TLDR: Add the .migrate_chore file mentioned in the original post even for passive migration.
I noticed unexpected behavior with passive migration on my node running v1.120.4. It seems that even if you are not running active migration, the .migrate_chore file needs to be present (even if the file is empty) for passive migration to work correctly. If the file is not present, passive migration will add data to the hash store but it will not remove the old migrated files from the piece store. As a result I believe I now have some duplicate customer data in both the old piece store and inside the new hash store. Hopefully garbage collection will deal with it.
I used the following linux command from inside my blobs folder to monitor the number of files in my piece store as customers were uploading and downloading files:
find . -type f | wc -l
PS: I also have WriteToNew set to true in addition to PassiveMigrate. The number of files in my blobs folder would remain constant until I added the .migrate_chore file(set to false or set to nothing at all), and now the blobs file count decreases slowly as passive migration works correctly as customers download their files.
If I set .migrate_chore to true, it enables active migration. I only want passive migration right now. I have it set to false and it’s still deleting files but in passive mode.
It’s not deleting files. Just saving new in hashstore.
Passive mode because of old files will be deleted by garbage collectors and retains in a couple of years and there will be only new files in hashstore.
After migrating to hashstore some nodes has stopped getting ingress. They really have free space and their dashboards shows ~2TB free space on each, but in multinode dashboard i see this
Of course it will not. Sometimes hashstore enabled nodes reports a negative free space.
The only workaround is to enable also a dedicated disk feature,
The dedicated disk feature is not calculating used space properly for the the dashboard, but at least it reports a correct free space to the satellites.