Introduce container awareness

[druesendieb]

trafficstars

Currently, Splunk is not aware of the cgroup it's running in and the assigned resources. This results in several problems:

• Resource utilization in Monitoring Console shows not the containers cgroups resources but the node's resources.
• If the cgroups set resource allocation is exceeded, searches will be ungracefully killed by the oom killer of the host OS. This can happen as Splunk thinks it has the node's resources available but is limited to the cgroups set resources.
• Workload management can't be used out of the box.

Solving this problem would help solving at least these as well:

• #375
• https://github.com/splunk/splunk-operator/issues/445
• https://github.com/splunk/splunk-operator/issues/76

[druesendieb]

Theres a Splunk idea which would help tremendously with solving the outcome of this problem: Ability to manually set CPU and RAM limits

[BGrasnick]

Any progress on this? It is a huge issue for our deployment (oomkill for big searches) and should be fixed asap

[druesendieb]

Any progress on this? It is a huge issue for our deployment (oomkill for big searches) and should be fixed asap

Neither this nor any of the linked Issues got public attention since posting this I am afraid :(

[pogdin]

CSPL-1481