Getting constant activity on all harddrives

[jeffmackinnon]

For the last couple of days all my harddrives have been getting hit with reads on the data drives and writes on the drives. I have tried shutting down all dockers and VMs, and turning off all NFS shares and nothing changes.

 

 

 

I checked htop, but nothing looks strange there.

 

Any tips on what I can look at next?

 

 

EDIT:

 

Some more troubleshooting has found that after updating all the dockers again, and shutting them all down the constant activity stops. Then I restarted Plex, its the only docker that is "critical" and the reads started.

 

The confusion now is that it is "reading" from an empty disk -

 

 

Disk 9 was just added to the system and doesn't have anything written to it.

Edited June 16, 2020 by jeffmackinnon
Some new information discovered.

[Squid]

It appears that you have dual parity, and have turbo mode (reconstruct write) enabled in disk settings, and some process is currently writing to disk 2 @ 54.6 KB/s.

 

Since you have zero cache-only shares, I'd surmise that you don't have a cache drive and the docker image is stored on disk 2 which is what would be resulting in the minimal write happening.

[jeffmackinnon]

2 hours ago, Squid said:

It appears that you have dual parity, and have turbo mode (reconstruct write) enabled in disk settings, and some process is currently writing to disk 2 @ 54.6 KB/s.

 

Since you have zero cache-only shares, I'd surmise that you don't have a cache drive and the docker image is stored on disk 2 which is what would be resulting in the minimal write happening.

I have system, appdata and domains set for my cache. Those are the only shares that use the cache.

 

 

A user on reddit suggested that docker.img was likely a culprit, and they weren't wrong. But when I moved it from disk2 to cache using mc and re-started the plex docker the same thing happened again.

 

There is a possibility that FileHistory is causing the activity, but I would expect to see a SMB stream active if that was the case.

 

I've changed the "prefer" shares to cache only, ran the mover, and verified that the system, appdata and domains are 100% on the cache