No subject
Fri Mar 19 13:53:46 EDT 2010
ing I couldn't reproduce the orphans with out doing something a little =
odd namely, ssh ... && =A0sleep 1d. Although the fact that a user r=
eported suggests that it may occur in the real world as well. The question =
is this: should an application explicitly clean up resources it allocates? =
or should an application rely on the user not only knowing that there is th=
e potential for a resource leak but also knowing enough to do the right thi=
ng to avoid it (eg ssh -tt ...)? In my opinion, as a matter of principle, i=
f PV spawns a process it should explicitly clean it up and there should be =
no way it can become an orphan. In this case the fact that the orphan can h=
old ports open is particularly insidious, because further connection attemp=
t on that port fails with no helpful error information. Also it is not very=
difficult to clean up a spawned process. What it comes down to is a little=
book keeping to hang on to the qprocess handle and a few lines of code cal=
led from pqCommandServerStartup destructor to make certain it's cleaned=
up. This is from the patch I submitted when I filed the bug report.<br>
<br>
+ =A0 =A0// close running process<br>
+ =A0 =A0if (this->Process->state()=3D=3DQProcess::Running)<br>
+ =A0 =A0 =A0{<br>
+ =A0 =A0 =A0this->Process->close();<br>
+ =A0 =A0 =A0}<br>
+ =A0 =A0// free the object<br>
+ =A0 =A0delete this->Process;<br>
+ =A0 =A0this->Process=3DNULL;<br>
<br>
I think if the cluster admins out there new which ssh options (GatewayPorts=
etc) are important for ParView to work seamlessly, then they might be will=
ing to open them up. It's my impression that the folks that build clust=
ers want tools like PV to be easy to use, but they don't necessarily kn=
ow all the in's and out's of confinguring and running PV.<br>
<br>
Thanks for looking at this again! The -tt option to ssh is indeed a good fi=
nd.<br>
<br>
Burlen<br>
<br>
pat marion wrote:<br>
<blockquote class=3D"gmail_quote" style=3D"border-left: 1px solid rgb(204, =
204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div class=3D"im"=
>
Hi all!<br>
<br>
I'm bringing this thread back- I have learned a couple new things...<br=
>
<br>
-----------------------<br>
No more orphans:<br>
<br>
Here is an easy way to create an orphan:<br>
<br>
=A0 $ ssh localhost sleep 1d<br>
=A0 $ <press control c><br>
<br>
The ssh process is cleaned up, but sshd orphans the sleep process. =A0You c=
an avoid this by adding '-t' to ssh:<br>
<br>
=A0$ ssh -t localhost sleep 1d<br>
<br>
Works like a charm! =A0But then there is another problem... try this comman=
d from paraview (using QProcess) and it still leaves an orphan, doh! =A0Go =
back and re-read ssh's man page and you have the solution, use '-t&=
#39; twice: ssh -tt<br>
<br>
-------------------------<br>
GatewayPorts and portfwd workaround:<br>
<br>
In this scenario we have 3 machines: workstation, service-node, and compute=
-node. =A0I want to ssh from workstation to service-node and submit a job t=
hat will run pvserver on compute-node. =A0When pvserver starts on compute-n=
ode I want it to reverse connect to service-node and I want service-node to=
forward the connection to workstation. =A0So here I go:<br>
<br>
=A0 $ ssh -R11111:localhost:11111 service-node qsub start_pvserver.sh<br>
<br>
Oops, the qsub command returns immediately and closes my ssh tunnel. =A0Let=
's pretend that the scheduler doesn't provide an easy way to keep t=
he command alive, so I have resorted to using 'sleep 1d'. =A0So her=
e I go, using -tt to prevent orphans:<br>
<br>
=A0$ ssh -tt -R11111:localhost:11111 service-node "qsub start_pvserve=
r.sh && sleep 1d"<br>
<br>
Well, this will only work if GatewayPorts is enabled in sshd_config on serv=
ice-node. =A0If GatewayPorts is not enabled, the ssh tunnel will only accep=
t connections from localhost, it will not accept a connection from compute-=
node. =A0We can ask the sysadmin to enable GatewayPorts, or we could use po=
rtfwd. =A0You can run portfwd on service-node to forward port 22222 to port=
11111, then have compute-node connect to service-node:22222. =A0So your jo=
b script would launch pvserver like this:<br>
<br>
=A0pvserver -rc -ch=3Dservice-node -sp=3D22222<br>
<br>
Problem solved! =A0Also convenient, we can use portfwd to replace 'slee=
p 1d'. =A0So the final command, executed by paraview client:<br>
<br>
=A0ssh -tt -R 11111:localhost:11111 service-node "qsub start_pvserver=
.sh && portfwd -g -c fwd.cfg"<br>
<br>
Where fwd.cfg contains:<br>
<br>
=A0tcp { 22222 { =3D> localhost:11111 } }<br>
<br>
<br>
Hope this helps!<br>
<br>
Pat<br>
<br></div><div><div></div><div class=3D"h5">
On Fri, Feb 12, 2010 at 7:06 PM, burlen <<a href=3D"mailto:burlen.loring=
@gmail.com" target=3D"_blank">burlen.loring at gmail.com</a> <mailto:<a hre=
f=3D"mailto:burlen.loring at gmail.com" target=3D"_blank">burlen.loring at gmail.=
com</a>>> wrote:<br>
<br>
<br>
=A0 =A0 =A0 =A0Incidentally, this brings up an interesting point about<br>
=A0 =A0 =A0 =A0ParaView with client/server. =A0It doesn't try to clean=
up it's<br>
=A0 =A0 =A0 =A0child processes, AFAIK. =A0For example, if you set up this =
ssh<br>
=A0 =A0 =A0 =A0tunnel inside the ParaView GUI (e.g., using a command inste=
