A huge amount of data writes

[OFark]

Hi all,

 

A query I have; I have an NVME drive failing SMART because it's passed it's probable amount of data it can store ok.

 

This is an unassigned drive, on it is Docker.img, appdata, libvirt, and my VM images. I have one VM (proxy gateway) that I setup a few weeks ago. I have a dozen Docker images, Plex, and the supporting containers . Nothing that I would consider to be write intensive to itself. I mean, I know Plex will store media files etc, but what I have seems out of the ordinary:

 

Power on time is 296.5 days, less than a year. 141TB??? It's a 128GB drive, 32gb for Docker.img, 90GB used in total.

 

I've done some reading and I know I could over-provision it, but is there something I can check for? Is there someway to identify whats being written so often to the drive so I could move it to rust?

 

I'm thinking of putting a 256GB in there, but if that's not going to last 2 years, it's a costly expenditure. 

[JorgeB]

Probably related to this:

 

[OFark]

I'm learning a lot here, but from what I can see this relates to a RAD-1 scenario, which I don't have. But yeah, I am seeing everything else.

[JorgeB]

1 hour ago, OFark said:

but from what I can see this relates to a RAD-1 scenario

It doesn't, it can happen with a single device, I only have a single NVMe for cache, and v6.8 was writing 3TB per day, with the fixes in v6.9 it decreased to less than 200MB per day.

[OFark]

10 hours ago, JorgeB said:

It doesn't, it can happen with a single device, I only have a single NVMe for cache, and v6.8 was writing 3TB per day, with the fixes in v6.9 it decreased to less than 200MB per day.

Good. I could swap this one to XFS, but tbh it's already dead. I'll put in a new one, XFS and hope 6.9 comes out sooner rather than later.