trying to repro CONTROLLER-1849 w/ artery

JamO Luhrsen

Tom, Robert, others,

the good news is I finally got a few things figured out and I can make
a local 3 node cluster work with artery (using udp).

The bad news is that I keep running out of shared memory (after not
too many iterations of my test).

I am launching my docker containers with 5G of /dev/shm, but something
is eating it up when I'm killing and restarting the controller in a
single container. Once it's gone, artery craps out and can't send any
messages, so the test is worthless after that.

I need to find a way to clean/flush whatever's using that memory after
each iteration, I think.

anyone with any expertise here?


Join { to automatically receive all group messages.