When a worker process in pmap
gets killed, the parent worker aborts with a ProcessExitedException
. Is there any way to prevent this? Fault tolerant `pmap` when worker goes down shows that it works when you specify a retry delay, but I don’t want to retry.
Example:
using Distributed
addprocs(2)
@everywhere function foo(i)
if myid() == 2
exit(1)
# throw(ErrorException()) # works as expected
end
sleep(2)
"jipeee"
end
pmap(foo, 1:3, on_error=x->-1)
# Worker 2 terminated.ERROR:
# ProcessExitedException(2)
# Stacktrace:
# ...
# This works, but we don't want to retry:
# pmap(foo, 1:3, on_error=x->-1, retry_delays=zeros(3))