1.78.2 slamming my array (Windows)

What else is planned?
There seem to be several filewalkers in place:

I am running into massive reads on startup which seems to be caused by the TrashChore:

And I don’t really see a reason why this is handled the way it is. As already said, with nodes getting bigger traversing files is getting more and more expensive and I don’t see a reason why we need to run through the trash files on every node startup.
What is even worse is that there is no config setting to control the trash traversing task, so it runs whenever it wants to run at whatever priority it chooses. That’s also not a good implementation.
To me it seems that for the trash it would be sufficient to collect the modification times once. This should be done when the piece is moved into the trash. This can be stored in a database. From that modification date then we can calculate when to finally delete the piece. Also it is sufficient to run this only once a day to find out which pieces to delete on the current day and then delete the pieces over the day as background task with low priority.
This is my idea of a better implementation.
And it should also include the temp folder, as it seems that massive amounts of data can get pile up there from the partial files there without ever deleted automatically:

1 Like