Thanks for the reply, Matus. Few more questions inline: On Wed, 2009-12-23 at 08:58 -0800, Matus UHLAR - fantomas wrote: > On 22.12.09 17:45, Manjusha Maddala wrote: > > Using LRU replacement policy, > > I hope you use at least "heap LRU" which is more effective. > > > I see many of the cached webpages for my > > site getting evicted. Could someone here tell me: > > > > 1. How often is the replacement done? > > when the percentage of configured cache_dir size crosses the cache_swap_high > value. So if my cache_swap_high is set to 96 and cache_dir is set to 100GB, only when the cache_dir size grows beypnd 96GB will the replacement task begin - is that right? Lets say I have 8 cache_dir directives in my squid.conf each pointing to a different directory. Now, if one of those directories is over the cache_swap_high value while all others are below cache_swap_low, will cache replacement work on just the first directory or on all others as well? > > > 2. Lets say the cache low water mark is set to 95, I understand that > > once the disk usage exceeds 95%, cache replacement kicks in. Once the > > process is completed, what would the disk usage drop to? > > the cleaning stops after cache usage falls under cache_swap_low value. > > > 3. I notice some pages in the cache_dir which are over a month old. > > Doing a "curl" to the squid cache with the URL of the cached page causes > > a TCP_REFRESH_MISS/FIRST_UP_PARENT entry to be logged in the access log. > > Which means, the page has become stale (by the refresh pattern rules) on > > the web cache and a new page is fetched from the parent site. From this, > > I infer that the page hasn't been referenced in the last one month and > > yet hasn't been marked for cache replacement while other newer cached > > pages have been flushed out. Why's that? > > apparently nobody asked for it and previous cleanings removed older objects. >