ad<br>
=A0 =A0 =A0 =A0of a manual connection), and you cancel the connection, it<=
br>
=A0 =A0 =A0 =A0will leave the ssh running. =A0You have to track down the s=
sh<br>
=A0 =A0 =A0 =A0process and kill it yourself. =A0It's minor thing, but =
it can<br>
=A0 =A0 =A0 =A0also prevent future connections if you don't realize th=
ere's a<br>
=A0 =A0 =A0 =A0zombie ssh that kept your ports open.<br>
<br>
=A0 =A0I attempted to reproduce on my kubuntu 9.10, qt 4.5.2 system, with<=
br>
=A0 =A0slightly different results, which may be qt/distro/os specific.<br>
<br>
=A0 =A0On my system as long as the process ParaView spawns finishes on<br>
=A0 =A0its own there is no problem. That's usually how one would expec=
t<br>
=A0 =A0things to work out since when the client disconnects the server<br>
=A0 =A0closes followed by ssh. But, you are right that PV never<br>
=A0 =A0explicitly kills or otherwise cleans up after the process it<br>
=A0 =A0starts. So if the spawned process for some reason doesn't finis=
h<br>
=A0 =A0orphan processes are introduced.<br>
<br>
=A0 =A0I was able to produce orphan ssh processes, giving the PV client a<=
br>
=A0 =A0server start up command that doesn't finish. eg<br>
<br>
=A0 =A0 =A0ssh ... pvserver ... && sleep 100d<br>
<br>
=A0 =A0I get the situation you described which prevents further<br>
=A0 =A0connection on the same ports. Once PV tries and fails to connect<br=
>
=A0 =A0on th eopen ports, there is crash soon after.<br>
<br>
=A0 =A0I filed a bug report with a patch:<br>
=A0 =A0<a href=3D"http://www.paraview.org/Bug/view.php?id=3D10283" target=
=3D"_blank">http://www.paraview.org/Bug/view.php?id=3D10283</a><br>
<br>
<br>
<br>
=A0 =A0Sean Ziegeler wrote:<br>
<br>
=A0 =A0 =A0 =A0Most batch systems have an option to wait until the job is<=
br>
=A0 =A0 =A0 =A0finished before the submit command returns. =A0I know PBS u=
ses<br>
=A0 =A0 =A0 =A0"-W block=3Dtrue" and that SGE and LSF have simil=
ar options (but<br>
=A0 =A0 =A0 =A0I don't recall the precise flags).<br>
<br>
=A0 =A0 =A0 =A0If your batch system doesn't provide that, I'd reco=
mmend<br>
=A0 =A0 =A0 =A0adding some shell scripting to loop through checking the qu=
eue<br>
=A0 =A0 =A0 =A0for job completion and not return until it's done. =A0T=
he sleep<br>
=A0 =A0 =A0 =A0thing would work, but wouldn't exit when the server fin=
ishes,<br>
=A0 =A0 =A0 =A0leaving the ssh tunnels (and other things like portfwd if y=
ou<br>
=A0 =A0 =A0 =A0put them in your scripts) lying around.<br>
<br>
=A0 =A0 =A0 =A0Incidentally, this brings up an interesting point about<br>
=A0 =A0 =A0 =A0ParaView with client/server. =A0It doesn't try to clean=
up it's<br>
=A0 =A0 =A0 =A0child processes, AFAIK. =A0For example, if you set up this =
ssh<br>
=A0 =A0 =A0 =A0tunnel inside the ParaView GUI (e.g., using a command inste=
ad<br>
=A0 =A0 =A0 =A0of a manual connection), and you cancel the connection, it<=
br>
=A0 =A0 =A0 =A0will leave the ssh running. =A0You have to track down the s=
sh<br>
=A0 =A0 =A0 =A0process and kill it yourself. =A0It's minor thing, but =
it can<br>
=A0 =A0 =A0 =A0also prevent future connections if you don't realize th=
ere's a<br>
=A0 =A0 =A0 =A0zombie ssh that kept your ports open.<br>
<br>
<br>
=A0 =A0 =A0 =A0On 02/08/10 21:03, burlen wrote:<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0I am curious to hear what Sean has to say.<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0But, say the batch system returns right away after =
the job<br>
=A0 =A0 =A0 =A0 =A0 =A0is submitted,<br>
=A0 =A0 =A0 =A0 =A0 =A0I think we can doctor the command so that it will l=
ive for<br>
=A0 =A0 =A0 =A0 =A0 =A0a while<br>
=A0 =A0 =A0 =A0 =A0 =A0longer, what about something like this:<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0ssh -R XXXX:localhost:YYYY remote_machine<br>
=A0 =A0 =A0 =A0 =A0 =A0"submit_my_job.sh && sleep<br>
=A0 =A0 =A0 =A0 =A0 =A0100d"<br>
<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0pat marion wrote:<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Hey just checked out the wiki page, nice! O=
ne<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0question, wouldn't this<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0command hang up and close the tunnel after =
submitting<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0the job?<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0ssh -R XXXX:localhost:YYYY remote_machine s=
ubmit_my_job.sh<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Pat<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0On Mon, Feb 8, 2010 at 8:12 PM, pat marion<=
br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<<a href=3D"mailto:pat.marion at kitware.co=
m" target=3D"_blank">pat.marion at kitware.com</a> <mailto:<a href=3D"mailt=
o:pat.marion at kitware.com" target=3D"_blank">pat.marion at kitware.com</a>><=
br></div></div>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<mailto:<a href=3D"mailto:pat.marion at kit=
ware.com" target=3D"_blank">pat.marion at kitware.com</a><div class=3D"im"><br=
>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<mailto:<a href=3D"mailto:pat.marion at kit=
ware.com" target=3D"_blank">pat.marion at kitware.com</a>>>> wrote:<b=
