On 6 May 2022, at 11:31, Benny Pedersen wrote:
> On 2022-05-06 11:25, Henrik K wrote: >> On Fri, May 06, 2022 at 11:08:21AM +0200, Niels Kobschätzki wrote: >>> Hi, >>> >>> I have a setup where the spamassassin-servers have actually no access to the >>> data of the mail-servers. Now I was looking into having per user >>> bayes-databases and saw that I can do that with a SQL-database. I have >>> already >>> a small galera-cluster and I wonder if spamassassin will work with it >>> because >>> of the limitations galera has. >>> The limitations are: >>> >>> * only innodb >>> * unsupported explicit locking >>> * a primary key on all tables is necessary >>> * no XA transactions >>> * no reliance on auto-increment >>> >>> Does anyone have experience with such a setup? >> >> I see no reason why it wouldn't work, none of the limitations should apply >> to SpamAssassin. Great :) I’d rather be safe than sorry and like to ask. > fair, its just that redis is more prefered to bayes imho, and postgresql is > high performance without being memory hungry But I read that redis doesn’t have per-user databases? And I probably would need new machines with lots of RAM for it, because I have no idea how much RAM is needed per user. And I already have a galera-cluster running and don’t want to set up yet another database-cluster (psql). Niels
signature.asc
Description: OpenPGP digital signature
