Hi,

I notice that after running a job through a Service (ROPfetch service in this case) that the RAM consumed by the Hython instances isn't cleared out.

Is there any way to have that happen? On high RAM jobs this is a bit crippling.

I took a look in the docs but couldn't find anything, though the PDG python docs are comprehensive and I might have missed it.

18.5.596 Win 10 Py 3

Cheers,

Peter B

Hi Peter,

Currently there's no way to tell the service to reset and clear its cache. To work around this at the moment, you would need to restart the service whenever the memory needs to be freed up. We're looking into being able to issue a reset command to the services, however, which will provide a way to clear the contents of the scene file and empty out any caches that the service maintains without needing to restart it.

OK thanks Brandon, no worries, I'm just not using services for now and eating the Hython startup cost, given the length of some of my processes it's bearable for now

Looking forward to the "clear scene" option for services though, sounds like what the doctor ordered for sure.

Cheers,

Peter B

BrandonA
Hi Peter,

Currently there's no way to tell the service to reset and clear its cache. To work around this at the moment, you would need to restart the service whenever the memory needs to be freed up. We're looking into being able to issue a reset command to the services, however, which will provide a way to clear the contents of the scene file and empty out any caches that the service maintains without needing to restart it.

Are there any updates on this? I'm running into the same problem of quickly running out of RAM on a long-runing service processing lots of geometry/work items

Is there anything planned to be able to combine batching/services in some way, where you can respawn a service for every nth work item, or after a certain RAM percentage is hit etc etc

Hello: Ping on this! we have the same issue and question! @Sidefx

"Are there any updates on this? I'm running into the same problem of quickly running out of RAM on a long-runing service processing lots of geometry/work items

Is there anything planned to be able to combine batching/services in some way, where you can respawn a service for every nth work item, or after a certain RAM percentage is hit etc etc "

Thanks!

Vince

vinyvince
Hello: Ping on this! we have the same issue and question! @Sidefx

"Are there any updates on this? I'm running into the same problem of quickly running out of RAM on a long-runing service processing lots of geometry/work items

Is there anything planned to be able to combine batching/services in some way, where you can respawn a service for every nth work item, or after a certain RAM percentage is hit etc etc "

Thanks!

Vince

Vince, did you find a solution? I think it should be possible now with the "Service Reset" node.
I'm trying a setup with a partition by range and custom priorities on the service reset to reset the services every n work items...