lkml.org 
[lkml]   [1998]   [Aug]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    SubjectRe: 100Mbps TCP stalls in 2.1.115
    From
    Date
    Andi Kleen <ak@muc.de> writes:

    > Zlatko Calusic <Zlatko.Calusic@CARNet.hr> writes:
    > >
    > > E. g. FTP between two 100mbit hosts, transfer rate is ~50kb/sec, but
    > > many times it recovers after few (or few tens of) seconds.
    > >
    > > Sometimes it works, sometimes not. Tcpdump logs available on request.
    > >
    > > Any clue?
    >
    > What does /proc/net/netstat say after such a stall?

    {atlas} [~]% cat /proc/net/netstat
    TcpExt: SyncookiesSent SyncookiesRecv SyncookiesFailedEmbryonicRsts
    TcpExt: 0 0 0 0


    > Also could you try to reproduce it with sockets that have the SO_DEBUG
    > flag set ?
    >

    I should take my favorite TCP application and recompile with one
    additional setsockopt in code, am I right?
    I never did SO_DEBUG's. :)

    I noticed that stalls are periodical, with period of about 2-3
    minutes.

    I made a long ping session, and came up with this:

    PING div.srce.hr (161.53.3.13): 56 data bytes
    64 bytes from 161.53.3.13: icmp_seq=0 ttl=253 time=1.3 ms
    64 bytes from 161.53.3.13: icmp_seq=1 ttl=253 time=1.1 ms
    64 bytes from 161.53.3.13: icmp_seq=2 ttl=253 time=1.1 ms
    64 bytes from 161.53.3.13: icmp_seq=3 ttl=253 time=1.0 ms
    64 bytes from 161.53.3.13: icmp_seq=4 ttl=253 time=1.1 ms
    ... everything's ok, when ping times are around 1ms
    64 bytes from 161.53.3.13: icmp_seq=147 ttl=253 time=1.0 ms
    64 bytes from 161.53.3.13: icmp_seq=148 ttl=253 time=1.0 ms
    ... from now on, things are dog slow...
    64 bytes from 161.53.3.13: icmp_seq=150 ttl=253 time=2.1 ms
    64 bytes from 161.53.3.13: icmp_seq=151 ttl=253 time=2.0 ms
    64 bytes from 161.53.3.13: icmp_seq=152 ttl=253 time=2.0 ms
    64 bytes from 161.53.3.13: icmp_seq=153 ttl=253 time=2.1 ms
    64 bytes from 161.53.3.13: icmp_seq=154 ttl=253 time=2.1 ms
    64 bytes from 161.53.3.13: icmp_seq=155 ttl=253 time=2.1 ms
    64 bytes from 161.53.3.13: icmp_seq=156 ttl=253 time=2.1 ms
    64 bytes from 161.53.3.13: icmp_seq=157 ttl=253 time=74.9 ms
    64 bytes from 161.53.3.13: icmp_seq=158 ttl=253 time=28.1 ms
    64 bytes from 161.53.3.13: icmp_seq=161 ttl=253 time=2.2 ms
    64 bytes from 161.53.3.13: icmp_seq=163 ttl=253 time=2.1 ms
    ...
    64 bytes from 161.53.3.13: icmp_seq=335 ttl=253 time=2.1 ms
    64 bytes from 161.53.3.13: icmp_seq=336 ttl=253 time=2.0 ms
    ... from now on, things are working correctly again...
    64 bytes from 161.53.3.13: icmp_seq=337 ttl=253 time=1.1 ms
    64 bytes from 161.53.3.13: icmp_seq=338 ttl=253 time=1.1 ms
    ... and so on...


    Notice I cut a long listing into relevant pieces (watch those icmp_seq
    numbers)!

    Could this mean that some counter is under/overflowing or something
    similar?

    Unfortunately, I can't boot older kernel right now, cause I'm not
    sitting at the machine, and wouldn't want to push my luck... :)

    Thanks for your attention!
    --
    Posted by Zlatko Calusic E-mail: <Zlatko.Calusic@CARNet.hr>
    ---------------------------------------------------------------------
    Do not put statements in the negative form.

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.rutgers.edu
    Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html

    \
     
     \ /
      Last update: 2005-03-22 13:43    [W:0.022 / U:0.472 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site