I’m frequently seeing this field reporting something like only 16gigs of average RAM usage, but then when I view the RAM usage on the local slave, it is much much higher. On some machines, I see over 64 gigs of RAM usage, even though the monitor tells me 16gigs. This is causing our artists to think their scene is much lighter than it actually is, and causing some of our machines to start swapping. Can the RAM usage fields be tuned to show actual total RAM usage, or at least something closer?
Thanks!
As it turns out, there is a problem with one of our scenes that is causing the final buckets of the vray render to hang, and the RAM usage just climbs and climbs. I guess since there’s no progress being made on the render, the RAM reporting to the monitor is not happening.
Where are you seeing this? I’m assuming the slave list, since we don’t set the CPU and RAM usage for the tasks in the Monitor until they are complete, but I just want to confirm.
The slaves update their state every 20 seconds or so, and should continue to do so even if the render has hung up.
Cheers,
The render was hung up for about an hour on the last render buckets in vray, with the 3dsmax process continually growing its ram usage to the max (64gigs) and then it would fluctuate around there. After a while, the task does finish and then the average/ peak ram usage is posted to the complete task entry. However, it just shows the 10-12gig ram usage that was the case while the task was mid-frame.
How long was the total render time for the task? Just wondering if that last hour where it grew to 64 gb wasn’t enough to really skew the results. Also, is the peak ram usage being recorded properly?
Nope, peak shows 12-16 gigs or so, that’s why Jon thought its not showing the right results. Its actually stuck for a substantial amount of time relative to the rendertime.
Weird. We’ve logged this as a bug, and we’ll try to figure out what’s going on here.
Thanks!
I’m not having any luck reproducing this strange behavior. The only thing I can think of is that an error is occurring when the slave tries to get the memory usage info from the process, and as a result it doesn’t get factored into the final calculation. Have you seen this behavior with any other jobs besides the ones with the runaway memory?