Hello, I have a perhaps different setup where I have a MariaDB Galera Cluster spanning two webservers with a third arbitrator server for Galera sitting "on the side". The webservers are utilizing Nginx and various PHP versions. Then I have a GlusterFS replication cluster consisting of three external servers with one slice mounted to /var/www/clients - to handle normal clients/resellers one slice mounted to /var/backup - to handle backups one slice mounted to /usr/local/ispconfig/interface/acme/.well-known - to handle acme.sh generated verification files Then I have a HAProxy, currently only one, that loadbalance traffic between the two webservers in a roundrobin fashion both for http and https(this took some work). All this is actually working quite ok. I can't have user-/group-quota as GlusterFS doesn't support it but being more for personal, and PoC, use I have no issue with that. I have a few issues though. 1. When the client activate "Letsencrypt" for SSL the mirror server doesn't get this information but I need to re-update the client again to get it to mirror over to the mirror server. I don't know if this is by-design as it could take a while before the ssl certificate is issued from acme.sh. 2. It looks like the mirror servers admin web ui doesn't use the "master" servers database ? So I can't really use the secondary servers admin web ui to handle the servers which is unfortunate .. are there away way to make it use the master db instead of the local server db ? 3. I've overcome this but as a note - when I setup the secondary server I choose to let it use "localhost" as master server and the master servers database .. this is working great but the user created didn't include the host "localhost" so had to copy the local server user with @server server to @localhost.
Sorry, not an answer, but could be useful anyway. GlusterFS supports quota... At least I'm using it. Are you on a recent release?
Yes that is very interesting, what are your mount options ? because everywhere I've read it just states that userquota is not supported ? Code: # gluster --version glusterfs 9.5
This should be /var/www Do you have client protection enabled in System > Main config? If so, disable it. You should only use the admin UI on the master. It should always use the master as master DB... That seems clear.
Ok, makes sense I'll correct that. Yes, I had.. what does this do ? It's not mentioned in the latest manual(3.1) Well.. it is a Galera Cluster so there are no real "master" and thus it makes sense to use "localhost" as the db can then be maintained even if they go split-brain or one server goes down. And if I have the DB's in a Galera it also makes sense to have the ISPConfig admin interface in a "master/master" fashion ? so the ISPConfig Admin UI still uses the "master" database but on the localhost ? I understand if this is not implemented and I'm fine with that.. I guess I got my explanation by the comments But as a feature request then.. perhaps it should be possible to have the ISPConfig Admin UI setup in master/master if the master database is kept in sync ? In a real HA setup I would probably opt to have a separate setup for just the ISPConfig Web UI and have that on multiple servers
It prevents client users from making changes. The ISPConfig DB shall be on the slave server itself. It is possible to use a cluster but it's not our recommendation You can host the master UI on several servers, but in our experience it can lead to more bad than good
