PDG Deadline - Farm Error

   1763   1   0
User Avatar
Member
1 posts
Joined: March 2020
Offline
I've spent the last couple days trying to get PDG running on my local farm. I've looked through the trouble shooting threads and similar threads with no luck in fixing the issue. When submitting, only the machine that I submit from actually works. The other two machines throw an error. I've also submitted from the other farm boxes that fail, and they work as long as I submit from them. So PDG is properly setup on the other boxes.

All slaves are pointing to a shared folder /home/jcoleman/mnt/afx_sd_01/

OS = Linux Mint 19.3 Tricia
Houdini Version = 18.0.416
Deadline Version = 10.1


Deadline Log Error:
2020-04-22 10:30:21: 0: STDOUT: Traceback (most recent call last):
2020-04-22 10:30:21: 0: STDOUT: File “/home/jcoleman/mnt/afx_sd_01/Sandbox/Jcoleman/Mutagen_Test/pdgtemp/77812/scripts/rop.py”, line 592, in <module>
2020-04-22 10:30:21: 0: STDOUT: args.server = socket.gethostbyname(hostname) + ‘:’ + port
2020-04-22 10:30:21: 0: STDOUT: socket.gaierror: Name or service not known



// Snippet of code from the rop.py, Line 592 error

rop.py line 592:

# cache the ip of the result server on the args to speed up rpc calls
if not args.server:
if not disable_rpc:
hostname, port = os.environ.split(':')
args.server = socket.gethostbyname(hostname) + ‘:’ + port
else:
args.server = ‘'

From what I can see the farm machines can’t connect to the PDG_RESULT_SERVER. Even though earlier in the log it shows:

Process Environment Variable PDG_RESULT_SERVER to AFX-WS-003:40493
PDG_RESULT_SERVER: AFX-WS-003:40493

I've tried as many different options in the Deadline Scheduler as I could. Submit Graph as job, all the different MQ job options (local / Farm), Enabling Data Layer Server, Auto Connect, Copy Plugin to Working Directory, Inherit Local Environment ect… At them moment I have my firewall disabled. And I'm not sure how exactly to trouble shoot this any further.

I've attached the full Deadline Log. Any help would be greatly appreciated.

Attachments:
Job_2020-04-22_10-30-46_5ea05516159c767bb9163dd2.txt (9.8 KB)

User Avatar
Member
571 posts
Joined: May 2017
Offline
I answered you via support, but I'll post here as well:

Could you enable `Use IP Address for PDGMQ` on the TOP Deadline node under the Deadline section? It might be that its not able to resolve the hostname to IP address due to lack of DNS.

Another option is to use our new MQ server. To do so, set PDG_USE_PDGNET=1 in your environment, then launch Houdini. This new feature will use IP addresses directly.
  • Quick Links