Hi,

I have constant writes to cache (<500B/s every 2-5s).

When I turn off all my dockers, this drops to 0.

I systematically checked every docker and isolated to radarr, sonarr and overserr.

When I stop these, the frequency drops down a fair bit but it still happens.

I used iotop and noticed the activity linked to the writes were kworker events:

kworker/R-kvfree_rcu_reclaim

kworker/u48:1-btrfs-endio-write

kworker/u48:5-btrfs-endio-meta

kworker/u48:1-events_power_efficient

kworker/u48:2-bond0

What is this? I'm assuming I have done something wrong.

tower-diagnostics-20260613-1117.zip

This is completely normal. These apps are very active for a number of reasons.

1. Background Databases: The apps constantly update their internal databases to record state changes, inner task progress, command statuses, and scheduled tasks. Background tasks run as often as every 90 seconds, causing persistent, low-level write activity.

2. Active Refresh & Monitoring: They operate with a hard-coded task (often every 1–15 minutes) to "Refresh Monitored Downloads" and regularly synchronize RSS feeds with your indexers.

3. Frequent Metadata & Log Updates: They routinely fetch or refresh artwork, cast data, and episode metadata. Additionally, they persistently write operational text logs to disk depending on your configured log levels.

Any app with a database, you can usually expect a lot of disk activity.

3 hours ago, strike said:

This is completely normal. These apps are very active for a number of reasons.

1. Background Databases: The apps constantly update their internal databases to record state changes, inner task progress, command statuses, and scheduled tasks. Background tasks run as often as every 90 seconds, causing persistent, low-level write activity.

2. Active Refresh & Monitoring: They operate with a hard-coded task (often every 1–15 minutes) to "Refresh Monitored Downloads" and regularly synchronize RSS feeds with your indexers.

3. Frequent Metadata & Log Updates: They routinely fetch or refresh artwork, cast data, and episode metadata. Additionally, they persistently write operational text logs to disk depending on your configured log levels.

Any app with a database, you can usually expect a lot of disk activity.

Thanks for that advice.

I thought it was less activity before but maybe my memory was wrong and didn't notice until recently.

Just now, bobalot said:

Thanks for that advice.

I thought it was less activity before but maybe my memory was wrong and didn't notice until recently.

Yeah, I wouldn't worry about it. With docker containers and maybe VMs, your cache drive will be almost constantly writing. I went down this rabbit hole myself in the early days of using unraid and learned that this is just something I have to live with. You would think that it wears down the drives faster, and it does to some degree, but not enough to worry about. It's just very small writes most of the time. You can look at your drives' health by clicking on it and checking the attributes tab. In the end, It's all the linux ISO downloads that are gonna wear it down faster than the small writes do.. :P