AWS Thinkbox Discussion Forums

No route to host (slaves)

My slaves are being detected and they are rendering successfully. But trying to connect to them or executing any “remote” commands returns “No route to host”. Obviously the Monitor is unable to connect to the slave, but I have no idea exactly what is the requirement to do so. My nodes are on different subnet than the workstations and usually I access them via ssh. So what’s the deal with Deadline?

This is actually going to be a networking problem. It means that the TCP connection you’re trying to make didn’t even make it to the machine in question, and there are some generic tricks that seem good over here:

maketecheasier.com/fix-no-r … ror-linux/

To add colour to the guide above, Deadline is trying to connect to the machine using the hostname it has on record (check the ‘Machine Name’ column in the Slave list). If they’re on a different network, I’m expecting that the packets aren’t making it through but if SSH is working maybe it’s a DNS issue?

Do you use the raw IP addresses when you SSH in? If so, you can either set up DNS to map those IPs to the machine names Deadline is trying to connect to (preferred since it can be a network-wide change) or you could override Deadline to use the IP addresses.

[attachment=0]2018-03-26 09_09_27-Slave Properties.png[/attachment]

You’re right. My local DNS doesn’t resolve the hostnames of the slaves.
I tried the second option - overriding the IP address. Seems to be working. Also, disabled the firewall on the slaves.

edit: works without adding the IPs manually as well. Seems like the firewall was the issue.

What kind of messaging do you think would have helped you dig into that? It’s not a terribly uncommon problem, so I’d like to spend some time designing a way to help users figure out what that means.

I don’t know if an ICMP ping message would have at least given an idea. Maybe just a better error message when it happens…

Hi mate.
I’m glad that you are open to improving the way users can troubleshoot issues.
For me, just adding somewhere in the documentation (if it is already added, then my bad) that firewall rule or turning firewall off may be needed for this particular problem would do it.
For a specific message I’d go the simple way - something like “Check network connectivity or firewall” on the slave. That seems way easier to figure out than just “No route” :smiley: Cheers!

Privacy | Site terms | Cookie preferences