> > Of course if the checkuser patch could consult a mysql database > > instead it > > would be cleaner but gotta live with what we have. If > anyone has a better > > solution let me know. > > > I run a cluster of qmail/vpopmail using mysql replication and > NFS, and it all works flawlessly. I'm planning on integrating > the chkuser patch shortly to start reducing the volume of > queued bounces I have to deal with.
To further Nick's comments... This (vpopmail front end servers with a common NFS store and users in mysql) is how we run now and it works great. The checkuser patch works perfectly with mysql because it makes calls to the libvpopmail.a to handle lookups, so if you have vpopmail compiled with the correct settings for mysql (vmysql.h and appropriate configure options) it works perfectly. A couple of gotchas though... If you are running a high concurrency on qmail-smtpd be sure that your mysql server has a similar max connection limit. If the checkuser cannot talk to the mysql server (too many connections) it will send out a rejection notice for a valid user (chkuser - 5.1.1 I think, been a while since I have seen one . Been there, done that, bought the cheap t-shirt.) I was never able to get the checkuser patch and our smtp-auth patch to work together. So if you are using an smtp-auth patch verify that both work together before moving it into production. And yes... It really does cut down on the double bounces... Previously before using the patch our qmail-queue use to run about 13K to 15K messages (and that was with a 2 day expire instead of the stock 7 day queuelength). We now run (combined among our servers) ~2000. Big difference. And we are very pleased with the checkuser patch... I worry about harvesting attacks against our server, as it lets the spammer know right away which address is valid and which isn't... But that really is a problem with the SMTP spec... Not the checkuser patch... We have been running the checkuser patch since July 2003 and have had no problems other than the SMTP-AUTH and mysql max_connections issues. Tom Walsh Network Administrator http://www.ala.net/
