Shard Allocation awareness stopped working

Hello. I configured forced shard allocation awareness on my cluster as described here some time earlier. And it was working well.

But now I find out that some indices with one shard and one replica allocated on the nodes with same label.
Nodes has approximately equal free space below 80% watermark. So I can't find another explanation of such cluster allocation behaviour.
Need help please.

I suggest the cluster allocation explain API :slightly_smiling_face: If you ask it to explain the allocation of one of the shards that you think is allocated wrongly then you should get the information you need. Share it here if you need help interpreting it.

Found a root of problem. Not all master nodes had attribute enumeration in config.

1 Like