Matus UHLAR - fantomas wrote:
I have noticed a strange behaviour of squid:
I have 4 disks of different size dedicated for cache_dirs in my machine.
With 4 cache_dirs of equal size (small enough to fit all disk sizes) all directories are used evenly.
After increasing to of the cache_dir located on the bigger disks the smaller cache_dirs are no longer used to store new objects.
What is the selection criteria when squid chooses the cache_dir?
Is there any explanation to this behaviour?
On 24.11 18:58, marc elsen wrote:
There is a parameter called
store_dir_selection_algorithm
in squid.conf.default , if I remember its name correctly. Check it out, by reading the comment.
the comment is not 100% clear. Does squid measure the disk speed (latency) or just number of requests?
in the first case, it would happen that 5X faster disk would get 5X more requests, not taking whole capacity...
I was thinking about configuring slower disk for larger files, while letting squid store small files only on the faster disk - that should do it. But I should probably check the sources :)
The comment I have seen is that there is round-robin as an alternative to least-load. Thats it. Quite clear, but not very much to read...
Is there more in the newer releases?
Your approach with using different types of disks is interesting, but I think I will switch to round-robin...
