-
Notifications
You must be signed in to change notification settings - Fork 1.8k
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
High unaccounted memory usage #10302
Comments
We have the same issue. Here is an example:
zfs 0.8.3, but we also observed the same issue on zfs 2.0.5. After analyzing /proc/slabinfo and /proc/spl/kmem/slab, it turns out there are huge amount of used space in kmalloc-512, kmalloc-8k, and etc. Here are the lines of kmalloc-512 and kmalloc-8k in /proc/slabinfo:
No idea why. We are very sure other applications on the server (such as Samba) do not use hundreds of GiB of memory. |
Same issue here, also proxmox 6 |
This issue has been automatically marked as "stale" because it has not had any activity for a while. It will be closed in 90 days if no further activity occurs. Thank you for your contributions. |
also seeing this on alma linux 9.1 (zfs-2.1.7-1) |
System information
Describe the problem you're observing
We have an issue where all our servers slab grow up to the point of filling the entire RAM. The server seems to be working fine even if our hypervisor is reporting over 95% ram usage. This is problematic because there is no way for us to accurately monitor the available ram to prevent the OOM to kick in and start killing processes. The only workaround is to reboot the server so that ram goes back down but over the next few week it will gradually climb up resulting in the same issue.
I don't have an exact way of reproducing the issue unfortunately as of now. I will try to spin up a test server in the next couple of days but was hoping my logs could help me shed some light on the issue.
Since this is an hypervisor there is a lot of read and write of small files (LXC containers fs)
Include any warning/errors/backtraces from the system logs
The server has 72GB of ram with only the following parameters set
Current ram usage on the hypervisor is 60GB/72GB. I've calculated the amount of ram used by all containers according to the hypervisor and it's around 15GB. The ARC is currently at 7GB according to
arc_summary
. That would mean 24GB of usage and leaving 36GB of unaccounted ram which is roughly 50% of the total memory of the server.Could it be that something is not releasing memory correctly in SUnreclaim ?
I tried the following commands which freed some memory but a big chunk still left.
Here the slabtop
Here are my ARC stat
And the pool (I know there is a SLOG and L2ARC but the issue also appear on other servers without them)
Any ideas ?
The text was updated successfully, but these errors were encountered: