Silly squid

Seems squid breaks down in tiny little pieces if the hard disk partition it runs on turns out to be too small to contain everything you would need according to the cache_dir setting.

My server runs squid, which helps keeping the downloaded volume of data traffic low. I mounted a 1G partition on the /var/spool/squid directory, and said squid by way of the second parameter to the cache_dir setting that there was 950MB for his usage. Which turned out to be just not enough; with the filesystem overhead and the 5% reserved blocks, the thing ended up overflowing. Bye-bye, proxy. Restarting it just gave me something in the logs, mentioning that there isn't enough space. Oh well, so I lowered the setting, and restarted squid.

That didn't work either.

Now, I can understand that they break down (although it would've been nice if they had handled it, perhaps by dynamically adjusting the size); what I cannot understand is that they break down after restarting, and after having those values adjusted.

Anyone with a bit of squid-knowledge care to shed some light on that mystery? For reference, this is what the setting looks like right now:

cache_dir aufs /var/spool/squid 930 16 256