That is. I need to monitor 5000 servers with metricbeat and I need help to know how many nodes and hardware is recommended.
Thank You!
Presumably, that would depend on how much log output these 5000 servers generate.
1 log line per minute isnt very much.
1 log line per second is noticable
5 log lines per second is "heavy" 
It’s about 5 or more log lines per second 
The only real answer here is to benchmark your workload yourself. It's definitely possible to index over 100,000 log lines per second on a fairly small cluster, but the details very much depend on the workload and the hardware you have available.