Error when trying to start mdce service on hosts for clustering (Parallel Computing)

Hello, I'm new to computer clustering and experienced some difficulties with it:

when I try to add a host in the admin center for the clustering, they all appear as reachable, no problem here. But then I try to start the mdce service for each of them separately (since I have macintosh + windows hosts) and get an error at the end.

mac hosts: the ssh connections seems to work (in the terminal it works definitely), then I got the following at the end:

Unable to write a test file (.mdcetestfile) in PIDBASE directory /var/run. This might be because you do not have adminstrator privileges on cluster machines.

windows hosts:

Unable to start the MDCS Remote Execution Service on host "192.168.2.123", which is required for managing the MDCE service.

Output from sc.exe:
[SC] OpenSCManager Error 5:
Permission denied

Trying to add the host that runs the admin center is no problem, everything works fine and all cores get connected.

Am I missing something? BTW, the user that is used for the ssh is of course a normal system user that has root privileges, but even though I give this user the right to write into that folder stated in the error output, I'm not sure wether I want any user to write in there or not, since it's pretty high level access, isn't it? And I don't know if this is the original source of the problem.

Thank you very much beforehand!

Answer 1

[Jason Ross]

• For the Mac, I believe that /var/run is going to be owned by root and of the group "daemon". If the user account you are logging in as can't write there, you will need to put the logs elsewhere where you can write. You can change the log directories by editing mdce_def.sh, which is in matlabroot/toolbox/distcomp/bin. You'll likely need to change PIDBASE, LOCKBASE, LOGBASE and CHECKPOINTBASE. This will also be moving the logs to a "non-standard" location, which you'll need to remember if you ever need to look at them for some reason.
• On the Windows host, there can be a few things going on -- one can be that you are running with UAC enabled, and that is preventing the service from being installed. Or it's also possible you are using an account that does not have administrative privileges to install the server. You can get around the UAC thing by doing the "mdce install" in an elevated prompt, but if you want to do this remotely, you need to switch off UAC in order to allow remote changes to the system, as there's no way to respond to the elevation prompt.

I'll also note that installation support is available by calling in, too.

I'd also encourage concentrating on the Windows or Mac hosts first, then move onto the other platform.