EIGRP router failure

Not sure that I understand topology (drawing would be nice), but typically, if you have Feasible Successor route in EIGRP topology database impact should be under 1 second. As soon as connection to Successor is lost route to Feasible Successor will be placed in routing table as new Successor route. This is the fastest convergence time that you can get with almost no impact on network at all (especially with bfd configured).

In This case R1 is at the Data Center , R3 is at the Disaster Recovery Site, the rest of the routers at the remote site
First , I want to know if best practice is to make R1 as Successor and R3 Feasible Successor.
Then if any of Remote Sites router crashes, whether it will impact the Data Center

First , I want to know if best practice is to make R1 as Successor and R3 Feasible Successor. and how to do it ?

Then if any of Remote Sites router crashes, whether it will impact the Data Center. I mean most our servers  are in the Data Center

if best practice is to make R1 as Successor and R3 Feasible Successor

Sure it is. You can manipulate routes by changing bandwidth and delay to adjust route costs, or you can use offset-lists to make routes more or less attractive.

What about this ?

Then if any of Remote Sites router crashes, whether it will impact the Data Center. I mean most our servers  are in the Data Center

If you have Feasible Successor for every significant route on your routers, you can route around anything under 1 second. if I remember correctly that fastest reroute time was around 200 ms (including dead neighbor detection with bfd). So, if configured properly there should be no significant impact.

I thought if he remote site fails completely..it will not impact other sites especially the Data Center ?

If you can change routing around currently used route fast enough (and there is no NAT) traffic path is typically not important at all. And Feasible Successor should give you "fast enough" switch to other path to location. The thing is - feasible successor is already in topology, but path is just not in use (by default). So router will just switch to that route - that is the reason for minimal impact in this case.

All our servers are in the DataCenter (behind Router 1), I do not see in which case the  Failure on remote sites will impact the Data Center.

In your topology that you provided there should be no problem, but in some cases you can have problem with EIGRP Stuck in active. That is situation when router loses path to some network and looses connection to that network. If there is many routers and answer from neighbors is not back fast enough (either alternative path provided to network or network is lost), in that case all neighbor relationships gets broken and are reestablished again. However - that problem is not solved by having feasible successor ready. :)
Looks like I went in wrong direction - fastest way to reroute by usage of feasible successor.

I believe an Outage on the remote site, will still cause EIGRP recalculation on all routers if they are in the same AS
Probably that time of recalculation can impact the servers.
I would like to know if there is a  way to isolate the remote sites failures from impacting the Datacenter...

Thanks Jovic

You're welcome.