Hi Guys, I want to setup elastic cluster with failover management. I have two machines available. Please guide me about Master and Slave nodes optimum strategy, so that I should not face any downtime issue due to failure of a node or machine or network (inter nodes). I have little heard about Split Brain issue that I also want to avoid.
how big are your machines (ram/cpu/ssd ?) ? what is your user case ? 50/50 indexation/search or more indexation than search querys ? your querys will use heavy aggregations/sort ? (i have 38 nodes a never had split brain issue)
Thanks for coming up on this so fast.
These are 2 linux machines with 4 core processor, 8gb RAM and SSD drive 500 GB.
User Case is average like on peak it could be concurrent 1K-2K users load at max. But apart from query there would be data updates (indexation) in between.
And also using aggregations in queries.
Please also mention, is multiple node on same machine is good approach (if I have to, and what should be strategy for master/salve)
yeah can be a good approach but when your servers are bigger! with only 8gb or RAM you need tu put 4gb to the heap and that is ok, but if you put less memory it can be a litte dangerous if you have multiple index and too many indexations and search in the same moment, you need to test multiple options but you can beginning with 2 nodes one for each serveur. after that if (marvel/bigdesk) you see that your heap never go up 2gb you can split your servers.
If you want HA then you really want 3 nodes, it helps prevent split brain.