If I cycle through running the *_thr programs on two machines running
a large number of iterations
(e.g. 1million) changing the message size each time, I get the error:
message of incorrect size received
each time, on the machine running local_thr.
It doesn't happen at lower iteration counts (below about 200K).
I think some msgs are getting lost. The behavior isn't completely repeatable:
sometimes the smaller message sizes work, sometimes not.
Here's what I've done:
On machine A:
set count=200000
foreach size ( 2 8 64 512 1024 2048 4096 16384 )
echo -n "$size $count " ; time ./remote_thr tcp://10.33.26.26:5555
$size $count
end
On machine B (10.33.26.26):
set count=200000
foreach size ( 2 8 64 512 1024 2048 4096 16384 )
time ./local_thr 'tcp://*:5555' $size $count
end
And the output I get is:
2 200000 0.030u 0.010s 0:00.05 80.0% 0+0k 0+0io 0pf+0w
8 200000 0.020u 0.030s 0:00.23 21.7% 0+0k 0+0io 0pf+0w
64 200000 0.060u 0.030s 0:00.36 25.0% 0+0k 0+0io 0pf+0w
512 200000 0.070u 0.120s 0:01.15 16.5% 0+0k 0+0io 0pf+0w
1024 200000 0.120u 0.180s 0:01.98 15.1% 0+0k 0+0io 0pf+0w
2048 200000 0.160u 0.300s 0:03.82 12.0% 0+0k 0+0io 0pf+0w
and
message of incorrect size received: expected 2, got 8
0.050u 0.010s 0:00.37 16.2% 0+0k 48+0io 1pf+0w
message of incorrect size received: expected 8, got 64
0.050u 0.020s 0:00.23 30.4% 0+0k 0+0io 0pf+0w
message of incorrect size received: expected 64, got 512
0.100u 0.090s 0:00.36 52.7% 0+0k 0+0io 0pf+0w
message of incorrect size received: expected 512, got 1024
0.050u 0.370s 0:01.15 36.5% 0+0k 0+0io 0pf+0w
message of incorrect size received: expected 1024, got 2048
0.180u 0.660s 0:01.99 42.2% 0+0k 0+0io 0pf+0w
<hang>
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev
