I'll look into that.
Gonna try and set it up tonight!
1 Like
So far I have Make installed, as well as gcc-libs and libc. I cloned the project, cd into the directory, and did "make install" and get argp.h no such file or directory".
I'm fairly advanced at getting things to work, but am perhaps not as technical as what you need.
Consider this as feedback from a general member of the forum with moderate technical chops, who wants to help.
My experience this evening is that the readme on github doesn't really give me much to go on. If I do "make install" and get no errors, what is the expected result? What should I see in my terminal if anything?
I installed the requirements on a Windows machine...Is that even compatible? May be worth mentioning in the readme what platform is expected and has been tested. If Windows/Linux/Mac are all usable, then there should be some platform specific docs.
Overall I think a summary would help.
"This is a TF-cluster project that if you succesfully compile on your X computer, will run in the background and cluster data from such and such source".
If from this feedback I'm obviously way too dense to be any help, then just ignore me, lol.
I tried the git zip download in a VM and it failed.....
Im not super tech but there is only so far I go to help.
This needs to be made accessible and usable by people far less computer savvy than yourself. I'll be making revisions as I can to fix these issues.
However, I should note that you can just 'make' to get the executable....on linux. None of the design has really taken into consideration running on a non-unix platform since the vast majority of research machines at this point run linux.
Thank you for you time and feedback! I'll post in here again when I push out the fixes.
git zip? I have no idea what you tried. If you could, please tell me your entire thought process, so I may work on the documentation.
Ok cool. I was a little skeptical of getting it to work on Windows, but just went ahead and tried. I will give it another shot on my Linux machine and let you know how it goes.
UPDATED:
Switched over to Linux (Ubuntu 16.04 LTS), cloned, cd into directory, then ran command "make install". Got almost to the end where it tried to run "install -C -t /usr/bin tf-cluster" and got a permission denied error (trying to create /usr/bin/tf-cluster). So I tried again as sudo make install, it re ran install -C -t /usr/bin/tf-cluster immediately, no error. So now the question is just, what is the next step to testing this out?
Follow the Source here link to the git site then tiop right on the files is download zip of files....It pulled the site down and stuck in in a VM to install.
I don't think you are using the correct process here.
Inside an already working linux install, use 'git clone' to pull down the repository, cd into the repo, and use 'make'.
I agree Im not 100% on tech....Ill spin up a new VM and try again tomorrow
On Linux if definitely looks like I've got it. I'm assuming that just need to launch tf-cluster, but judging from the error message I get, it looks like I need to pass in a "correlation file" as an argument. What file would that be in the project?
run.sh contains an example run. Also, if you pass '--help' it'll give you usage information.
Ok cool, I did run.sh --help and got an output, so that is cool. Says it took %255 percent of my CPU and took 13.72 seconds. What's the calculation for CPU percentage that gives a result over %100?
As next steps, would it make sense for me to try and get run.sh to run repeatedly? The summary output said no socket messages were sent or received, and there was no file system output. Does this mean you're not getting a log of the work my CPU did?
LMK what next steps help the most! Looking through the code base might take a few days though, not sure how fast I could get up to speed on that.
The '--help' is only meaningful when passed to the tf-cluster program. run.sh is a script which gives an example usage. The resulting CPU usage over 100% is where 100% fully uses one CPU core, and since it is multithreaded it can use more than one core and so gets more than 100% CPU. Getting run.sh to run repeatedly will not do anything interesting or useful. A log from this kind of program should be done by redirecting stdout and stderr. This allows for pipelining programs. As far as for what I should work on, I already know enough now to keep me busy for awhile. I just need to learn how to set up a compute grid to run this program, so almost a voluntary botnet :P