Kibana data node monitoring stops after five minutes

[mbby]

I've a cluster with five nodes, running 8.4.1.
One of the nodes had write rejections. We restarted the node and everything went back to a working state.
But after five minutes no monitoring data was displayed anymore.
If I restart the node monitoring data is coming in but after five minutes it stops again (short time after the dashboard shows 0 unassigned shards).

Just to mention: we had to delete some indices today because we were nearly running out of disk space (e.g. .ds-filebeat...).

It looks like the cluster is still working as expected.
If I do a _cat/nodes?v I could see the node and the data (like CPU) changes.
Kibana reports 11 nodes but only shows 10 in the list.

If I click on enter setup mode, the node is shown but greyed out.
All nodes are marked as self monitored.

Does someone know what is broken and how to fix it?

[jessgarson]

Thanks for reaching out, @mbby. I have a follow up question, did you get any error messages? If, so what did they say?

[mbby]

Yes. Just found the log. It says:
failed to flush export bulks

[mbby]

[mbby]

Just seen: this node only has replicas. Don't know why.

[mbby]

The system recovered itself. I reduced the load. But it took some hours.

[jessgarson]

Thanks for sharing your solution, @mbby.