View Single Post
  #1   (View Single Post)  
Old 10th October 2008
PatrickBaer PatrickBaer is offline
Fdisk Soldier
 
Join Date: May 2008
Posts: 81
Default Project layout, any comments?

Morning,

so I have made up my mind on the OS question. FreeBSD will be the choice for the main servers, while OpenBSD shall be working in the DMZ.

This is the config I have in mind:

Failover: Haven't completely made up my mind yet. Redundant systems should be able to not just monitor ICMP connectivity, but service availability and switch when the second host fails:

Scenario 1, Host A is active and crashes completely: HB-request or equivalent detects outage, Host B takes over and becomes active by bringing up IP-adress on virtual interface.

Scenario 2, Host B is active and imapd fails: Host A detects failed service and sends a "become inactive"-request to Host B, Host B turns down virtual interface, brings up virtual interface and becomes active.

Proxyservers:

OpenBSD, failover as described above
no immediate replication besides FTP. Changed config data replicated hourly via cron and rsync.
Bind9
Proxy
Spamassassin, amavisd, clamd
ftp,sftp, replication via ???

Domain Controller:

FreeBSD, failover as described above
MqSQL master-master Cluster
SMTP,imapd,pop3, configs AND emails stored in MySQL DB, hence immediate replication
DHCP (no idea about replication yet)
LDAP, Samba PDC, storage in MySQL

Fileserver:

Linux, maybe switched to FreeBSD in the future, currently no need to change
nfs, samba without any replication. In case of a hardware or OS failure, the 16 disks can be mounted in a spare case and brought up as a replacement. OS held on a spare hdd.


The PDC part and NT Domain introduction will be completely new, the other services do already run, more or less. The failover is more than buggy.



With this setup, it will(should) be able to switch back and forth between Host1 and Host2 at will. The master-master databases should make them absolutely consistent, even if the failover happens in the middle of hundreds of write transactions. The MySQL database is (I hope!) able to handle 50 to 100 users accessing the emails simultaneously and the fiber-channel connection for the replication should speed up things.

I am open for comments and suggestions! (Yes, that's a question
Reply With Quote