Skip to content

davidgiga1993/pollect

Repository files navigation

pollect - python data collection daemon

pollect is a daemon for collecting system and application metrics in periodical intervals. (similar to collectd). It's designed to require very little dependencies and to be easily customizable.

Architecture

 ------------                           ----------
 | executor |  -- result of sources --> | writer |
 ------------                           ----------
       1
       | Calls sources to probe data
       n
 ---------------                                   
 |   source    |   
 ---------------

pollect uses executors which contain sources for probing the data. The data is exported using the collection name.

For persisting the data writers are used. They can be defined globally (for all executors) or on a per executor level.

By default, the tick time is defined globally, but can be changed on a executor level.

Usage

pip install pollect
pollect --config config.json [--dry-run]

Note: You can use either json or yml files for configuration.

Docker

Place your config.[json|yml] and any custom sources into your working directory and run

docker run -v $(pwd):/pollect -p 8000:8000 davidgiga1993/pollect:latest

Config

Here is an example for collecting the load average every 30 seconds as well as sampling the response time of google every 2 min and exporting it as prometheus metrics

---
tickTime: 30
threads: 4
writer:
  type: Prometheus
  port: 8000
executors:
  - collection: pollect
    sources:
      - type: LoadAvg
  - collection: slowerMetrics
    tickTime: 120
    sources:
      - type: Http # See Http below for more details
        url: https://google.com
        labels: # Additional labels/tags for the metrics
          # It is also possible to access env variables anywhere
          # in the config
          system: prod
          home: ${HOME}

A more advanced configuration sample can be found in the pollect.[json|yml] file.

Metric names

The metric names are automatically build out of the collection, source and value name. Example:

Pattern: ${collection}_${sourceType}[_${sourceName}][_$valueName}]

Collection Source type Source name (optional) Resulting metric name
pollect Http pollect_Http
pollect Http test pollect_Http_test
pollect Process pollect_Process_load_percent
pollect_Process_virtual_memory

Sources

A source collects data in regular intervals. Depending on the source there are multiple configuration parameters available.

Certain sources might need additional dependencies.

The following parameters are available for all sources:

Param Desc
name Name of the metric (prefix)
labels Dict of static labels

Http response time Http

Measures the http response time in milliseconds

Param Desc
url Url to the web service. Can be a list of strings as well (the url will be added as label)
timeout Timeout in seconds (default 15)
statusCode The expected status code (default any non error)
proxy Http proxy which should be used (defaults to none respecting environment variables. Set to '' to use no proxy regardless of environment)

Disk usage DiskUsage

Disk usage statistics. Requires shutil package

Load average LoadAvg

System load average. Linux only

Memory usage MemoryUsage

System memory usage. Requires psutil package

Process stats Process

Information about one or more processes

Param Desc
name Name of the metric
procRegex Name of the process(es) - Regex
memory True to enable memory metrics (default true)
load True to enable cpu load metrics (default true)

Interface Interface

Collects NIC statistics. Requires psutil package

Param Desc
includeTotal Enables incremental counter data
include Explicitly includes nics. Can be empty
exclude Excludes nics. Can be empty

IO IO

Collects IO statistics. Requires psutil package

Param Desc
include Explicitly includes disks. Can be empty
exclude Excludes disks. Can be empty

HDD smart data SmartCtl

Wrapper for the linux smartctl tool. Collects SMART data

Param Desc
attributes Name of smart attributes which should be included
devices List of regex for matching disks which should be included

Sensors Sensors

Wrapper for the linux sensors utility. Collects sensor data such as temps and voltages

Param Desc
include Name of chips which should be included. Can be empty
exclude Name of chips which should be excluded. Can be empty
useBaseUnits Set to False to report the raw values in their reported unit (for example mV instead of V)

DNS server statistics Bind

Bind DNS server statistics.

Param Desc
url URL to the bind statistics
views Views which should be included

SNMP SnmpGet

Wrapper for the snmpget binary.

type: SnmpGet
name: Procurve
# Host which should be contacted
host: 10.1.100.1
# Community string (public by default)
communityString: public
# Metrics which should be collected
# Each metric can query one or more oids
metrics:
  # Name of the metric
  - name: throughput
    # Processing mode:
    # - undefined (default): No processing
    # - rate: Computes the change per second, compensating for value overflows
    mode: rate
    # OID which should be probed
    oid: iso.3.6.1.2.1.16.1.1.1.4.1

  # It is also possible to define ranges
  - name: interface_link_state
    # The parameter in the oid will be replaced with the range number
    oid: iso.3.6.1.2.1.2.2.1.5.${if}
    range:
      from: 1
      to: 10 # From and to are inclusive
      label: "if" # The label will be attached to the metric

Plex server Plex

Collects plex mediaserver statistics. This requires local IP addresses to be allowed without authentication.

Param Desc
url URL to plex. Use the IP of the NIC instead of localhost

Plex server Zfs

Provides simple ZFS pool sizing and performance metrics.

Fritzbox WAN Fritzbox

Connects to the fritzbox api and collects WAN statistics. Requires the fritzconnection package.

EVCC Evcc

Exposes the values shown in the EVCC web-ui as metrics. Requires the websocket-client package.

