Global Stats with NFS

[CFT6Server]

We have the Log Server data saving to an NFS mount and looks like this is throwing the global stats off. We are not seeing the "Data Written" and "I/O Size" data. I am guessing this could be due to the nature of us moving the data location. Is this something we could adjust and fix?

global stats.JPG

[jolson]

This metric is pulled from an elasticsearch query. Please run the following on one of your problem nodes:

Code: Select all

curl -s -XGET 'http://localhost:9200/_nodes/stats?pretty' | grep disk

Let us know the results. My guess is that either elasticsearch can't calculate disk read/write actions via NFS or the GUI is having trouble communicating with the subsystem. I think the former is more likely.

[CFT6Server]

Looks like you are right. There are no disk stats. I ran the same command on a test node which had local disks and it is returning a number of disk statistics.

[jolson]

My guess is that the elasticsearch API does not support monitoring the disk stats on a remote NFS mount. This guess might be taken further based on a quote from the elasticsearch team:

Do not place the index on a remotely mounted filesystem (e.g. NFS or SMB/CIFS); use storage local to the machine instead.

I'm doing some further digging - I'll let you know what I find out.

[jolson]

Since elasticsearch cannot read the data here, we cannot report it properly in the GUI. I traced the code down to this section, view-able via github: https://github.com/elastic/elasticsearc ... monitor/fs

I suggest filing a bug report to the elasticsearch team, since I cannot find an existing bug regarding this issue. https://github.com/elastic/elasticsearch/issues

In the meantime, you could use something like iostatto monitor your disk performance.