Hello,
Just to continue about a previous thread: 'Running >10 jobs in parallel on the same remote machine'.
My Environment: OS (Centos 6.4 - 64 bits, GNU parallel 20140322).
Sorry for this long and boring email but it can helps others.
It's surely obvious that updating the parameter 'MaxStartups' induced also update 'MaxSessions' but it take me sometimes to find it.
Ole, could you update the warning to update 'MaxStartups' and 'MaxSessions'?
Below the action.
It's nice to have this warning form parallel
BEGIN_OUTPUT_PARALLEL
...
parallel: Warning: ssh to peritmp1 only allows for 13 simultaneous logins.
You may raise this by changing /etc/ssh/sshd_config:MaxStartup on peritmp1.
Using only 12 connections to avoid race conditions.
....
Computers / CPU cores / Max jobs to run
1:local / 32 / 32
2:peritmp1 / 16 / 12
END_OUTPUT_PARALLEL
It's true
that sshd_config with parameter MaxStartups must to be updated '#MaxStartups 10:30:60'.
Ok , i reproduce the issue.
BEGIN_OUTPUT
# for i in {1..100}; do (ssh root@peritmp1 uname &) 2>&1; done | grep ssh_exchange| wc -l
47
#
END_OUTPUT
I updated MaxStartup to 'MaxStartups 100:30:1000' and restarted sshd daemon. So now no issue from CLI.
BEGIN_OUTPUT
# for i in {1..100}; do (ssh root@peritmp1 uname &) 2>&1; done | grep ssh_exchange| wc -l
0
#
END_OUTPUT
With parallel, i have always the issue:
BEGIN_OUTPUT_PARALLEL
...
parallel: Warning: ssh to peritmp1 only allows for 9 simultaneous logins.
You may raise this by changing /etc/ssh/sshd_config:MaxStartup on peritmp1.
Using only 8 connections to avoid race conditions.
Computers / CPU cores / Max jobs to run
1:local / 32 / 32
2:peritmp1 / 16 / 8
END_OUTPUT_PARALLEL
I updated sshd_conf, so updated '#MaxSessions 10' to 'MaxSessions 100' and restarted sshd.
Now, it's fine.
BEGIN_OUTPUT_PARALLEL:
...
Computers / CPU cores / Max jobs to run
1:local / 32 / 32
2:peritmp1 / 16 / 15
END_OUTPUT_PARALLEL
Regards,
Phoebus