This content has been marked as final. Show 2 replies
On one of my decent size installation we had the same two options that you are thinking about. The index was highly available (4 dgraphs: 2 copies across 2 data centers), there was a 15 sec ping on the load balances and a program to hit Endeca and Corda every X minutes (X=5-20 mins, depending on the time of day) that would send out an email/page if any of the nodes was down.
Do you have any specific concerns or requirements?
This is real problem and I even see when dgraph goes into poor performance state.
Ping donot help because it just says server is up which is really up and running but very poorly responding query.
I think endeca recommends to use least connection algo, but in this case no much help which is resulting few query just bad experience.