As we are using the very useful codem-transcode for our transcoding solution, we recently had some problems using the scratch directory.
In our cluster-setup consisting of 10 nodes, we are using a local scratch-directory /tmp/ on every node, and a shared directory /shared/ for finished transcoded files.
Unfortunately every node signaled a working progress as finished, before it was copied from the local scratch directory to the shared-directory.
I changed this behaviour. In our fork of codem-transcoder, every transcoding-progress ends with a copy of the file from the scratch-directory to the destination folder.
As we are using the very useful codem-transcode for our transcoding solution, we recently had some problems using the scratch directory.
In our cluster-setup consisting of 10 nodes, we are using a local scratch-directory /tmp/ on every node, and a shared directory /shared/ for finished transcoded files.
Unfortunately every node signaled a working progress as finished, before it was copied from the local scratch directory to the shared-directory.
I changed this behaviour. In our fork of codem-transcoder, every transcoding-progress ends with a copy of the file from the scratch-directory to the destination folder.
Have anice day