Hight IO wait times and usage *fixed*

[shimi269]

I been experiencing freezes and slowdowns when downloading via LFTP (using NerdPacks) on Unraid, I have managed to narrow it down to some IO issues, as put in the screenshots. But I'm at a loss as to what is causing the issues. I have tried running the download with Docker turned 'on' and 'off' and also by setting the paths to 'mnt/user' and 'mnt/cache', but BTRFS still spikes to 99% usage, and if downloading to 'user' SHFS to 99%. If I download to 'cache' BTRFS still spikes to 99% usage but also LFTP to 99% (as seen in the screenshot below).

 

 

The dips/freezes in the network, are matching to the spikes in the storage writes, as well as the 99% spikes (screenshot above). I download to the cache only, which then invokes mover every night. My Docker containers are stored on a sperate SSD, and I've eliminated them being the issue, by disabling Docker and running LFTP and still getting the issue.

 

 

Server Spec:

i7 2600k

32GB RAM

x2 250GB Crucial SSDs

x4 8TB WD Drives 

tower-diagnostics-20201125-0958.zip

Edited November 26, 2020 by shimi269
Spelling and Grammer

[shimi269]

*Update*

Change the cache drive from BTRFS to XFS, in the first test it appears to have made a difference. IO still sees some spikes but the network and download doesn't pause. Will carry on testing to see if the issue was with the Filesystem choice.

 

[shimi269]

After extensive testing, the issue is now resolved, it was just changing the cache drives file system to XFS and not using BTRFS