Stop timeout monitoring
With the SA server, it appears that the numpy savetxt
function is particularly slow. When dealing with big simulations (500x500 grid), its slowness causes the server to remain inside the write_output
function for too long. This ultimately leads the launcher to consider the server as timed out and kill it.
The new features supported through this MR are:
- send a message asking the launcher to stop looking for time-out (this would be sent just before writing the result files)
- make the time-out parameter
server_ping_interval
an option accessible to the user (make sure this is consistent withjob_update_interval
)
Edited by SCHOULER Marc