unable to lock shared memory data lock

[Adrian Wyer]

getting this error note

new 9,1.26 install, trying to render in houdini 21.0.729 with rs 2026.7.0

maya seemed to work just fine, but since then the client app just keeps restarting....

any ideas?

[Adrian Wyer]

farm/clients restart 'seems' to  have fixed it for now....

[Adrian Wyer]

ok, more in the ongoing saga....

got a second machine updated... it picks up the houdini/rs job in the farm 

in viewer it shows that the new machine (fluid001) is rendering frames, but when the frame completes, the label changes to fluid002 (this machine is currently disabled)

small issue, but will be a problem in future when trouble shooting renders

[RR, Schoenberger]

Hi

Lock Error:

The lock error might be caused by thew GPU info app.
But if there is an incompatibility, then the GPU Info should be restarted and not the rrClient.
The issue will be investigated.

 

Client name:

I have to take a look at the job.

Please select the job (or multiple) in rrControl.

Then execute "Export debug info - selected job(s)" via the menu "Debug".

Please upload via www.RoyalRender.de/upload/

 

 

 

regards,
Holger Schönberger

Craftsman and Keeper of the Royal Render Flame

Please use the rrKnights Tavern
or our support system for new questions.

 

[RR, Schoenberger]

Hi

 

The lock issue  should not happen any more.
It was related to a log message in rrClientWatch, which should not show a dialog, but did.
The main download was updated.
Quickfix: www.RoyalRender.de/download/RoyalRender__9.1.26__updateFrom_9.1.26+.zip

[Adrian Wyer]

having updated more machines, we're noticing that the clients are constantly restarting, switching between demo/licensed.. losing connection to server then restarting, then sitting idle, then repeating the sequence

it's not job specific as there's currently nothing really in the farm

what log files do you need me to export to see why they are constantly restarting?

in the meantime i'll update from the quickfix

thanks

[Adrian Wyer]

also the earlier issue with jobs in the viewer reporting having been rendered by the wrong machine is still happening, have uploaded a job debug for that one

[Adrian Wyer]

ok, latest, the new quickfix seems to have sorted the constant restarting of clients and the LOCK error

i have 2 machines out of 7 that will not pick up jobs, they are installed/configured and appear in the control app

the weird thing is, if a submit a job to all the clients, the job shows up in control UNASSIGNED to those two machines....

i figured it could be something to do with the machine list in config, so uninstalled the client from both, and removed the machines from the client list in config

then reinstalled the client app.. waited until the machine again showed up in the farm, but again it won't let me assign jobs

[RR, Schoenberger]

Hi

> but again it won't let me assign jobs
Two possible reasons:
The render app is not listed for the client.
It is a GPU job,  but the client has not found the GPU.

You can verify both in rrConfig, menu Clients, tab Per Client.
On the first tab on the right you can see the GPU count.
On the last tab the installed render apps.

[Adrian Wyer]

thanks!! it was the GPUs... both machines have 2 but RR config said 0... set it to two and the machines picked up straight away!

great stuff

a

[Adrian Wyer]

any thoughts on how to get the farm to recognise the gpus and clock them ?

noticed that even if i force it to use 2 gpus on a machine, the GPU Speed field is zero...

a

[RR, Schoenberger]

Hi

 

Which machines have the issue?
Then I can take a look in the app logs that you send with the Job debug export.

Which GPUs do they have?

 

 

Note:

The new application that should detect all GPUs can be run via

%RR_ROOT%/bin/win64/rrGpuInfo -v -single

[RR, Schoenberger]

GPU issue:
I found it in the logs.
Machine no 006.
It was a problem that it tried to benchmark the GPU while another process was trying to retrieve the GPU usage at the same time. Will be fixed.

 

Houdini client name issue.
The job was reset, therefore I do not have access to all information any more.
At the moment it states that frames have been assigned to Fluid005.

Do you know if the issue Fluid001/Fluid001 happened after an older reset?
If it was after the reset, then the issue was perhaps that some old information was not reset.

[Adrian Wyer]

good to know about the gpu issue, will run the rrgpuinfo when nothing else is running

will keep an eye on renders and see if i can replicate the client name issue, nothing rendering currently and the farm has been swept clean of logs/jobs

will report back next week

thanks!

[Adrian Wyer]

i can manually set the number of GPUs in the config app, but after a while of rendering it get's reset to 0 on 3 of the machines...

i ran gpu info   ( %RR_ROOT%/bin/win64/rrGpuInfo -v -single) on one machine, it reported 2 gpus, but didn't affect the number of gpus reported in config (does it take some time to update?)

for now i'll have to open config periodically and update the problem machines back to 2 gpus manually

[RR, Schoenberger]

Hi

 

Please zip and upload the folder RR/sub/log:

www.RoyalRender.de/upload/

 

Then I can take a look at the log files.
 

[Adrian Wyer]

done, thank you sir!

a

[RR, Schoenberger]

Hi

 

The rrClient was unable to execute the rrGpuInfo executable.

Please apply this update:

www.RoyalRender.de/download/RoyalRender__9.1.26a__updateFrom_9.1.26+.zip

 

 

While checking the log files I found another issue.

The rrClient was not able to load rendered files to verify that they are written correctly.

Can you export a debug info of the job {8P5M}  via rrControl and upload it as well?

[Adrian Wyer]

sorry i was away for a couple of days and someone flushed the old jobs from the farm...

will apply the patch later today....

as for the loading rendered files, would that have anything to do with the thumbnails not working in the job info panel? i believe it might be because 

of the icio profile in the exrs... in maya some jobs have to be set to LEGACY color management, and i think this causes RR to not read the files;

[RR, Schoenberger]

Hi

 

> loading rendered files, would that have anything to do with the thumbnails not working in the job info panel?

Yes, this message could be caused by the error I have seen in job {8P5M}.

It should not be an OCIO issue, in that case the image would be without OCIO transform applied.
The message was “unable to read file”. As if it did not exist on your fileserver.

[Adrian Wyer]

finally got around to updating RR ( 9.1.26a)

seems like a couple of the clients still think they have no GPU

tried running workstation installer on one thinking that might get it to acknowledge the GPU, but not made a difference

any ideas?

will it have saved a config for that machine stating it has no GPUs somewhere that needs a refresh/delete?

do i need to restart the farm/client for the new update to take effect?

thanks

a

[RR, Schoenberger]

Hi

 

Perhaps we should do a remote session that I can take a look?

If you send an email to support@RoyalRender, then we can arrange a date.

[Adrian Wyer]

thanks for your help this morning... good to have everything working (aside from thumbnails, but we'll live with that :)  )

a

[RR, Schoenberger]

Thanks for the update.

FYI:
There is a different thread about thumbnails in this rrKnights Tavern (22.6.).
It might solve your issue as well once we have a new public version with that fix.