Re: [knot-resolver-users] cache filesystem size and setting a cache size limit

Hi Christoph. On 5/5/19 2:05 PM, Christoph wrote: I can't see what's wrong there yet.  For me LMDB used by kresd allocates `data.mdb` file of exactly the chosen size and `lock.mdb` of 8 MiB.  The implementation is basically just calling mdb_env_set_mapsize() [1] [1] http://www.lmdb.tech/doc/group__mdb.html#gaa2506ec8dab3d969b0e609cd82e619e5 It's possible tmpfs has some overhead, so 50 MiB might be a too small margin for the mount size if the size is e.g. in gigabytes.  In any case, tmpfs is supposed to only consume memory used by files inside, so mounting with larger size should be for free, except for giving up a kernel-level upper bound on size. I believe it works, but perhaps you can send more details so we can reproduce it, assuming the deviation is significant.  Before that you'd best check what `ls -l` reports on the files.  I think there are some minor caveats, some mentioned in [1] - IIRC low limit of 10 MiB, rounding to 4 KiB, truncation sometimes impossible because it never moves written data. --Vladimir