An ILM policy had a conflict with curator last night. It wasn't working because of closed index.
I reopen it and retry ilm policy and seems ok.
But my cluster still yellow and cannot reallocate shard because of this :
Data too large, data for [<transport_request>] would be [30660165828/28.5gb], which is larger than the limit of [30601641984/28.5gb], real usage: [30660162168/28.5gb], new bytes reserved: [3660/3.5kb], usages [request=0/0b, fielddata=7205127741/6.7gb, in_flight_requests=3660/3.5kb, accounting=10134822598/9.4gb]]; ], allocation_status[no_attempt]]]"
I think it is because of too hight heap size memory.
How can I reduce heap size memory ? Close indexes can reduce it ?