FFMPEG Crash! in FFAStrans Manager

[vijm1972]

Hi Steipal,

Crash happens every time on different servers, some times it's happens on same server too.
FFAStrans executable are in Isilon NAS Central storage. We have multiple workflow's on this storage and no network related issues.
No..... FFAStrans crash even after 5/6 hours of working, sometimes they crash while processing jobs.

I have tested the same environment using local drive on one of my server, I have mapped the drive on all the servers and opened FFAStrans executable using same mapped drive from all the servers. I haven't face any crash issue while processing jobs using mapped drive.

How many maximum number of servers I can add to manager? currently I have added 25 servers but only 23 servers process jobs at a time. To confirm that I had dropped 60 source files but only 23 jobs processed at a time. Is that because of CPU roof which I had configured on each servers?

I also configured remove from history option in Configure menu, when I dropped same 60 source files again, out of 60 files 5/6 files not processed. Even I am not getting any error for them.

Thanks
Vijay

[admin]

So, if I understand correctly; FFAStrans will crash only when running from an UNC-path on the Isilon NAS?

There is no built in maximum servers in FFAStrans but it will all probably depend on the server where FFAStrans resides. What are the "CPU roof" and "Core usage" settings on your node servers?

Have you tried to somehow modify the folder where your dropped the 60 files either by adding or remove files?

Also, which version of FFAStrans are you using?

-steipal

[vijm1972]

Yes you understand correctly; FFAStrans will crash only when running from an UNC-path on the Isilon NAS.

On all the servers "CPU roof" is 75 and "Core usage" is 1.

I haven't tried to modify my folder structure. If I rename the same 60 files and reprocess them again all the files processed successfully in 1st attempt. If if I drop them again some files got failed.

FFAStrans Version 0.5.1

[admin]

Ok, do you know if the issues are the same if you limit node servers to f.ex. 3-5?

The "Core usage" directly reflects the amount of simultaneous jobs one node can process. If you set your servers to 2 then you would have up to 46 simultaneous jobs.

I will try and look further into the issue regarding not picking up new files. Maybe I need to add an option to force deep seek for new files, although this would lead to some impact on idle CPU-usage.

-steipal

[vijm1972]

If I run FFAStrans using NAS location on only one server then also FFAStrans get crash.

I have some other application running on same nodes, that's why I kept the "Core usage" 1. Is it possible to process single job using multi core CPU? Other application's are primary applications and that's why I kept the "CPU roof" 75.

Actually I am not dropping new files, re dropping same files again then only some files not process. If I drop new files, all files process successfully.

[Ysh]

Hi, Steipal !

Fist of all I want to express my appreciation for a such awesome product you are working on.

Now we've faced the same problem as a topic starter.
Running FFAStrans in farm mode from a shared folder and getting a

FFAStrans workflow executor
and
FFAStrans processor manager

error c000000d every couple of hours.

Running from a mapped drive didn't solve the problem.

Is there any suggestion how to fix it ?

[admin]

Hi Ysh, thank you for your kind words!

I'm not sure what this error is but you need to try running it from a different computer, and also try running it from a local path, not network. See if it behaves the same way. I know there is sometimes issues when running applications from a network share and I have plans on addressing that in a future release.

-steinar

[Ysh]

Hi, Steipal !

I'm running on 3 different servers, and all this have the same issue.
I tried to run FFAStrans from local path and have no problems, but have an issue that one file is encoded on all 3 servers. And there is a monitoring issue when I run locally. I cannot monitor the farm. I can see only jobs that runs on one server at a time.