Certain source files can cause knitr to prompt and wait for input, and when this happens in a WW make job, for some reason it doesn't terminate after the 5 minutes that is currently the time limit on our wikis.
At least, the initial report I received seemed to say that knitr was prompting for input. When I make a pdf from the problem file on my laptop, it does not prompt, but it does go into an infinite make recursion. Since that should never happen, I'm going to fix that and then see what we have.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I have dealt with the make recursion. That could bring a server to its knees. But pe-make is supposed to be using rlimit to keep that from happening. I will have to look into that.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
OK - now that I have the project on my laptop, and all the utilities installed that it uses (AFAICT), I am finding that it invokes xppaut which prompts and waits. Why this should fail to get terminated, and possibly cause server crashes, I don't know.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
(back on my laptop) just prints the prompt once and gets killed at 30 seconds. It also behaves well when I pipe it into a file and run it in the background, both at home and on the cluster, and also when I use 'sudo -u apache -i' to run it in the background as apache into a file. I am puzzled.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
So, got the project running in my home wiki and it does the repetitive output thing there - so I can use that to diagnose the problem without crashing yushan nodes.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
When it runs amok on my laptop, it gets killed at the time limit, at least if I set the time limit to 20 seconds. I could see having more trouble when it runs for 5 minutes. The log file generated on n06 is 3G.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Oh, and possibly it crashes the server... I waited 5 minutes for it to terminate on n06, and now n06 isn't accepting ssh connections
but is still responding to http requests
At least, the initial report I received seemed to say that knitr was prompting for input. When I make a pdf from the problem file on my laptop, it does not prompt, but it does go into an infinite make recursion. Since that should never happen, I'm going to fix that and then see what we have.
I have dealt with the make recursion. That could bring a server to its knees. But pe-make is supposed to be using rlimit to keep that from happening. I will have to look into that.
OK - now that I have the project on my laptop, and all the utilities installed that it uses (AFAICT), I am finding that it invokes xppaut which prompts and waits. Why this should fail to get terminated, and possibly cause server crashes, I don't know.
When I do it with pe-make on my laptop, it terminates the process at the time limit.
And when I do it with pe-make on n06
it terminates after 30 seconds.
But in thesis.pdf.make.log in the WW directory is a horror, endlessly repeating
"(r)un (c)d (l)ist " is the prompt from xppaut, that doesn't get a response. Maybe xppaut doesn't behave well when its output is being piped?
(back on my laptop) just prints the prompt once and gets killed at 30 seconds. It also behaves well when I pipe it into a file and run it in the background, both at home and on the cluster, and also when I use 'sudo -u apache -i' to run it in the background as apache into a file. I am puzzled.
So, got the project running in my home wiki and it does the repetitive output thing there - so I can use that to diagnose the problem without crashing yushan nodes.
</dev/null does not help
When it runs amok on my laptop, it gets killed at the time limit, at least if I set the time limit to 20 seconds. I could see having more trouble when it runs for 5 minutes. The log file generated on n06 is 3G.
The problem behavior of xppaut can be reproduced outside the wikis by running
where missing.ode is not an actual file.