Kubernetes resource requirements

The resource requirements for the IBM Cloud® Monitoring agent depend on the size and load of the monitored host. The greater the activity the greater the resource requirements.

Monitoring can consume between 5 KB to 20 KB of bandwidth. The following can affect the bandwidth required:

The number of metrics
The number of events
The number of Kubernetes objects
The enabled products and features

When data is collected you will see a spike in bandwidth usage while the file is being ingested.

You do not want to place bandwidth shaping or caps on the agent. Doing so can affect data sent to the collection service.

In general, in larger clusters, the agent requires more memory. In servers with a larger number of cores the agent requires more CPU cores to monitor system calls. CPU cores are used on the host and Kubernetes nodes visible to the agent as proxies.

Consider the following environment definitions.

Table 1. Example environment sizings
Environment size	Description
Small	8 or less CPU cores 10 or less Kubernetes nodes
Medium	Between 9 and 32 CPU cores Between 10 and 100 Kubernetes nodes
Large	More than 32 CPU cores More than 100 Kubernetes nodes

The following recommendations can improve behavior instead of when you use default values. Your environment might require different resources depending on your requirements. The values should only be used for general information.

Table 2. Configuration considerations based on environment sizes
Cluster configuration	Small	Medium	Large
Kubernetes CPU request	1	3	5
Kubernetes CPU limit	1	3	5
Kubernetes Memory Request	1024 MB	3072 MB	6144 MB
Kubernetes Memory Limit	1024 MB	3072 MD	6144 MB
Dragent Memory Watchdog	512 MB	1024 MB	2048 MB
Cointerface Memory Watchdog	512 MB	2048 MB	4096 MB

The Monitoring agent has its own memory watchdog to prevent runaway memory consumption on the host in case of memory leaks. The agent configuration file can be modified to meet your environment needs.

By default the following is configured:

watchdog:
  max_memory_usage_mb: 1024
  max_memory_usage_subprocesses:
    sdchecks: 128
    sdjagent: 256
    mountedfs_reader: 32
    statsite_forwarder: 32
    cointerface: 512
    promscrape: 640

The max_memory_usage value corresponds to the dragent process in the agent. All values are specified in MB. For example if you want to change the configuration to match the watchdog settings for a large environment you would specify the following:

watchdog:
  max_memory_usage_mb: 2048
  max_memory_usage_subprocesses:
    sdchecks: 128
    sdjagent: 256
    mountedfs_reader: 32
    statsite_forwarder: 32
    cointerface: 4096
    promscrape: 640

The value for promscape depends on the amount of timeseries and label data to be scraped on a particular node. The cluster size does not affect promscape memory usage.