The ultimate guide to LVM Caching


Also, anyone currently running any cool lvm cache setups? or want to share their setups or lessons learned for me to incorporate into the video?

I shot a video ages ago… editing backlog… updating it with some of the new data/metadata features. One quirk I found was that with large metadata pools it could degrade performance e.g. nvme + hdd was always fine but 280gb optane + 1tb nvme would be slower than either individually… but a smaller cache for just metadata worked really well with optane like 50g or so for 1tb of flash (and I doubt it even used the full 50g)


What specific storage use case(s) do you have in mind (if any)?

Power user, mainly. So lots of games, steam and otherwise, lot of local storage for movies music etc. Assuming no Nas.

Also maybe the edge use case where you have a laptop that has nvme and SATA SSD.

1 Like

The guide I made the other week was on a proxmox server . I put all the virtual drives (qcow) for the VMs on the lvmcache setup I created.


At work, I setup a lvmcache on the /var file system for a postgresql database (for spacewalk server). It was nice.


I would love to see you compare it to the other software some other YT channels have tested… (optane memory on intel, senseMI , other software that linus mentioned… etc )

Write back or through?

My proxmox is using writeback.

The spacewalk server at work is using default write through (but I could have changed it to writeback)

1 Like

The man page says to only use 10% of cache lv for the meta lv

For gits and shiggles, I’d be curious to see how LVM cache performance compares to bcache. I’ve been using bcache for my workstations with decent results, but don’t really trust it yet on production systems.

This thread is among top Google search results. Please edit the original post to include some information, because currently the title is very misleading.