On Sun, Aug 14, 2005 at 01:03:58AM -0400, Lyle wrote: > The consequence of predicting an average chunk size that is "too big" (waves > hands a bit) is that the available space in the cache will not be fully > utilized whenever the actual chunk size is smaller than that. The > consequence of predicting an average chunk size that is "too small" is that > cache reclamation will be slightly less than optimally efficient, boot will > be imperceptibly slower, and a tiny bit of disk space will be wasted on > empty chunk files.
I guess I'd rather fall on the side of 'too big' then. FYI, an 8G disk cache takes quite awhile to scan on boot with the old tuning parameters. While we are patching things, can we patch 'fs getcacheparms' to be able to show number of chunks used/free as well? For instance, I'd like to get some better idea if I'm limited by number of chunks, or space, particularly on this machine: [EMAIL PROTECTED]:~$ fs getcacheparms AFS using 8277622 of the cache's available 10485760 1K byte blocks. _______________________________________________ OpenAFS-devel mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-devel
