EDIT: I realized this sounded like the new feature that allows multiple slaves on the same machine and found the documentation has an explanation for exactly this situation. I added the “MultipleSlavesEnabled=False” line to the deadline.ini file in the node’s client installation but it only helped half way. I now only have one node show up in the monitor, but that node is still creating two slave instances and checking out 2 licenses when started as part of starting deadlinelauncher. Is there a way to enable automatic slave creation within deadlinelauncher but not have it start multiple slaves?
I’m having an odd problem with the launcher, but it could very well be due to having an odd setup of my render slaves.
Setup: I’m booting Linux slaves (currently only 2 total) from a PXE image stored on my server so that the slaves can run without a hard disk (less power, heat, central management) That also means that they share the same root filesystem, since that lives on an NFS share on the server, and I suspect this may be creating the problems. The machines are called node1 and node2, and have different IP addresses successfully assigned to them.
Problem: When I start “deadlinelauncher -nogui” with only one slave machine turned on out of two, deadlinelauncher creates 2 slave processes on that one machine (when I check with ‘ps aux’ command), and deadline monitor picks up 2 slaves named Node1 and Node2, although both have the same IP address and machine name in the information list. It is also using two licenses.
-
If I then turn on / start the launcher on the second node, it never picks up jobs or registers with the monitor, I assume because deadline thinks it already has, or it has run out of licenses.
-
The odd part is that if I shutdown deadlinelauncher on all the nodes and run just “deadlineslave -nogui” it all works perfectly - each machine gets its own slave, IP address shows up right in the monitor, etc. The disadvantage of that is that I lose all the benefits of having the launcher running.
-
Even if I turn of the ‘Start slave at startup’ option in the deadline.ini file, one machine’s launcher instance seems to intercept messages for both physical machines. Although with that setup it does successfully launch only one instance when I request it in the monitor on my workstation.
-
I tried creating a different user account for each node so that deadlinelauncher was being started by different users, but that didn’t seem to change anything
I hope that was clear since it’s kind of a weird setup. I’d like to keep the linux nodes running diskless, though if that just isn’t a possibility, I can switch back to a more traditional setup.
thanks for any help that can be offered.
- Eric