[[6.8.3] docker image huge amount of unnecessary writes on cache]

Switched to XFS, rebuilt the docker.img with the same dockers as I had before. Now `loop2` is around 9-30MB/min.

 

For those who are still on BTRFS perhaps it's worth trying removing the docker.img and re-adding all dockers. It might help. I should have tried that first.

 

My original post: https://forums.unraid.net/bug-reports/stable-releases/683-docker-image-huge-amount-of-unnecessary-writes-on-cache-r733/?do=findComment&comment=9019

 

Edit: after having it run for almost a day, the `loop2` write seems to be stable at around 1.7GB/h.

 

[[6.8.3] docker image huge amount of unnecessary writes on cache]

Unraid 6.8.3

Cache: Crucial MX500 2TB SSD - BTRFS w/o encryption

 

Just want to chip in and say that I have similar issue, write on the cache drive hovers between 5-10MB/s constantly. `iotop` reports huge amount of writes to `loop2` for no reason, or so it seems.

 

I disabled all active dockers and started enabling them one by one to see how it affects `loop2`. Out of all dockers that I have, `nginx-proxy-manager` seems to have the most effect on it. Without it `loop2` writes around 100MB/min and the cache drive in the UI shows as low as few KB/s or even 0 for writes, which is for a total of 14 dockers, arguably some of them aren't that heavy to begin with or don't do much IO in the first place. Those 100M might be related to the dockerd logs which are reported as `dockerd -p /var/run/dockerd.pid --log-opt max-size=10m --log-opt max-file=1 --storage-driver=btrfs --log-level=error` in `iotop`, althogh these report less than 1MB/min. Then there are `shfs /mnt/user -disks 4095 2048000000 -o noatime,allow_other -o remember=330` processes which report 2-30MB/min. With `nginx-proxy-manager` docker enabled (plus all the other dockers), the write to `loop2` jumps to around 400MB/min, the dockerd log processes go to around 2-3MB/min, and shfs ones jump to 70-90MB/min each.

 

I used to have 6 Crucial SSDs in RAID10 for cache, which ran for about 2 years. Upon inspecting them with Crucial tool, all of them reported around 260TBW which is crazy high for 2 years of really not that intensive load.

 

I've seen people mention `pihole` and `nzbhydra2`, which I also run, but they don't seem to affect it overall as much as `nginx-proxy-manager` does.