HI Maciej,
OK, please be mindful that RedEye is only at release 1.0 and there are more things that we will introduce into it over time. Our first objective was to get the failover working reliably and smoothly, which it does. Failover, with zaptel cards fitted to each box, takes about 12 seconds from point of failure to up and running on the "B" (backup) server. This time delay is almost entirely due to the UDEV startup delay when bringing up zap trunk drivers on the the B server. In a near release we will significantly reduce the failover time by pre-loading the drivers on the B server.
HA 1.0 is running in production in at least one commercial deployment (with another two sites just about to be installed) and it is behaving well in the field.
Synchronization is currently limited to "Loose" and "Lazy". Loose means no automatic synch (you have to manually do it yourself). Lazy will keep the SARK schemas in synch between the two servers. However, it WON'T keep voicemail files or CDR's in synch and you need to be aware of this.
The way Lazy works is by setting up an SMB share across the two servers and periodically comparing the B server SARK schema with the A server. IF the two schemas differ then the A server schema will be copied to the B server and a "soft" commit issued. The soft commit will generate all of the asterisk files, tftp server and so forth but it won't start/restart asterisk on the B server. Asterisk will ONLY be started on the B server if it senses a failure on the A server.
Due to the nature of HA, we don't believe we can run the cluster in server-gateway mode so it has to be server-only.
Finally, due to the nature of SME server and its startup mechanism, we cannot currently support DHCP server in this environment. So DHCP has to be provided by a third server.
Kind Regards
Jeff