Add'l Pollers and Hot Standby questions

gork

Hello all,

Currently (3) primary sites with +9K objects and no DR or local pollers.
All sites are connected by multi-gig switching (ring).

The main site has the NPM server which is also a poller.
The database is not installed on this server.

I am planning on adding another poller and a Hot Standby server in the DR (secondary) site (there will also be a copy of the database at this location geoclustered). Nodes will be migrated to this poller to alleviate stress on the primary and localize traffic to this active datacenter. The Hot Standby server will take over if the primary server fails (they are not on the same subnet).

Planning on adding a third poller at the third site which is large (+3K objects). Nodes will be migrated to this poller to localize traffic.

With Hot Standby it's not entirely clear but here is what I understand will happen - please correct me if I'm wrong (#3 is a question):

1. If the Primary NPM server fails the hot standby server will take over. Traffic going to that server (other than the exceptions like VMWare, etc.) will now go to the Hot Standby server until the primary is restored. Until the primary NPM server is restored certain poller features will be unavailable.

2. If any of the Pollers fail they will transfer to the Hot Standby server until the server is restored.

3. If multiple Pollers and/or the Main server fail but the Hot Standby will take - all, some, none - of the traffic from the other downed Pollers.

4. From what I understand the traffic will move - not instantly but over time - from the Poller/NPM server to the Hot Standby server until that device is restored. Since they use the same database there should be only minor gaps.

Thanks in advance.

savell

The hot-standby server can perform polling duties for a single failed poller at a time.

You configure the hot-standby to be either automatic or manual.

In the automatic case on the hotstandby server, you configure the pollers that you wish to perform a hotstandby function for, and the time parm (i.e. how long the poller needs to be down before the hotstandby starts up).

It might be worth asking Sales about the future of the hot-standby server. Now they are moving towards the jobengine type model of polling (and the non APM, IPSLA module support), there must be changes planned. I'm hoping that the hot-standby becomes redundant with some kind of ability to dynamically move nodes between polling engines (at least without stopping services as we do now).

I don't think it's anything to be concerned about however - in my experience SolarWinds are good at ensuring your investment is protected.  

Dave.