced this behaviour on Red Hat 6.0. I traced it to a
    problem
    > with the cflowd process. If you run something like 'netstat -an' you will
    > see what processes are bound to what ports. cflowd should be bound to
    port
    > 2056 as default, however in this "error state" I noted that nothing is
    bound
    > to 2056. As you experienced, stopping and restarting cflowd does not
    help.
    > This appears to be due to the fact that when cflowd goes into the "error
    > state" it looses track (for want of a better term) of its shared memory
    > segments and semaphore arrays. Use 'ipcs -a' to display these. I have
    > found that if these are removed, using the ipcrm command, cflowd will
    again
    > restart successfully.
    >
    > I found that under heavy load cflowd would mess up in this manner every
    few
    > days. I ended up writing a script that checks to see if port 2056 is in
    > use, and if it isn't (cflowd has become messed up) then I shut down
    > cfdcollect and cflowdmux, remove the shared memory and semaphores, and
    start
    > it up again. Works like a charm.
    >
    > I can relate to your frustration in finding a program that will not run
    > properly until after the box has been rebooted. Hehe, it's the sort of
    > behaviour I'd expect to see on a windows machine. :-)
    >
    > --
    > "Buying a car because it's reliable is like marrying
    > someone because they are punctual" - Jeremy Clarkson
    >
    >
    >
    > > -----Original Message-----
    > > From: Soren Petersen [mailto:sp@int.tele.dk]
    > > Sent: Tuesday, 16 January 2001 14:09
    > > To: cflowd@caida.org
    > > Subject: Cflowd dies, cfdcollect can't reconnect until after reboot.
    > >
    > >
    > > Hi,
    > >
    > > Cflowd has been acting strange ever since I upgraded to cflowd-2-1-b1.
    > > The cflowd process dies at a random interval without leaving
    > > a trace in
    > > syslog. When I start it back up, cfdcollect can no longer connect to
    > > cflowd, which it shows by writing the following in syslog:
    > >
    > > Jan 16 13:38:40 machine cfdcollect[490]: [ID 365783 local6.error] [E]
    > > connect(4,1 31f6c,16) (host 127.0.0.1 port 2056) failed: Connection
    > > refused {CflowdServer.cc:112}
    > >
    > > Even if I shut down cflowd, cflowdmux and cfdcollect and
    > > start them all
    > > back up again, cfdcollect still can't connect. The only way
    > > I've been able
    > > to make it work again has been to reboot the machine.
    > >
    > > Has anyone else seen this behavior, and does someone have a
    > > solution to
    > > offer?
    > >
    > > - Soren
    > >
    > > --
    > > cflowd mailing list
    > > cflowd@caida.org

    --
    cflowd mailing list
    cflowd@caida.org
    

    -- cflowd mailing list cflowd@caida.org



    This archive was generated by hypermail 2b29 : Tue Jan 16 2001 - 11:26:35 PST