On Wed, 13 Jan 1999, Richard van Drimmelen wrote:
> Suppose that within one month my cache will grow bigger than the
> cache_dir size of 1500. That means no entries in the cache will have
> been removed because they have not been accessed for over one month.
> Will squid try to write to the cache (and create errors because the disk
> is full) or is a smarter algoritm involved ?
According to the source:
/*
* storeExpiredReferenceAge
*
* The LRU age is scaled exponentially between 1 minute and
* Config.referenceAge , when store_swap_low < store_swap_size <
* store_swap_high. This keeps store_swap_size within the low and high
* water marks. If the cache is very busy then store_swap_size stays
* closer to the low water mark, if it is not busy, then it will stay
* near the high water mark. The LRU age value can be examined on the
* cachemgr 'info' page.
*/
time_t
storeExpiredReferenceAge(void)
...
Also, when storeExpiredReferenceAge() is less than 300 Squid stops
caching objects. See store_check_cachable_hist.no.lru_age_too_low.
Alex.
