1

When trying to disable JVM swapping I disabled paging on a Windows machine and rebooted:

enter image description here

When looking at the cluster health via Elastic HQ, I see that the JVM swaps very much anyway:

enter image description here

Why is it so? Is the reading not accurate or is swap occuring despite the paging being 0 MB?

WoJ
  • 3,365
  • 8
  • 46
  • 75

1 Answers1

0

It's a bit weird, but any way you can disable the swapping behaviour by enabling ]mlockall.

mlockall will try to lock the process address space into RAM, preventing any Elasticsearch memory from being swapped out. This can be done, by adding this line to the config/elasticsearch.yml file:

bootstrap.mlockall: true

Elastic HQ uses the following method to get the swap value:

formula:"stats.os.swap.used_in_bytes / 1024 / 1024"

IIRC, ES uses sigar to offer "stats.os.swap.used_in_bytes" and sigar uses "Paging Files" performance counter to finally get the number of bytes of paging files used by the ES.

You can double check this counter value by issuing 

Get-Counter '\Paging Files\% Usage'

in windows power shell.

Enzo Wang
  • 121
  • 2