Re: [HTCondor-users] condor_ssh_to_job issues

[Carl Edquist <edquist@xxxxxxxxxxx>]

Thanks Rolf,

In this case i don't know if condor is doing ssh multiplexing under the 
hood or not, though i suppose that's one possible explanation.  I have a 
feeling it's not though, as i think it happens even when the ssh session 
closed wasn't the first one opened.  (not 100% sure though.)  I'm thinking 
this is something the condor team may be able to fix.

Carl

...

PS

By the way, outside of condor_ssh_to_job, when i use ssh multiplexing, i 
never do ControlMaster=auto anymore.  It adds guesswork to keeping track 
of whether there is still a master present and which session is the master 
that can't be closed.

Instead, i explicitly set up a ControlMaster with no command to run in the 
background

	ssh -MNf host

and connect to it in other windows without mentioning the ControlMaster 
setting

	ssh host

and then take down the background ControlMaster explicitly when i'm done 
with all of them

	ssh -O exit host

(Where the ControlPath pattern for host can be set in your ~/.ssh/config, 
or else by including -o ControlPath=... in the ssh command lines)

Fun times..!  :)


On Wed, 6 Mar 2019, Rolf Seuster wrote:

> re 2, I guess this would be the case if the ControlMaster feature is 
> used here. Basically with ControlMaster,
> ÂÂÂÂÂÂÂ all subsequent ssh connections to that host go through the same 
> connection.
>
> There might be ways to avoid this, but this would come with a penalty. 
> That might point you into the
> right direction - I never got ssh_to_job to work :(
>
> HTH, Rolf
>
> On 2019-03-06 5:22 p.m., Carl Edquist wrote:
> > Two somewhat annoying issues i ran into with the otherwise *very cool* 
> > condor_ssh_to_job (for a docker universe job, in case that matters).
> >
> >
> > 1. I've noticed that when i condor_ssh_to_job (for an interactive ssh 
> > terminal session), and then i resize my terminal, bash's readline 
> > support gets all confused about where to wrap lines.
> >
> > This is true even if i do a "kill -WINCH $$" from bash, which suggests 
> > to me that the pty created for the condor_ssh_to_job session has wrong 
> > information that's not getting updated.
> >
> > Maybe the SIGWINCH from my terminal is not making its way to the pty 
> > created because the sshd is outside the docker container's PID 
> > namespace, or maybe the pty just doesn't update its size info after it 
> > gets created.
> >
> > Any ideas?
> >
> >
> > 2. It seems that if i condor_ssh_to_job multiple times to keep 
> > multiple ssh sessions open to the same job (which works fine), then 
> > when i close the first of them, the rest are immediately cut off.
> >
> > This came as a surprise and seems like maybe a bug, or maybe an 
> > abusive misfeature.
> >
> >
> >
> > Thanks..!
> > Carl
> > _______________________________________________
> > HTCondor-users mailing list
> > To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx 
> > with a
> > subject: Unsubscribe
> > You can also unsubscribe by visiting
> > https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users
> >
> > The archives can be found at:
> > https://lists.cs.wisc.edu/archive/htcondor-users/
>
>
> _______________________________________________
> HTCondor-users mailing list
> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users
>
> The archives can be found at:
> https://lists.cs.wisc.edu/archive/htcondor-users/