Kubernetes Metadata Module (mmkubernetes)

Module Name:

mmkubernetes

Author:

Tomáš Heinrich Rich Megginson <rmeggins@redhat.com>

Purpose

This module is used to add Kubernetes <https://kubernetes.io/> metadata to log messages logged by containers running in Kubernetes. It will add the namespace uuid, pod uuid, pod and namespace labels and annotations, and other metadata associated with the pod and namespace.

Note

This only works with log files in /var/log/containers/*.log (docker –log-driver=json-file, or CRI-O log files), or with journald entries with message properties CONTAINER_NAME and CONTAINER_ID_FULL (docker –log-driver=journald), and when the application running inside the container writes logs to stdout/stderr. This does not currently work with other log drivers.

For json-file and CRI-O logs, you must use the imfile module with the addmetadata=”on” parameter, and the filename must match the liblognorm rules specified by the filenamerules (filenamerules) or filenamerulebase (filenamerulebase) parameter values.

For journald logs, there must be a message property CONTAINER_NAME which matches the liblognorm rules specified by the containerrules (containerrules) or containerrulebase (containerrulebase) parameter values. The record must also have the message property CONTAINER_ID_FULL.

This module is implemented via the output module interface. This means that mmkubernetes should be called just like an action. After it has been called, there will be two new message properties: kubernetes and docker. There will be subfields of each one for the various metadata items: $!kubernetes!namespace_name $!kubernetes!labels!this-is-my-label, etc. There is currently only 1 docker subfield: $!docker!container_id. See https://github.com/ViaQ/elasticsearch-templates/blob/master/namespaces/kubernetes.yml and https://github.com/ViaQ/elasticsearch-templates/blob/master/namespaces/docker.yml for more details.

Configuration Parameters

Note

Parameter names are case-insensitive.

Module Parameters and Action Parameters

KubernetesURL

type

default

mandatory

obsolete legacy directive

word

https://kubernetes.default.svc.cluster.local:443

no

none

The URL of the Kubernetes API server. Example: https://localhost:8443.

tls.cacert

type

default

mandatory

obsolete legacy directive

word

none

no

none

Full path and file name of file containing the CA cert of the Kubernetes API server cert issuer. Example: /etc/rsyslog.d/mmk8s-ca.crt. This parameter is not mandatory if using an http scheme instead of https in kubernetesurl, or if using allowunsignedcerts=”yes”.

tls.mycert

type

default

mandatory

obsolete legacy directive

word

none

no

none

This is the full path and file name of the file containing the client cert for doing client cert auth against Kubernetes. This file is in PEM format. For example: /etc/rsyslog.d/k8s-client-cert.pem

tls.myprivkey

type

default

mandatory

obsolete legacy directive

word

none

no

none

This is the full path and file name of the file containing the private key corresponding to the cert tls.mycert used for doing client cert auth against Kubernetes. This file is in PEM format, and must be unencrypted, so take care to secure it properly. For example: /etc/rsyslog.d/k8s-client-key.pem

tokenfile

type

default

mandatory

obsolete legacy directive

word

none

no

none

The file containing the token to use to authenticate to the Kubernetes API server. One of tokenfile or token is required if Kubernetes is configured with access control. Example: /etc/rsyslog.d/mmk8s.token

token

type

default

mandatory

obsolete legacy directive

word

none

no

none

The token to use to authenticate to the Kubernetes API server. One of token or tokenfile is required if Kubernetes is configured with access control. Example: UxMU46ptoEWOSqLNa1bFmH

annotation_match

type

default

mandatory

obsolete legacy directive

array

none

no

none

By default no pod or namespace annotations will be added to the messages. This parameter is an array of patterns to match the keys of the annotations field in the pod and namespace metadata to include in the $!kubernetes!annotations (for pod annotations) or the $!kubernetes!namespace_annotations (for namespace annotations) message properties. Example: [“k8s.*master”,”k8s.*node”]

srcmetadatapath

type

default

mandatory

obsolete legacy directive

word

$!metadata!filename

no

none

When reading json-file logs, with imfile and addmetadata=”on”, this is the property where the filename is stored.

