Not sure what you're asking, the master just checks for bi-directional connectivity: the slave can receive the master's ping and the master can receive the slave's pong. If that's not happening the health check will time out.
On Sun, Aug 9, 2015 at 6:40 PM, sujz <[email protected]> wrote: > Hi all, > I ran into this problem with these steps: > 1: Start master and slave successfully. > 2: Stop slave by pressing Ctrl+C. > 3: After slave stopped, restart slave within 75s, it prompts this error: > Slave asked to shut down by master because 'health check timed out'. > > After reading the code and searching on the internet, I knew that after > slave being observed disconnected, master continuously to send > PingSlaveMessage for MAX_SLAVE_PING_TIMEOUTS times, during each time > waiting for pong message from slave for SLAVE_PING_TIMEOUT, so the total > waiting time is > MAX_SLAVE_PING_TIMEOUTS * SLAVE_PING_TIMEOUT=75s, if slave restarts within > 75s, master believes this slave failsover and accepts re-register request, > otherwise, removes this slave. > > For my situation, within 75s, slave restarts occasionally failed with > message 'Slave asked to shut down by master because 'health check timed > out'. I also found the bug report MESOS-2679 discussing this problem, but > I think it didn't explain what master checks and what kinds of reasons may > cause health check timed out, could anybody be kind to give more > explaination? So I can restart slave successfully all the time. > > Thanks very much and best regards! >
