Hi, I am trying to keep NFS client inodes and dentry caches in memory for as long as possible using /proc/sys/vm/vfs_cache_pressure = 1. One side-effect of this is that the culling runs in cachefilesd seem to be suspended until a certain number of files (or blocks) are ejected from cache? I believe this is reduce the number of culling iterations and only run when we think there might be some files to clear up? But if we never drop anything from inode/dentry cache, then we want to run culling periodically and not rely on the count of recently freed inodes (otherwise nothing will happen). I believe this functionality was introduced here: https://kernel.googlesource.com/pub/scm/linux/kernel/git/dhowells/cachefilesd/+/4f881c327573cce04977c116330d0a9e05366c60%5E%21/#F2 But the actual cachefilesd.conf keyword (described in the docs/man/config) doesn't seem to work (or I am using it wrong). Setting "resume_thresholds - -" or even "resume_thresholds 10 10" just results in an error: cachefilesd[55348]: /etc/cachefilesd.conf:21:CacheFiles gave config error: Operation not supported I'm using the RHEL7 supplied cachefilesd v0.10.9. I might try editing the cachefilesd code to set "f_resume_threshold = ULLONG_MAX" (instead of the default "5") and recompile to see if that disables this functionality and helps culling with my vfs_cache_pressue = 1 case. Daire -- Linux-cachefs mailing list Linux-cachefs@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cachefs
