Flood stage exceeded is incorrectly logged

[PhaedrusTheGreek]

Tested on

• ES 6.0.0-beta1
• Mac OSX Sierra

Node is reporting flood stage exceeded:

[2017-08-14T09:29:37,167][WARN ][o.e.c.r.a.DiskThresholdMonitor] [Lq-N3tZ] flood stage disk watermark [95%] exceeded on [Lq-N3tZ8QY6B648xAA8H0g][Lq-N3tZ][/Users/jay/Workspace/elasticsearch-6.0.0-beta1/data/nodes/0] free: 11.1gb[9.9%], all indices on this node will marked read-only

But the math doesn't add up (flood stage is 95% and there's 9.9% free disk space

And indexing is still allowed, even on indices created before the flood stage log message appears.

{
    "_index": "foo",
    "_type": "bar",
    "_id": "7",
    "_version": 1,
    "result": "created",
    "_shards": {
        "total": 2,
        "successful": 1,
        "failed": 0
    },
    "_seq_no": 0,
    "_primary_term": 1
}

[jasontedor]

Is this a clean install, are you sure that you did not set the flood state setting to a disk space free instead of a disk percentage free? Can you show the output of GET /_cluster/settings?include_defaults=true&filter_path=**.flood_stage&pretty=true and GET /_nodes/settings?filter_path=**.flood_stage&pretty=true?

[PhaedrusTheGreek]

Seems that this line is comparing with the high watermark accidentally

[jasontedor]

You're right: #26204