VRay resumable render file failing after power outage

[Art48]

Hi,

We recently rendered an image which took about 4 hours to complete (so no big deal) and when it was 90% completed we had a total power outage.

I thought: Whelp - good that there is resumable rendering! This was run via backburner and when I restarted the job everything went normal when starting up until the point where VRay loads the resumable file. It just sat there and it did SOMETHING, but apparently it failed. There were RAM fluctuations and 8% CPU usage (meaning one core did something). RAM-wise it seemed that the image was loaded since it occupied quite some space there, and the fluctuations were generally like constantly getting higher and dropping down a bit inbetween. But generally the RAM usage climbed. I left it running for about 30 minutes and it still wasnt loading any image so I cancelled and re-started the rendering and removed the resumable VRIMG file and it worked fine.

Is this a known issue?
I may be able to provide a "corrupted" VRIMG, if I find the right one on our backup since I deleted it.

[radoslav.platikanov]

Hello,

Can you send me a stripped down version of the scene to radoslav.platikanov@chaosgroup.com? I only need the render settings, you can delete everything else. The .vrimg file will be helpful too if you can get it.

[Art48]

Just sent the mail with the max file contained in it. Also uploading the VRIMG to your FTP right now - ETA 2 hours.

Edit: It may be worth noting that the file output directories are mapped drives on a server in the local network. Backburner works fine with this though since the drive letters are the same on both machines.

[radoslav.platikanov]

Thank you for the files. The image doesn't seem to be corrupted actually. The resuming process itself is rather slow due to the large resolution, the number of render elements and the small bucket size, but it resumes successfully in 2-3 minutes in my tests. My best guess is that it's slow due to the network location of the output file, but I can't reproduce it on our network. You can try resuming the file from a local drive to verify this.
Anyway, the resuming process with a large number of buckets can be optimized, especially with network locations. Sorry for the inconvenience.

[Art48]

Hi,
I will try this, resuming from local drive instead of network. We recently had a case where we resumed a render from a network location. It wasn't a big one and it worked flawlessly. I will let you know how it goes.

[Art48]

I just tested it and when the VRIMG is read/written on a local drives the 2-3 minutes you mentioned seemed about right here, too and it works. I mean we don't have a SAN here but a gigabit ethernet and a NAS with fast HDDs (sadly no SSDs yet) and if stuff doesn't get loaded after nearly 30 minutes on our (not unusual) network configuration I have to say I question the usability of this (awesome) feature for us. We render huge images for print all the time .

Please improve the loading speeds and ideally provide me with a progress bar where I can see how much of the reading process is done. Currently on this case it just hangs at "Preparing direct light manager" for the 2-3 minutes and then starts rendering.

Also did I understand it right that the loading speed is related to the number and thus also the size of the buckets? Would that mean if I increase the bucket size the loading would be faster? Any hard numbers on this (f.e. double the bucket size leads to half the loading time or something similar)?

[radoslav.platikanov]

Thanks for the update. We'll improve the loading speed and add a progress estimation at some point.

About the bucket size - yes, the loading speed is affected by the number of buckets. Larger buckets will improve the performance, but I'm not sure by how much exactly.