[Spread-users] Messages getting trapped

[Doug Palmer]

We're seeing an odd behaviour with some of our programs.

We're running spread on windows XP using 4.0.0.

If we send a stream of messages, then the first 788[*] messages will be
processed and delivered. No more messages are then delivered. During
this time, we see the memory used by the spread daemon rise from about
2800k to 6300k in the task manager and then stop; the rising memory
seems to be a good indicator of something going wrong. sptmonitor shows
the group leader to have a large number of s_retrans, which keeps
growing after everything else has stopped.

We can sometimes disconnect and sometimes not. If we do cleanly
disconnect, then when we try to run the same program again, we get a -6
error, which suggests that the daemon still thinks that the connection
is present.

Has anybody seen anything like this before?


[*] For a particular configuration, it is consistently 788 messages. For
other configurations, we get different results, but the number of
messages seems quite stable for a specific configuration.