[Metricbeat] Process modules produces high load / cpu usage

[adsr]

Hi,

after upgrading metricbeat from 7.11.2 to 7.16.3 I notice very high load/cpu usage for the metricbeat process.
I narrowed it down to the process module in combination with a python process.
Furthermore I tried several versions of which at least 7.16.x, 7.15.x and 7.14.x are affected

• Version: 7.16.3 (+7.15.x, +7.14.x)
• Operating System: CentOS 7

Config:

#==========================  Modules configuration ============================

metricbeat.modules:
  - module: system
    metricsets:
      - cpu
      - core
      - load
      - memory
      - network
      - process
    ticks: False
    enabled: True
    period: 10s
    process.cgroups.enabled: True
    processes:
      - .*
  - module: system
    metricsets:
      - diskio
      - uptime
    enabled: True
    period: 60s
  - module: system
    metricsets:
      - filesystem
      - fsstat
    enabled: True
    period: 60s
    filesystem.ignore_types: []

  - module: docker
    metricsets:
      - cpu
      - memory
    enabled: True
    period: 10s
    hosts:
      - "unix:///var/run/docker.sock"
  - module: docker
    metricsets:
      - container
      - event
      - healthcheck
      - info
      - network
    enabled: True
    period: 60s
    hosts:
      - "unix:///var/run/docker.sock"


#================================ General =====================================

# The 'name' property refers to the host name
# This field is added as "host.name" to the event
name: my-host
fields:
  env: production
tags:
  - myhost
max_procs: 2

#============================== Processors ====================================

processors:
  - add_host_metadata:
      netinfo.enabled: true
      cache.ttl: 60m

#================================ Outputs =====================================

output.logstash:
  enabled: true
  hosts:
    - logstash:6033
  index: metricbeat
  compression_level: 3
  worker: 1
  loadbalance: true
  pipelining: 0
  slow_start: true

output.console:
  enabled: false
  pretty: true

#================================ Logging =====================================

logging.level: info
logging.selectors: []

logging.to_syslog: false
logging.to_files: true
logging.files:
  rotateeverybytes: 10485760 # = 10MB
  keepfiles: 7

Any input is highly appreciated!

Workarounds:

• downgrade to 7.13.4
• disable the process module