Yep - been after that for a while. To date I've found nothing within my capabilities that can do it; to be fair I've not seen anything bar MySQL server farms that will do it - and the memory requirements for that exclude it from realistic options.
Which seems to leave the answer, as you suggest, in either SA itself or the connector. In our case it may be that changes can be made that allow the connector to specify which MySQL SA should use. One issue that arises from this approach is synching up the data. I'd expect that in an established corpus minor changes in the synch's would mean negligible changes in scores. If we ever get it working I may find out :-D Nigel On Tue, 14 Jun 2005 13:58:01 -0400, "Matthew Yette" <[EMAIL PROTECTED]> wrote: >I would think this would just involve modifying the db connection >functions inside the SA code to verify the primary connection is >established, and if not , roll onto a second (backup) server. > >-- >Matthew Yette >Senior Engineer - NOC/Operations >MA Polce Consulting, Inc. >[EMAIL PROTECTED] >315-838-1644 (w) >315-356-0597 (f) >AIM/Yahoo: MAPolceNOC >MSN: [EMAIL PROTECTED] > >-----Original Message----- >From: Nigel Frankcom [mailto:[EMAIL PROTECTED] >Sent: Tuesday, June 14, 2005 1:52 PM >To: JamesDR >Cc: users@spamassassin.apache.org >Subject: Re: Re: shared SQL DB > > >Hi All, > >It's 5 SA actually :-D - and I've not noted any problems that didn't >come down to slight differences in config between SA servers, once >standardised they run much the same. > >The single MySQL is working well, my only concern being that I still >haven't managed to eliminate a potential single failure point (MySQL). > >My ideal would be to have a means of automatically selecting a slave or >copy of the main MySQL in the event the Connector can't get a response >from it (the master). It's all well and good having a manual option for >master/slave but that implies that you're watching your servers 24/7 - >oh that I had the time. > >One of the reasons for putting in so many SA boxes here was to allow for >multiple failures. > >If anything, moving SA off the SQL box has sped things up. > >One caveat here, email is coming in from a Win32 mailserver >(MTSProfessional) so email and SA are divorced from each other. > >HTH > >Nigel > >On Tue, 14 Jun 2005 13:20:03 -0400, JamesDR <[EMAIL PROTECTED]> >wrote: > >>Ronan McGlue wrote: >>> Hi all, >>> running 2 spamd servers and want to share the BAYES DB between >>> them... >>> so that one never gets too swayed and that both the servers are >>> contantly synced etc... >>> >>> anybody currently running an SQL backend off multiple Spamd servers?? >>> and if so care to part with some knowledge??? >>> >>> thanks >> >>We had (up until 2 months ago) 2 spamd servers hanging off one mysql >>server. It worked quite well. We could train on either of the spamd >>boxes (sa-learn not autolearn.) AWL also worked well. I know of >someone >>who has 4 or so boxes hanging off of one mysql server. He doesn't >report >>any significant load from SA (SQL wise) causing all around slowness. >>Maybe he'll chime in and add anything that I've missed.