r>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Actually I didn't write the notes at th=
e <a href=3D"http://hpc.mil" target=3D"_blank">hpc.mil</a><br></div>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<<a href=3D"http://hpc.mil" target=3D"_b=
lank">http://hpc.mil</a>> <<a href=3D"http://hpc.mil" target=3D"_blan=
k">http://hpc.mil</a>><div><div></div><div class=3D"h5"><br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0link.<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Here is something- and maybe this is the pr=
oblem that<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Sean refers<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0to- in some cases, when I have set up a rev=
erse ssh<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0tunnel from<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0login node to workstation (command executed=
from<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0workstation) then<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0the forward does not work when the compute =
node<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0connects to the<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0login node. However, if I have the compute =
node<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0connect to the<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0login node on port 33333, then use portfwd =
to forward<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0that to<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0localhost:11111, where the ssh tunnel is li=
stening on<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0port 11111,<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0it works like a charm. The portfwd tricks i=
t into<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0thinking the<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0connection is coming from localhost and all=
ow the ssh<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0tunnel to<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0work. Hope that made a little sense...<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Pat<br>
<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0On Mon, Feb 8, 2010 at 6:29 PM, burlen<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<<a href=3D"mailto:burlen.loring at gmail.c=
om" target=3D"_blank">burlen.loring at gmail.com</a> <mailto:<a href=3D"mai=
lto:burlen.loring at gmail.com" target=3D"_blank">burlen.loring at gmail.com</a>&=
gt;<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<mailto:<a href=3D"mailto:burlen.loring@=
gmail.com" target=3D"_blank">burlen.loring at gmail.com</a><br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<mailto:<a href=3D"mailto:burlen.loring@=
gmail.com" target=3D"_blank">burlen.loring at gmail.com</a>>>> wrote:=
<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Nice, thanks for the clarification. I am gu=
essing that<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0your<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0example should probably be the recommended =
approach rather<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0than the portfwd method suggested on the PV=
wiki. :) I<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0took<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0the initiative to add it to the Wiki. KW le=
t me know<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0if this<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0is not the case!<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<a href=3D"http://paraview.org/Wiki/Reverse=
_connection_and_port_forwarding#Reverse_connection_over_an_ssh_tunnel" targ=
et=3D"_blank">http://paraview.org/Wiki/Reverse_connection_and_port_forwardi=
ng#Reverse_connection_over_an_ssh_tunnel</a><br>
<br>
<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Would you mind taking a look to be sure I d=
idn't miss<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0anything<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0or bollix it up?<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0The sshd config options you mentioned may b=
e why your<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0method<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0doesn't work on the Pleiades system, ei=
ther that or<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0there is a<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0firewall between the front ends and compute=
nodes. In<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0either<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0case I doubt the NAS sys admins are going t=
o<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0reconfigure for<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0me :) So at least for now I'm stuck wit=
h the two hop ssh<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0tunnels and interactive batch jobs. if ther=
e were<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0someway to<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0script the ssh tunnel in my batch script I =
