> Hello All, > > I have mail server and also backup mail server, so when the first one is > down, the mails are stored on the second mail server, and when the main > (first) mail server is up, the second server sends them to the main > server, and users can receive their mails. > > Now the question is the following. How can users receive the massages > immediately from the backup mail server, before main mail server is > restored ?. Simply put - there is no real graceful way of doing this except to build a distributed (modular) mail platform and ensure redundancy at each of the various points in the model. In this case, you would want two POP servers that share a common disk area and a disk unit that expoerts its volumes to the front end servers (in this case POP servers, but would also need to do it to the SMTP servers for inbound mail as well) Once you do this you run into issues with mail box locking - which is where the Maildir format comes in quite handy as a user can be reciving their mail an the SMTP server can write new inbound messages at the same time without file system locks. You could then setup a DNS round robin type system for load distribution or something such as a layer4 switch to ensure automatic failover (layer4 would be superior in this reguard, tho DNS RR with a short TTL and updates should a system drop may also work well) Usually in this case, if your mail system breaks then put up a temp notice for your users and tell them you are working to get it fixed and spend the $$ in getting decent hardware and redundant disks as you will find that this will improve uptimes to a far greater degree than anything else. > > Is there a possibility to do mirroring mail servers under RH9 ???. And if > there is so how ? > Who can recommend me a good solution for this task ? > Also I would like to do the same thing for DNS server. Please help !!!!! DNS is rather simple to do in this reguard as DNS has redundancy built into the protocol. All you need to do is setup n DNS boxes and then mirror the bind zone files and configuration files across them - you can also do some rather neat things with automatic failover using routing protocols (we used to run a cluster of 4 DNS servers that talked OSPF and assumed each others IP addresses in event of failure - worked a little too well however and we lost all 4 servers over the space of 6 months at one point and never noticed till the last one died <grin> - monitoring would need to be setup properly) I used to work on a system where I had rsync copying the zone files out to two DNS servers and then a shell script that checked for files transfered and remotely ssh'd out to the remote DNS server and did an ndc reload if files had been updated. There are other more graceful ways to do the same thing (hidden primary with DNS Notify's sent out to the public facing slave servers or even using a common database and getting the servers to slave zones from that) but it really depends again on how complex you want to make things and usually, simple is better. Also, remember, if doing things like automated restarts on name servers - stagger the updates otherwise you may end up restarting all your DNS servers at one which tends to be bad [tm] HTH. -- Steve. -- redhat-list mailing list unsubscribe mailto:redhat-list-request@xxxxxxxxxx?subject=unsubscribe https://www.redhat.com/mailman/listinfo/redhat-list
