Don't use workerTimeout of 60 by default, use 120 instead.
When cloning repositories (5gb) takes longer then 60 seconds, we run into the issue, that the connection resets with:
19:57:23.140369 pkt-line.c:80 packet: sideband< \1
19:57:25.900806 pkt-line.c:80 packet: sideband< PACK ...
packet_write_wait: Connection to xxx.xxx.xxx.xxx port 22: Broken pipe
fatal: the remote end hung up unexpectedly
fatal: early EOF
fatal: index-pack failed
And the log says:
[2019-10-18T13:53:06.957658 #77963] WARN -- : Unicorn::WorkerKiller send SIGQUIT (pid: 77963) alive: 354297 sec (trial 1)
[2019-10-18T13:54:08.251687 #10] ERROR -- : worker=1 PID:77963 timeout (61s > 60s), killing
If the workerTimeout is increased to 120 instead we get:
20:19:31.576297 pkt-line.c:80 packet: sideband< \1
20:19:36.545481 pkt-line.c:80 packet: sideband< PACK ...
20:31:09.031982 pkt-line.c:80 packet: sideband< 0000
and the project (5gb) is checked out properly. So I expect, even though the checkout takes 12 minutes here, the 60 seconds is definitely not enough to not beeing killed by the WorkerKiller.
Edited by Jan Schütze