would be<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0golden...<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0By the way I put the details of the two hop=
ssh tunnel<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0on the<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0wiki as well, and a link to Pat's <a hr=
ef=3D"http://hpc.mil" target=3D"_blank">hpc.mil</a><br></div></div>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<<a href=3D"http://hpc.mil" target=3D"_b=
lank">http://hpc.mil</a>> <<a href=3D"http://hpc.mil" target=3D"_blan=
k">http://hpc.mil</a>><div><div></div><div class=3D"h5"><br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0notes. I don't dare try to summarize th=
em since I've never<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0used portfwd and it refuses to compile both=
on my<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0workstation<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0and the cluster.<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Hopefully putting these notes on the Wiki w=
ill save future<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0ParaView users some time and headaches.<br>
<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Sean Ziegeler wrote:<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Not quite- the pvsc calls ssh with both the=
tunnel options<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0and the commands to submit the batch job. Y=
ou don't even<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0need a pvsc; it just makes the interface fa=
ncier. As long<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0as you or PV executes something like this f=
rom your<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0machine:<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0ssh -R XXXX:localhost:YYYY remote_machine s=
ubmit_my_job.sh<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0This means that port XXXX on remote_machine=
will be the<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0port to which the server must connect. Port=
YYYY (e.g.,<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A011111) on your client machine is the one on=
which PV<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0listens. You'd have to tell the server =
(in the batch<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0submission script, for example) the name of=
the node and<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0port XXXX to which to connect.<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0One caveat that might be causing you proble=
ms, port<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0forwarding (and "gateway ports" i=
f the server is running<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0on a different node than the login node) mu=
st be enabled<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0in the remote_machine's sshd_config. If=
not, no ssh<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0tunnels will work at all (see: man ssh and =
man<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0sshd_config). That's something that an =
administrator<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0would need to set up for you.<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0On 02/08/10 12:26, burlen wrote:<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0So to be sure about what you're saying:=
Your .pvsc<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0script ssh's to the<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0front end and submits a batch job which whe=
n it's<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0scheduled , your batch<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0script creates a -R style tunnel and starts=
pvserver<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0using PV reverse<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0connection. ? or are you using portfwd or a=
second ssh<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0session to<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0establish the tunnel ?<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0If you're doing this all from your .pvs=
c script<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0without a second ssh<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0session and/or portfwd that's awesome! =
I haven't been<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0able to script<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0this, something about the batch system prev=
ents the<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0tunnel created<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0within the batch job's ssh session from=
working. I<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0don't know if that's<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0particular to this system or a general fact=
of life<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0about batch systems.<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Question: How are you creating the tunnel i=
