Process

The process collector can monitor various running processes in the system, acquire and analyze various metrics when the process is running, Including memory utilization rate, CPU time occupied, current state of the process, port of process monitoring, etc. According to various index information of process running, users can configure relevant alarms in Guance, so that users can know the state of the process, and maintain the failed process in time when the process fails.

Warning

Process collectors (whether objects or metrics) may consume a lot on macOS, causing CPU to soar, so you can turn them off manually. At present, the default collector still turns on the process object collector (it runs once every 5min by default).

Configuration¶

Preconditions¶

The process collector does not collect process metrics by default. To collect metrics-related data, set open_metric to true in host_processes.conf. For example:

[[inputs.host_processes]]
    ...
     open_metric = true

Collector Configuration¶

Host InstallationKubernetes

Go to the conf.d/host directory under the DataKit installation directory, copy host_processes.conf.sample and name it host_processes.conf. Examples are as follows:

[[inputs.host_processes]]
  # Only collect these matched process' metrics. For process objects
  # these white list not applied. Process name support regexp.
  # process_name = [".*nginx.*", ".*mysql.*"]

  # Process minimal run time(default 10m)
  # If process running time less than the setting, we ignore it(both for metric and object)
  min_run_time = "10m"

  ## Enable process metric collecting
  open_metric = false

  ## Enable listen ports tag, default is false
  enable_listen_ports = false

  ## Enable open files field, default is false
  enable_open_files = false

  ## only collect container-based process(object and metric)
  only_container_processes = false

  # Extra tags
  [inputs.host_processes.tags]
  # some_tag = "some_value"
  # more_tag = "some_other_value"
  # ...

Once configured, restart DataKit.

Can be turned on by ConfigMap Injection Collector Configuration or Config ENV_DATAKIT_INPUTS .

Can also be turned on by environment variables, (needs to be added as the default collector in ENV_DEFAULT_ENABLED_INPUTS):

ENV_INPUT_HOST_PROCESSES_OPEN_METRIC

Enable process metric collecting

Type: Boolean

input.conf: open_metric

Default: false
ENV_INPUT_HOST_PROCESSES_PROCESS_NAME

Whitelist of process

Type: List

input.conf: process_name

Example: .*datakit.*,guance
ENV_INPUT_HOST_PROCESSES_MIN_RUN_TIME

Process minimal run time

Type: Duration

input.conf: min_run_time

Default: 10m
ENV_INPUT_HOST_PROCESSES_ENABLE_LISTEN_PORTS

Enable listen ports tag

Type: Boolean

input.conf: enable_listen_ports

Default: false
ENV_INPUT_HOST_PROCESSES_TAGS

Customize tags. If there is a tag with the same name in the configuration file, it will be overwritten

Type: Map

input.conf: tags

Example: tag1=value1,tag2=value2
ENV_INPUT_HOST_PROCESSES_ONLY_CONTAINER_PROCESSES

Only collect container process for metric and object

Type: Boolean

input.conf: only_container_processes

Default: false
ENV_INPUT_HOST_PROCESSES_METRIC_INTERVAL

Collect interval on metric

Type: Duration

input.conf: metric_interval

Default: 30s
ENV_INPUT_HOST_PROCESSES_object_interval

Collect interval on object

Type: Duration

input.conf: object_interval

Default: 300s

Metric¶

For all of the following data collections, a global tag named host is appended by default (the tag value is the host name of the DataKit), or other tags can be specified in the configuration by [inputs.host_processes.tags]:

 [inputs.host_processes.tags]
  # some_tag = "some_value"
  # more_tag = "some_other_value"
  # ...

`host_processes`¶

Collect process metrics, including CPU/memory usage, etc.

Tags

Tag	Description
container_id	Container ID of the process, only supported Linux
host	Host name
pid	Process ID
process_name	Process name
username	Username

