bigbadbillyc

I think there can only be one Publisher

sillejo

I believe that is a SQL limitation. But since the subscribers can run on their own, why would you need a redundant publisher? Just keep a spare box off to the side in a lab enviorment with all the same info (IP, Name etc.) then if the Pub goes down, put the Lab box in play and restore from backup.

__________________
Deus ex machina

teltek

***additional to the hot standby server comment from silliejo***

This could work? Maybe?

It is possible in most intelligent switches to enable and disable each port.
SO...
You could potentially maintain your standby on an disabled switch port configured identically to your Master Server, then with a second NIC (non-server) connect it live for database synchronizations, or am I off-the-wall?

This way when you occur an outage you would be in hot standby by disabling the OOS system and enabling the standby simply through port management .... no?

Opinions are very welcome.

Cheers [[]]D

sillejo

Even with the 2nd NIC you won't be able to do SQL sync because the database ties the Server name into the sync. In your case you are trying to emulate the publisher which would not replicate to itself, eve on a second NIC.

Are you only running 1 server? If you have a subscriber then you have no worries about an outage. The only thing you would not be able to do is make database changes until you get the publisher back up. Restoring from backup only takes, at most, an hour. Taking into account boot-up time, ARP cache, and back-up restore (using Cisco BARS) you'r only out for about 1 1/2 hours max.

You could do the same scenario with the disabled Switchport, but you would still want to take into account the restore factor.

__________________
Deus ex machina

Mike4kz

Why would you need redundand Publisher? You should do backups, and if server ever crashed completely, just restore it from the backup. Or you may have a couple "spare" HDDs, swap them periodically, and store in the safe place. Then, in case of disaster, you could just boot server with that drive, and you will have Publisher up within a couple minutes.

In addition, why you need redundant Subscribers - if one crashed, second will pick everybody. If you have your Publisher crashed, your Subscribers will continue functioning, and only thing you will loose - ability to make changes (i.e. add phones, users, etc.).

EricP

Right now we are running a SL-100. Almost everything is redundant, and switch's activity without manual intervention. This was a area where I thought there was a little bit of a gap with Cisco so I was curious if anyone was using a work around. Its probably not a big deal since all calls will function as normal but it bothers me coming from a Nortel system that is almost fully redundant.

sillejo

I'm not sure how a publisher with mirrored hard drives redundant power supplies, able to fully cluster with multiple subscribers with the same hardware does not qualify as fully redundant? With a hot spare server sectioned off in a seperate VLAN you could even keep the thing on all the time. Say the Publishers building blows up. The Back-up is off site, accessible by the hot spare. Restore takes less than an hour, run remotly, then go change the routing so the Hot spares VLAN is now the same as the lost Publishers VLAN and you are back to a full system. In the meantime the subscribers have taken care of everything. I would say from time of discovery, you would have the new pub up in less than an hour and a half assuming a decent connection from the BARS backup to the hot spare Pub.

I guess I would need an explaination on how the Nortel would handle this type of DR situation better. I'm not a Nortel guy.

EricP

We have a SL-100. Each bay has redundant processors, power supplies, and in some cases link cards. They are logically separated in two different shelves. These systems can change processor activity without causing a interruption. A comparison would be from one subscriber to another. The front end is also redundant and can switch activity without a interruption. The reason why I asked is because it seams that the TFTP server not being able to have a hot standby by that can take over without manual intervention is a weakness. Yes I do understand that the phones will not be affected, only administration will be.

Thanks
Eric

sillejo

There is a TFTP server running on each CM so I think you should be able to point to redundant DHCP servers and those could have seperate TFTP servers defined. So lets say Pub has DHCP scope for Voice as well as TFTP#1 and the whole server goes down. You should have the Voice VLAN defined with a backup DHCP server and that Server could have a seperate TFTP server defined. Therefore if the Pub goes down, any new devices needing TFTP services will be told of the available TFTP via the DHCP info on boot.

this should be part of a fully meshed design. Even with DHCP running on the CM"s themselves you shpould be able to create this