n your<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0batch script?<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Sean Ziegeler wrote:<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Both ways will work for me in most cases, i=
.e. a<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0"forward" connection<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0with ssh -L or a reverse connection with ss=
h -R.<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0However, I find that the reverse method is =
more<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0scriptable. You can<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0set up a .pvsc file that the client can loa=
d and<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0will call ssh with<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0the appropriate options and commands for th=
e<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0remote host, all from the<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0GUI. The client will simply wait for the re=
verse<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0connection from the<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0server, whether it takes 5 seconds or 5 hou=
rs for<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0the server to get<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0through the batch queue.<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Using the forward connection method, if the=
server<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0isn't started soon<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0enough, the client will attempt to connect =
and<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0then fail. I've always<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0had to log in separately, wait for the serv=
er to<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0start running, then<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0tell my client to connect.<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0-Sean<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0On 02/06/10 12:58, burlen wrote:<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Hi Pat,<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0My bad. I was looking at the PV wiki, and<b=
r>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0thought you were talking about<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0doing this without an ssh tunnel and using<=
br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0only port forward and<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0paraview's --reverse-connection option =
. Now<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0that I am reading your<br></div></div>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<a href=3D"http://hpc.mil" target=3D"_blank=
">hpc.mil</a> <<a href=3D"http://hpc.mil" target=3D"_blank">http://hpc.m=
il</a>> <<a href=3D"http://hpc.mil" target=3D"_blank">http://hpc.mil<=
/a>> post I see<div class=3D"im">
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0what you<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0mean :)<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Burlen<br>
<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0pat marion wrote:<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Maybe I'm misunderstanding what you mea=
n<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0by local firewall, but<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0usually as long as you can ssh from your<br=
>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0workstation to the login node<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0you can use a reverse ssh tunnel.<br>
<br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0___________________________________________=
____<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Powered by <a href=3D"http://www.kitware.co=
m" target=3D"_blank">www.kitware.com</a> <<a href=3D"http://www.kitware.=
com" target=3D"_blank">http://www.kitware.com</a>><br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<<a href=3D"http://www.kitware.com" targ=
et=3D"_blank">http://www.kitware.com</a>><br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Visit other Kitware open-source projects at=
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<a href=3D"http://www.kitware.com/opensourc=
e/opensource.html" target=3D"_blank">http://www.kitware.com/opensource/open=
source.html</a><br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Please keep messages on-topic and check the=
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0ParaView Wiki at:<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<a href=3D"http://paraview.org/Wiki/ParaVie=
w" target=3D"_blank">http://paraview.org/Wiki/ParaView</a><br>
<br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Follow this link to subscribe/unsubscribe:<=
br>
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0<a href=3D"http://www.paraview.org/mailman/=
listinfo/paraview" target=3D"_blank">http://www.paraview.org/mailman/listin=
fo/paraview</a><br>
<br>
<br>
<br>
<br>
<br>
<br>
</div></blockquote>
<br>
</blockquote></div><br>
--00c09f8fe5d9990ad1048573fe15--
More information about the ParaView
mailing list