Metrics

Metric	Description
cpu_usage	CPU usage, the percentage of CPU occupied by the process since it was started. This value will be more stable (different from the instantaneous percentage of `top`) Type: float Unit: percent,percent
cpu_usage_top	CPU usage, the average CPU usage of the process within a collection cycle Type: float Unit: percent,percent
mem_used_percent	Memory usage percentage Type: float Unit: percent,percent
nonvoluntary_ctxt_switches	From /proc/[PID]/status. Context switches that nonvoluntary drop the CPU. Linux only Type: int Unit: count
open_files	Number of open files (Linux only) Type: int Unit: count
page_children_major_faults	Linux from /proc/[PID]/stat. The number of major page faults for this process. Linux only Type: int Unit: digital,B
page_children_minor_faults	Linux from /proc/[PID]/stat. The number of minor page faults for this process. Linux only Type: int Unit: digital,B
page_major_faults	Linux from /proc/[PID]/stat. The number of major page faults. Linux only Type: int Unit: digital,B
page_minor_faults	Linux from /proc/[PID]/stat. The number of minor page faults. Linux only Type: int Unit: digital,B
proc_read_bytes	Linux from /proc/[PID]/io, Windows from `GetProcessIoCounters()`. Read bytes from disk Type: int Unit: digital,B
proc_syscr	Linux from /proc/[PID]/io, Windows from `GetProcessIoCounters()`. Count of `read()` like syscall`. Linux&Windows only Type: int Unit: count
proc_syscw	Linux from /proc/[PID]/io, Windows from `GetProcessIoCounters()`. Count of `write()` like syscall`. Linux&Windows only Type: int Unit: count
proc_write_bytes	Linux from /proc/[PID]/io, Windows from `GetProcessIoCounters()`. Written bytes to disk Type: int Unit: digital,B
rss	Resident Set Size Type: int Unit: digital,B
threads	Total number of threads Type: int Unit: count
vms	Virtual memory size Type: int Unit: digital,B
voluntary_ctxt_switches	From /proc/[PID]/status. Context switches that voluntary drop the CPU, such as `sleep()/read()/sched_yield()`. Linux only Type: int Unit: count

Object¶

`host_processes`¶

Collect data on process objects, including process names, process commands, etc.

Tags

Tag	Description
container_id	Container ID of the process if the process is running in container, Linux only
host	Host name
name	Process object name field, consisting of `[host-name]_[pid]`
process_name	Process name
state	Process status. Linux only
username	Username

Metrics

Metric	Description
cmdline	Command line parameters for the process Type: string Unit: -
cpu_usage	CPU usage, the percentage of CPU occupied by the process since it was started. This value will be more stable (different from the instantaneous percentage of `top`) Type: float Unit: percent,percent
cpu_usage_top	CPU usage, the average CPU usage of the process within a collection cycle Type: float Unit: percent,percent
listen_ports	The port the process is listening on Type: string Unit: -
mem_used_percent	Memory usage percentage Type: float Unit: percent,percent
message	Process details Type: string Unit: -
nonvoluntary_ctxt_switches	From /proc/[PID]/status. Context switches that nonvoluntary drop the CPU. Linux only Type: int Unit: count
open_files	Number of open files (only supports Linux, and the `enable_open_files` option needs to be turned on) Type: int Unit: count
page_children_major_faults	Linux from /proc/[PID]/stat. The number of major page faults of it's child processes. Linux only Type: int Unit: digital,B
page_children_minor_faults	Linux from /proc/[PID]/stat. The number of minor page faults of it's child processes. Linux only Type: int Unit: digital,B
page_major_faults	Linux from /proc/[PID]/stat. The number of major page faults. Linux only Type: int Unit: digital,B
page_minor_faults	Linux from /proc/[PID]/stat. The number of minor page faults. Linux only Type: int Unit: digital,B
pid	Process ID Type: int Unit: -
proc_read_bytes	Linux from /proc/[PID]/io, Windows from `GetProcessIoCounters()`. Read bytes from disk Type: int Unit: digital,B
proc_syscr	Linux from /proc/[PID]/io, Windows from `GetProcessIoCounters()`. Count of `read()` like syscall`. Linux&Windows only Type: int Unit: count
proc_syscw	Linux from /proc/[PID]/io, Windows from `GetProcessIoCounters()`. Count of `write()` like syscall`. Linux&Windows only Type: int Unit: count
proc_write_bytes	Linux from /proc/[PID]/io, Windows from `GetProcessIoCounters()`. Written bytes to disk Type: int Unit: digital,B
rss	Resident set size Type: int Unit: digital,B
start_time	process start time Type: int Unit: timeStamp,msec
started_duration	Process startup time Type: int Unit: timeStamp,sec
state_zombie	Whether it is a zombie process Type: bool Unit: -
threads	Total number of threads Type: int Unit: count
vms	Virtual memory size Type: int Unit: digital,B
voluntary_ctxt_switches	From /proc/[PID]/status. Context switches that voluntary drop the CPU, such as `sleep()/read()/sched_yield()`. Linux only Type: int Unit: count
work_directory	Working directory (Linux only) Type: string Unit: -

Process

Configuration¶

Preconditions¶

Collector Configuration¶

Metric¶

host_processes¶

Object¶

host_processes¶

Is this page helpful? ×

`host_processes`¶

`host_processes`¶