Context: A team of operators work with large batch files up to 10GB in size in a third party application. Each file contains thousands of images and after processing every 50 images, they hit the save button. The work place has unreliable power and if the power goes out during a save, the entire file becomes corrupt. To overcome this, I am writing a small utility using the FileSystemWatcher to detect saves and create a backup so that it may be restored without the need to reprocess the entire batch.
Problem: The FileSystemWatcher does a very good job of reporting events but there is a problem I can’t pinpoint. Since the monitored files are large in size, the save process takes a few seconds. I want to to be notified once the save operation is complete. I suspect that every time the file buffer is flushed to disk, it triggers an unwanted event. The file remains locked for writing whether or not the a save is in progress so I cannot tell that way.
Creating a backup of the file DURING a save operation defeats the purpose since it corrupts the backed file.
Question:
- Is there a way to use the FileSystemWatcher to be notified after the save operation is complete?
- If not, how else could I reliably check to see if the file is still being written to?
Alternatives: Any alternative suggestions would be welcome as well.
There’s really no direct way to do exactly what you’re trying to do. The file system itself doesn’t know when a save operation is completed. In logical terms, you may think of it as a series of saves simply because the user clicks the Save button multiple times, but that isn’t how the file system sees it. As long as the application has the file locked for writing, as far as the file system is concerned it is still in the process of being saved.
If you think about it, it makes sense. If the application holds onto write access to the file, how would the file system know when the file is in a “corrupt” state and when it’s not? Only the application writing the file knows that.
If you have access to the application writing the file, you might be able to solve this problem. Failing that, you might be able to get something with the last modified date, creating a backup only if the file isn’t modified for a certain period of time, but that is bound to be buggy and unreliable.