Leaderboard

Ok, still poking things to see what happens. I am sure others have already done most of this testing but didn't have time to read the whole thread. So while moving the docker image to an XFS formatted array drive (decided to use unbalance instead of mover so I can watch the progress) dockers were disabled obviously. With the dockers disabled entirely, I noticed zero writes to btrfs-transacti or loop after 5 minutes as you would except. In fact I saw no writes of any kind, exactly as I would expect. When I tried to start docker after moving it to XFS array drive, it would not start, so I restarted the server. It then started up ok. So now watching the writes with docker on an XFS array drive. After 10 mins I am at 50mb writes to loop2 and 120mb writes to btrfs-transacti, which netdata shows is indeed still being written to the cache. That is still over 5TB a year to the cache for no reason at all. By increasing the dirty writeback time to 3 mins it drops the writes in 10 minutes to 35mb for loop2 and 70mb for btrfs-transacti. Not nearly as much of a drop as when docker was on the cache but still noticeable. This is confirmed by the writes increasing on the cache in the main view. In 45 mins of up time I have about 7500 writes on both the array drive with the docker and my cache pool. BTW, what does each "write" represent in the main view? If I disable the docker service, then btrfs-transacti doesn't even show up. Anyone that has gone the path of mapping the docker image directly without loop 2, does btrfs-transacti still cause writes to the cache? I might have to go that route if the official fix is months away. Was checking the smart status of my SSD's and in just over a week since installing unraid they all ticked down 1 of the lifetime in the smart stats. Took me well over a year to do that when using windows. Sure glad I caught it now.