On Sat, 2 Jun 2012 01:38:23 -0700 (PDT) rtsit wrote:
> >> Since it's over the threshold of 112500 I figured that it would > >> expire the extra tokens. > That's hardly ever going to work. > > [GAP #1]: That's harly ever going to work BECAUSE... > > The lowest bayes_expiry_max_db_size you should use is: 100,000/0.75 = > 133,333 > > [GAP #2]: The way to fix that is by setting the variable.... It looks like I misunderstood what you were saying. When you wrote "the threshold of 112500", I thought you had set bayes_expiry_max_db_size at 112500 instead of the default of 150000, whereas it's actually the target number of tokens 0.75*bayes_expiry_max_db_size >... dbg: bayes: expiry check keep size, 0.75 * max: 112500 In that case it looks there's not much you can do. It's trying to remove 34399 tokens, and it's looking for an atime cutoff that will remove up to that number. > May 25 20:55:13.468 [8792] dbg: bayes: 691200 37284 (8 days) > May 25 20:55:13.468 [8792] dbg: bayes: 1382400 0 (16 days) Cutting retention to 8 days removes a few too many tokens, 16 days would remove none, so it has no solution. What will happen is that it will eventually find a solution at 16 days, or 8 days if the account receives a higher volume of mail.
