On 2021-02-11 9:54 am, Alex wrote:
> Hi, > There is no real question, but what I would like to find out is (and to ask), > does it scale and are any pitfalls? Naturally, we would look at doing HA, but > am asking for that any comment, any tip, any opinion on using redis for > bayes. Been using it from day one (I'm party to blame we have this) and it > scales VERY well. Bayes processing bottleneck has become a thing of the past. > Pifalls? none so far. I wouldn't go back anymore. Obviously, it's global > only, no per user. Is there an easy/efficient way of converting an existing mariadb bayes database to redis? Perhaps "sa-learn --backup", set up redis, then restore? I know I've been less than successful in the past when migrating from one version of mariadb to another, so just wondering how successful this approach would be. The problem I'm having with bayes in mariadb is being able to use a central database server for the database, while reading and updating it from remote systems. Will redis solve this problem? # sa-learn --dump magic 0.000 0 3 0 non-token data: bayes db version 0.000 0 11083 0 non-token data: nspam 0.000 0 48363 0 non-token data: nham 0.000 0 3709015 0 non-token data: ntokens 0.000 0 1372117134 0 non-token data: oldest atime 0.000 0 1613055126 0 non-token data: newest atime 0.000 0 0 0 non-token data: last journal sync atime 0.000 0 1606461007 0 non-token data: last expiry atime 0.000 0 0 0 non-token data: last expire atime delta 0.000 0 0 0 non-token data: last expire reduction count I've had good luck with using mariadb and galera to share the spamassassin database across systems. I run a small 3-node setup for email, 2x servers running dovecot replicating to each other, and a 3rd galera quorum server. Mariadb is master-master across all 3 nodes, so changes on any one are replicated to all the others via vpn. Works well, and for the amount of data in the spamassassin database, it replicates very quickly.
