I setup a mesos cluster with zookeeper, 3 master nodes and 12 agents nodes. There are three nodes on which running master and agent processes. When run c++ test framwork, it usually failed, but somtimes the four tasks can finished successfully.
In one agent node, I see stderr log below:
I0420 22:49:05.532886 17577 exec.cpp:162] Version: 1.2.0 I0420
22:49:05.556433 17582 exec.cpp:237] Executor registered on agent
3f442a52-2e2f-4799-8c1a-3a05b27120b9-S1 I0420 22:49:05.811959 17579
exec.cpp:415] Executor asked to shutdown
Also there are many connection closed logs in master node:
E0420 22:48:51.964684 22212 process.cpp:2426] Failed to shutdown
socket with fd 30: Transport endpoint is not connected E0420
22:48:54.974160 22212 process.cpp:2426] Failed to shutdown socket with
fd 30: Transport endpoint is not connected E0420 22:48:56.192914 22212
process.cpp:2426] Failed to shutdown socket with fd 27: Transport
endpoint is not connected E0420 22:48:57.999858 22212
process.cpp:2426] Failed to shutdown socket with fd 30: Transport
endpoint is not connected E0420 22:49:00.994969 22212
process.cpp:2426] Failed to shutdown socket with fd 30: Transport
endpoint is not connected E0420 22:49:03.994499 22212
process.cpp:2426] Failed to shutdown socket with fd 30: Transport
endpoint is not connected E0420 22:49:05.999225 22212
process.cpp:2426] Failed to shutdown socket with fd 30: Transport
endpoint is not connected E0420 22:49:11.194205 22212
process.cpp:2426] Failed to shutdown socket with fd 27: Transport
endpoint is not connected E0420 22:49:26.196691 22212
process.cpp:2426] Failed to shutdown socket with fd 27: Transport
endpoint is not connected E0420 22:49:41.198381 22212
process.cpp:2426] Failed to shutdown socket with fd 27: Transport
endpoint is not connected
It seems the master try to close a connection between itself and one agent, but the connection was closed. I guess that.
How to fix this or find the real reason of this problem.