Param Desc
host Host and port of the EVCC instance (e.g. localhost:7070)

Viessmann API Viessmann

Collects sensor data from viessmann heatpumps

Param Desc
user Username of viessmann account
password Password of account

These information are only required for the first data collection. Afterwards a viessmann_token.json file is created to cache the oauth credentials.

Homematic IP HomematicIp

Collects temperature and humidity data from homematic IP. Requires the homematicip package.

Param Desc
authToken Auth token
accessPoint Access point id

SMA Energy Meter SmaEnergyMeter

Collects data from the SMA Home Manager 2.0. The data is received via multicast from the meter. If this doesn't work for you, there is also a possibility to configure the meter to unicast the values directly to your machine.

Param Desc
hostIp The primary IP address of the host running pollect.
Required for multicast

SMA PV Modbus SmaPvModbus

Collects data from SMA Photovoltaik inverters. Requires tcp modbus to be enabled on the inverted.

Requires the pymodbus dependency.

Param Desc
host IP or hostname of the inverted
port Optional: modbus port, 502 by default

TP-LINK EAP TpLinkEap

Collects wifi statistics of the TP-LINK EAP series devices. This uses the rest api of the device. Note that the devices only support one session at a time, meaning you will be logged out from the web UI in regular intervals.

Param Desc
url URL to the admin UI
user Username
password Password

Openhab Openhab

Collects the values of all channels of all items in openhab. This contains more data than the original metrics exporter of openhab (since it doesn't include all items).

Param Desc
url URL to openhab

Audi MMI MMI

Connects to the audi MMI backend and collects data. Requires the audi api package. Note: This pacakge is currently broken due to API changes.

Param Desc
credentials Path to the credentials.json
vin VIN of the car that should be crawled

Google Play Developer Console Gdc

Provides app statistics from the google play developer console. Requires the google-cloud-storage package.

Important each fetch will call the google cloud storage api to check for updates so make sure to call is less frequent (every 30min or so).

Sample config:

{
  "type": "Gdc",
  // Name of the bucket - this can be found at the bottom of the bulk export page in the GDC
  "bucketName": "pubsite_prod_rev_1234",
  "apps": [
	{ // All apps for which the statistics should be crawled
	  "package": "my.app.package",
	  "name": "My App"
	}
  ],
  // Key file for the service account which has read access to the GDC
  "keyFile": "api-project-1234.json",
  // Name of the folder where the crawled files should be stored
  "dbDir": "db"
}

Apple appstore connect AppStoreConnect

Collects download statistics from apple

{
  "type": "AppStoreConnect",
  "keyId": "ASDH123123",
  "keyFile": "my_api_key.p8",
  "issuerId": "asdasd-123123asd-asd12312-asd",
  "vendorNumber": "882223",
  "dbDir": "db"
}

Http Ingress source HttpIngress

Requires the gevent package This source starts a simple http webserver and where you can post metrics to. It's intended if you want to push metrics to pollect, instead of using the default pull probes.

- type: HttpIngress
  name: Ingress
  port: 9005 # Listener port
  metrics: # You can define multiple metrics
    sample_metric: # Name of the metric
      type: counter # Optional, gauge by default, counter will cause the value to increment by X every time
      labels: # Labels for this metric
        - host

You can update the metrics using a simple http json post:

curl -X POST http://pollect:9005 \
-H 'Content-Type: application/json' \
--data-binary @- << EOF
{
    "metrics": {
      "sample_metric": {
        "value": 124
        "labels": {
          "host": "my-hostname"
        }
      }
    }
}
EOF

Certificate source Certificate

Returns the expiry date of a https certificate. Requires openssl binary and pyOpenSSL.

- type: Certificate
  name: cert
  url: https://google.com

Writers

A writer represents the destination where the collected data is written to.

Dry run DryRun

Prints the collected data to the stdout

Prometheus http exporter Prometheus

Exports the data via a prometheus endpoint. The port can be configured using portas configuration:

writer:
  type: Prometheus
  port: 9001

Https support

Pollect has a custom prometheus exporter which supports https. This requires the gevent package.

writer:
  type: PrometheusSsl
  port: 8000
  key: key.key
  cert: cert.pem

Multithreading

By default, pollect executes all sources of a collection in parallel with 5 threads. Different collections are executed in separate threads as well, meaning that multiple long-running probes in one collection can't block another collection. If the writer supports partial writes (for example prometheus) the result of each source will be immediately available. Writers which do not support partial writes will receive the data once all probes have completed.

Extensions

This example shows how to add your own collectors

Source

extensions/SingleRandom.py:

# Single random value with parameter
class SingleRandomSource(Source):
    def __init__(self, data):
        super().__init__(data)
        self.max = data.get('max')

    def _probe(self):
        return random() * self.max

extensions/MultiRandomSource.py:

# Multiple random values
class MultiRandomSource(Source):
    def _probe(self):
        return {'a': random(), 'b': random()}

config.json:

{
  "sources": [
	{
	  "type": "extensions.SingleRandom",
	  "max": 100
	},
	{
	  "type": "extensions.MultiRandom"
	}
  ]
}

A similar principle is used for the writers. Take a look at the sourcesand writers folders for more examples.