tomweber-sas
commented
6 years ago @bikashdube Well, I've tried this out myself and I'm not getting any problems. I initially suspected the
ods _all_ close;you were submitting, as that would interfere with how saspy works with regard to getting output; it uses specific ODS statements to get and return output. However, proc setinit doesn't generate any output, it just writes to the log, which is simply text, so that isn't happening to be a problem for this code you submitted. I wouldn't submit that when running any code you expect to get output from though, as that would keep you from getting any ouput. Also, the 'duplicate' options, saspy's STDIO access method requires specific SAS system startup options to be able to work. Those options are submitted by saspy to run SAS with this access methods. You happen to put them into your options key also, which isn't needed. The other options like saswork and memsize, fullstimer... things like that are what the options: key is for. The specific startup options
-nodms -stdio -terminal -nosyntaxcheck -pagesize MAXwill be applied by saspy.
So, that gets us back to what is the problem here? One thing I'm noticing trying to run the sas_kernel is that it's taking a long time to start. I don't know why, but could that be what's happening? How long are you waiting for it to run?
But, let's do something proactive to see what's happening. We can't get much diagnostics from the sas_kernel, so let's try the python kernel and saspy directly. So 2 thisgs to start. Did you run
/usr/bin/ssh -t sashost.domain.com /opt/sas -WORK /opt/Default_saswork_v9 -noterminal -nodms -nosyntaxcheck -pagesize MAX -fullstimer -nofmterr -memsize 3072M -nodms -stdio -terminal -nosyntaxcheck -pagesize MAXfrom the command line to see if there's actually something keeping this from working?
The other thing to try is to just run these same things using saspy and then we can look at the whole saslog and see if there's anything suspicious, assuming you are successfully connecting to SAS; the run from the command like would tell us that.
I've uploaded 2 notebooks, one sas_kernel and the other saspy, to show both of these doing the proc setinit and a proc print, with and without the ods all close to show that that isn't causing a problem, but it will keep you from getting output.
These are in the saspy-example repo in the issues folder: https://github.com/sassoftware/saspy-examples/blob/master/Issue_examples/Issue176_SAS_kernel.ipynb https://github.com/sassoftware/saspy-examples/blob/master/Issue_examples/Issue176_saspy.ipynb
I'm not having much luck with github rendering these notebooks. Sometime it does, but sometimes is doesn't. You can download any of these notebooks on that site and then upload them into your jupyter and they should resolve to see them, and you can also then run them then also.
The saspy version is simply running the following code, in case you can't get it to render either:
import saspy
sas = saspy.SASsession()
sas
ll = sas.submit('''
ods _all_ close;
proc setinit;
run;
''')
print(ll['LOG'])
ll = sas.submit('''
ods _all_ close;
proc print data=sashelp.cars(obs=5);run;
''')
print(ll['LST'])
len(ll['LST'])
ll = sas.submit('''
proc print data=sashelp.cars(obs=5);run;
''')
print(ll['LOG'])
from IPython.display import HTML
HTML(ll['LST'])To see the full log in saspy, you use the saslog() method of the SASsession object:
sas = saspy.SASsession()
print(sas.saslog())Can you try out these things and see what that shows?
Thanks, Tom
bikashdube
@tomweber-sas
So I resurrected an old environment where things used to work fine. Did a pipdate to upgrade all packages in the environment first.
When I do a simple SAS code submit in a SAS Kernel cell, I am having to terminate the session as no response comes back.
Code I am submitting:
Here is the log after I terminate and confirm termination.
I am not sure where the duplicate options are coming after the 3072M option. Here is the full sascfg_personal.py
When I run the ssh command in terminal, I get a successful response. So the ssh keys etc are all setup correctly.
Here is a sample response (partial shown)
Similar error occurs when I try this in a Python3 kernel notebook as well.