-
Notifications
You must be signed in to change notification settings - Fork 1.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Allow pools to share a single l2arc cache #9859
Comments
While probably not impossible. Sharing an l2arc device in this fashion would be exceptionally difficult to implement since it requires associating a single device with multiple pools. This is something which is expressly prohibited by the design. Regardless, I've tagged this issue as a requested feature so it can be tracked. |
We did a design for this a few lives back. The arc.c changes are minor and relatively easy to implement and test. The hard part is managing a new spa type that does not include any data vdevs. Those changes are much more significant and impact many user-facing interfaces. We had the advantage of a restricted and highly automated user interface, so it was feasible for us. That said, it was a strange design decision to tie the cache devices to pools when ARC itself is shared amongst pools, except for cstor where each pool has its own ARC. |
Any update about this feature request ? |
would also love to have this ... would be awesome! |
Stumbled upon this while searching if this was possible to use the same L2ARC/ZLOG devices for distinct zfs pools. I also have servers that would benefit from this feature being implemented. Not assiduously counting on it, but still thinking this would be a good idea. |
Pull-Request #14060 is about implementing this. |
Interested in this. I see the PR appears to have stalled but hopefully gets picked back up. Would be great if L2ARC was similar to ARC and shared between all pools to maximize the benefit and use of capacity. Understood implementation is not as simple, but hope it gets there. |
Describe the problem you're observing
I have multiple pools of spinning disks with varying levels of activity over time, and a cache device on NVME.
The cache is quite effective - especially with #9582 - but manually partitioning the NVME into multiple independent caches is leaving some of the caches cold for long periods while other caches are struggling for space, which is far from ideal for performance/space/expense reasons.
It would be wonderful if multiple pools could point to the same cache device and have l2arc manage their relative needs.
The text was updated successfully, but these errors were encountered: