I am running Deadline 5.2 on test machines and the slaves keep dropping out after a while. I have the slaves set to Continue Running After Current Task Completion, but the slave still just goes offline after a while of inactivity.
I am only testing it on 2 machines using the Free License mode. Could this be the reason? Is this a thing with the free license mode that they just go offline after a while and you have to restart the deadline slave?
Which OS are you running?
Are you able to log into that slave and check it?
Is this a workstation that is going into sleep mode?
I’ve seen this issue where automatic updates are on in Windows and Windows wants to shut itself down and this kills any renderer
Yeah things are really good here. Was a bit crazy to begin with but has all settled down.
Windows 7 64bit
Yeah the slave is still accessible to connect via remote desktop
No I don’t think so.
Yeah I have had that on our nodes back in london but here that is not the case. The machine does not automatically restart. It stays on. Just the slave closes.
Not sure if there are any logs for that slave that might have information on its status and logs why it closes?
Hi,
Just to clarify, is the deadline slave crashing? Shutting down gracefully or something else?
Please describe what happens! If its the slave shutting down, can you RC the Deadline launcher and provide the slave log reports from that machine? (explore log directory option in the launcher).
Also how many slaves do you have in your monitor slave list? If its more than 2, then the slave might be shutting down due to more than 2 slaves in the list and being in free mode. I assume you have no license?
Mike
To be honest I am not sure if the slave is shutting down gracefully or not. Only once have I logged into the slave and see that it says not responding. I have not seen the Windows program error dialog saying Deadline Slave crashed. The slave just turns off.
There are only 2 machines listed in the slave list. No more.
I am still not sure where this slave log is meant to be or show.
I load Monitor and right click on the slave and go to Slave Reports - View Error reports
This is what it has. Only two of these Stalled Slave Reports but the deadline slave has dropped a lot more times than that.
To be clear I have not experienced the slave dropping during a render. From what I have seen it has only been when the slave has been inactive for a while.
Current House Cleaner Information
Machine Performing Cleanup: rendernode1
Version: v5.2.0.47700 R
Stalled Slave: Rsxxxxtmp
Last Slave Update: 2012-08-27 20:31:33
Current Time: 2012-08-28 09:51:56
Time Difference: 13.34 hrs
Maximum Time Allowed Between Updates: 0 s
Current Job Name:
Current Job ID:
Current Job User:
Current Task Names:
Current Task Ids:
Do not have enough information to identify stalled job/task.
Need at least a job name or a task name.
Attempting to check all active jobs to see if they have this slave rendering a task.
Setting slave’s status to Stalled.
Setting last update time to now.
U think it might be something to do with a machine name conflict.
I just noticed the slave crash because it said it saw another machine with the same name and so deadline thought there were more than 2 slaves in the list so it dropped it.
The issue is that my computer name is called rendernode1 yet deadline is reading the machine name as Rsxxxxtmp even though it is not called that anymore.
I tried reinstalling the repository and client on all machines but the same problem happened.
I have now downgraded back to 4.1 release (what we are using in london studio ans is stable) and it is running fine. Deadline sees the slave as the correct name and it is no longer dropping. Left it over night and the slave is still running (something that never happened with 5.2)
What you’re seeing is that the Slaves are being a little bit more security concious than they should be in license free mode (it happens from time to time). It should show that in the regular Slave logs. Those logs are easier to find than you might think. Just go to the Slave UI, then its “Help” menu, then “Explore log folder”. We do a lot of checks in license free demo mode to make sure people can’t game the system.
To actually get a license for testing, either send sales@thinkboxsoftware.com a message for a 30 day trial, or give me a call at +1-866-419-0283. We’re pretty liberal with the testing licenses.
The multiple name thing is a bug in the 5.x multi-slave mode. We got crafty and put the Slave names on the local machine in “C:\ProgramData\Thinkbox\Deadline\Slaves”, but it turns out that if you change the host name, those names don’t change to reflect that. So! Quick fix is to blow away that ‘Slaves’ folder on the local machines.
Feel free to give me a ring to discuss these sorts of errors too. I’m a bit put out that I didn’t see this thread sooner.
Ah ok. I deleted 5.2 and the repository so do not have those log folders any more. But if I upgrade again I will make note.
Here in NY we actually only have one render node so the license free mode is all I need to really use.
I was testing 5.2 here as it was a lot easier to manage than with all our nodes we have in London. They are still on 4.1 and I was wary about upgrading as “if it ain’t broke, don’t fix it” so thought with the setup in NY it would be a good opportunity to try it.
I do not think we will be getting any more render nodes in NY. The plan is to utilise cloud rendering and we are testing a service out for that.
The one render node we have here is a fall back render node for things that I can render quickly locally.
However I am still interested in testing Deadline here for our London office. What is the situation with Deadline 6.0? Is it in beta yet?
Deadline 6 supports spawning instances on cloud compute clouds such as ec2 directly within the monitor, and when launched get added as slaves which can be managed as normal rendering slaves. our siggraph demo was rendering from the tradeshow floor across local machines in la, ec2 slaves and machines in the Canada office at the same time, on the same job. The demo involved adding additional ec2 slaves live which picked up the job and finished it before the demo was over.
How would it work in terms of licensing on the cloud slaves? Do you set up these instances? Do we pay you for that or is it our responsibility to set up the cloud instances? How will licensing work? Have lots of questions about it.
Was your demo at siggraph recorded at all and available to watch on the web?
we rent licenses by the week and we are reviewing an automated online system to charge by the hour, day, week or whatever makes sense so that clients can scale in this fashion easily. imho, it makes sense to buy more deadline licenses and use them as you want - i dont care if they are local, remote or whatever.
3rd party licensing would be your responsibility.
the system would basically allow you to manage slaves on remote cloud compute resources pretty much in the same way you would use them locally.