Rolling upgrade problem from 6.8 to 7.1.1

I think my cluster didn't work yet. It didn't respond for REST API such as GET _cat/nodes.

The other ES can't connect to the elected master. They just printed like this message.

[2019-06-20T18:18:34,434][INFO ][o.e.x.m.e.l.LocalExporter] [node-mdata-637] waiting for elected master node [{node-mdata-638}{UtHL00IMRzihkL_erqCHPA}{wUWHEoNeRkK66PRZNoaGNg}{172.31.6.38}{172.31.6.38:9300}{ml.machine_memory=33449705472, ml.max_open_jobs=20, xpack.installed=true}] to setup local exporter [default_local] (does it have x-pack installed?)

And I found a error message

[2019-06-20T18:39:05,786][WARN ][o.e.n.Node ] [node-mdata-637] timed out while waiting for initial discovery state - timeout: 30s

Why was this message printed even I set "cluster.publish.timeout: 90s" in elasticsearch.yml?