On 6/2/2014 11:15 AM, Geoffrey Carman wrote:
> If the Remote Loader is running on 2 DC's, one for cold failover, and
> you use a DNS name for the server name, and when you fail over, you
> change the DNS entry to point at the cold server.
> What happens to the Password filter events, if they are pointed at a DNS
> entry and the DNS changes underneath it?
> Will it start forwarding to the new RL once DNS changes and it notices?
> Anyone ever tried that? I have not, but it seems like an easy way to
> avoid all the DC server reboots.

Another thought just hit me.

Two parameters:
Password Sync Timeout - how long RL stores passwords without forwarding
to engine.
DC Passwords TimeToLive - how DC's cache a password without forwarding
to RL before discarding them.

These are stored in DirXML-ShimConfigInfo attribute. This is flagged Per
Replica in schema, which means this can differ per replica.

Assume you run in the model where Password Filter is sending events to 2
RL's (Main and failover instance).

Can you have the two engines send different values for that, so as it
forwards to RL1 (Main RL) have the DC time to live set to like 1 hour,
so you can survive a normal outage issue. But the forwards to RL2 (the
DR RL) set to 5 min, so you avoid the replay issue, but also avoid
losing events during short outages...