dstmetadatapath

type

default

mandatory

obsolete legacy directive

word

$!

no

none

This is the where the kubernetes and docker properties will be written. By default, the module will add $!kubernetes and $!docker.

allowunsignedcerts

type

default

mandatory

obsolete legacy directive

boolean

off

no

none

If “on”, this will set the curl CURLOPT_SSL_VERIFYPEER option to 0. You are strongly discouraged to set this to “on”. It is primarily useful only for debugging or testing.

skipverifyhost

type

default

mandatory

obsolete legacy directive

boolean

off

no

none

If “on”, this will set the curl CURLOPT_SSL_VERIFYHOST option to 0. You are strongly discouraged to set this to “on”. It is primarily useful only for debugging or testing.

de_dot

type

default

mandatory

obsolete legacy directive

boolean

on

no

none

When processing labels and annotations, if this parameter is set to “on”, the key strings will have their . characters replaced with the string specified by the de_dot_separator parameter.

de_dot_separator

type

default

mandatory

obsolete legacy directive

word

_

no

none

When processing labels and annotations, if the de_dot parameter is set to “on”, the key strings will have their . characters replaced with the string specified by the string value of this parameter.

filenamerules

type

default

mandatory

obsolete legacy directive

word

SEE BELOW

no

none

Note

This directive is not supported with liblognorm 2.0.2 and earlier.

When processing json-file logs, these are the lognorm rules to use to match the filename and extract metadata. The default value is:

rule=:/var/log/containers/%pod_name:char-to:_%_%namespace_name:char-to:_%_%contai\
ner_name_and_id:char-to:.%.log

Note

In the above rules, the slashes \ ending each line indicate line wrapping - they are not part of the rule.

filenamerulebase

type

default

mandatory

obsolete legacy directive

word

/etc/rsyslog.d/k8s_filename.rulebase

no

none

When processing json-file logs, this is the rulebase used to match the filename and extract metadata. For the actual rules, see filenamerules.

containerrules

type

default

mandatory

obsolete legacy directive

word

SEE BELOW

no

none

Note

This directive is not supported with liblognorm 2.0.2 and earlier.

For journald logs, there must be a message property CONTAINER_NAME which has a value matching these rules specified by this parameter. The default value is:

rule=:%k8s_prefix:char-to:_%_%container_name:char-to:.%.%container_hash:char-to:\
_%_%pod_name:char-to:_%_%namespace_name:char-to:_%_%not_used_1:char-to:_%_%not_u\
sed_2:rest%
rule=:%k8s_prefix:char-to:_%_%container_name:char-to:_%_%pod_name:char-to:_%_%na\
mespace_name:char-to:_%_%not_used_1:char-to:_%_%not_used_2:rest%

Note

In the above rules, the slashes \ ending each line indicate line wrapping - they are not part of the rule.

There are two rules because the container_hash is optional.

containerrulebase

type

default

mandatory

obsolete legacy directive

word

/etc/rsyslog.d/k8s_container_name.rulebase

no

none

When processing json-file logs, this is the rulebase used to match the CONTAINER_NAME property value and extract metadata. For the actual rules, see containerrules.

busyretryinterval

type

default

mandatory

obsolete legacy directive

integer

5

no

none

The number of seconds to wait before retrying operations to the Kubernetes API server after receiving a 429 Busy response. The default “5” means that the module will retry the connection every 5 seconds. Records processed during this time will _not_ have any additional metadata associated with them, so you will need to handle cases where some of your records have all of the metadata and some do not.

If you want to have rsyslog suspend the plugin until the Kubernetes API server is available, set busyretryinterval to “0”. This will cause the plugin to return an error to rsyslog.

