On Thu, Mar 18, 2021 at 01:36:55AM -0000, SaveTheRbtz wrote: > Is it possible (pun intended) to compromise on > `CONFIG_SQUASHFS_DECOMP_MULTI` (which uses num_online_cpus) instead of > `CONFIG_SQUASHFS_DECOMP_MULTI_PERCPU` (which uses `num_possible_cpus`)?
It's better, but still not that great of a compromise in my opinion. It's fine for systems with small numbers of CPUs. I still don't think it scales well for large CPU systems. For instance, one machine I have access to has 208 online cores. There we would have 26M of RAM allocated just for the decompression caches for each squashfs mount, which probably isn't fatal since large-cpu systems likely also have large amounts of RAM, but it's also pretty wasteful since it's highly unlikely that 208 squashfs decompressions for the same super block will be going in parallel. It's possible to convince us to change to that option, but we'd at least need more information about what problems the current setting is causing and some testing to confirm that changing the option helps. Please file a new bug for this request. You can subscribe me to the bug. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1636847 Title: unexpectedly large memory usage of mounted snaps To manage notifications about this bug go to: https://bugs.launchpad.net/snappy/+bug/1636847/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs