Elastic search fails every alternate day...
Because... 
Probably they are unhappy about something? My very first guess would be to look at logs...
Yeah, they are unhappy... but they don't tell what they want...
i have observed that after 5 days, they fail... do they get tired.
When i delete the indices, it works. But that's not a solution
I am interested to diagnose them. how can i check whats wrong.
I tried all the CAT command, everything looks green and happening... but shards are failing. 
If they are failing you probably have a trace in logs.