sslpartialchain

type

default

mandatory

obsolete legacy directive

boolean

off

no

none

This option is only available if rsyslog was built with support for OpenSSL and only if the X509_V_FLAG_PARTIAL_CHAIN flag is available. If you attempt to set this parameter on other platforms, you will get an INFO level log message. This was done so that you could use the same configuration on different platforms. If “on”, this will set the OpenSSL certificate store flag X509_V_FLAG_PARTIAL_CHAIN. This will allow you to verify the Kubernetes API server cert with only an intermediate CA cert in your local trust store, rather than having to have the entire intermediate CA + root CA chain in your local trust store. See also man s_client - the -partial_chain flag. If you get errors like this, you probably need to set sslpartialchain=”on”:

rsyslogd: mmkubernetes: failed to connect to [https://...url...] -
60:Peer certificate cannot be authenticated with given CA certificates

cacheexpireinterval

type

default

mandatory

obsolete legacy directive

integer

-1

no

none

This parameter allows you to expire entries from the metadata cache. The values are:

  • -1 (default) - disables metadata cache expiration

  • 0 - check cache for expired entries before every cache lookup

  • 1 or higher - the number is a number of seconds - check the cache for expired entries every this many seconds, when processing an entry

The cache is only checked if processing a record from Kubernetes. There isn’t some sort of housekeeping thread that continually runs cleaning up the cache. When an record from Kubernetes is processed:

If cacheexpireinterval is -1, then do not check for cache expiration. If cacheexpireinterval is 0, then check for cache expiration. If cacheexpireinterval is greater than 0, check for cache expiration if the last time we checked was more than this many seconds ago.

When cache expiration is checked, it will delete all cache entries which have a ttl less than or equal to the current time. The cache entry ttl is set using the cacheentryttl.

cacheentryttl

type

default

mandatory

obsolete legacy directive

integer

3600

no

none

This parameter allows you to set the maximum age (time-to-live, or ttl) of an entry in the metadata cache. The value is in seconds. The default value is 3600 (one hour). When cache expiration is checked, if a cache entry has a ttl less than or equal to the current time, it will be removed from the cache.

This option is only used if cacheexpireinterval is 0 or greater.

This value must be 0 or greater, otherwise, if cacheexpireinterval is 0 or greater, you will get an error.

Statistic Counter

This plugin maintains per-action statistics. The statistic is named “mmkubernetes($kubernetesurl)”, where $kubernetesurl is the KubernetesURL setting for the action.

Parameters are:

  • recordseen - number of messages seen by the action which the action has determined have Kubernetes metadata associated with them

  • namespacemetadatasuccess - the number of times a successful request was made to the Kubernetes API server for namespace metadata.

  • namespacemetadatanotfound - the number of times a request to the Kubernetes API server for namespace metadata was returned with a 404 Not Found error code - the namespace did not exist at that time.

  • namespacemetadatabusy - the number of times a request to the Kubernetes API server for namespace metadata was returned with a 429 Busy error code - the server was too busy to send a proper response.

  • namespacemetadataerror - the number of times a request to the Kubernetes API server for namespace metadata was returned with some other error code not handled above. These are typically “hard” errors which require some sort of intervention to fix e.g. Kubernetes server down, credentials incorrect.

  • podmetadatasuccess - the number of times a successful request was made to the Kubernetes API server for pod metadata.

  • podmetadatanotfound - the number of times a request to the Kubernetes API server for pod metadata was returned with a 404 Not Found error code - the pod did not exist at that time.

  • podmetadatabusy - the number of times a request to the Kubernetes API server for pod metadata was returned with a 429 Busy error code - the server was too busy to send a proper response.

  • podmetadataerror - the number of times a request to the Kubernetes API server for pod metadata was returned with some other error code not handled above. These are typically “hard” errors which require some sort of intervention to fix e.g. Kubernetes server down, credentials incorrect.

  • podcachenumentries - the number of entries in the pod metadata cache.

  • namespacecachenumentries - the number of entries in the namespace metadata cache.

  • podcachehits - the number of times a requested entry was found in the pod metadata cache.

  • namespacecachehits - the number of times a requested entry was found in the namespace metadata cache.

  • podcachemisses - the number of times a requested entry was not found in the pod metadata cache, and had to be requested from Kubernetes.

  • namespacecachemisses - the number of times a requested entry was not found in the namespace metadata cache, and had to be requested from Kubernetes.

Fields

These are the fields added from the metadata in the json-file filename, or from the CONTAINER_NAME and CONTAINER_ID_FULL fields from the imjournal input:

$!kubernetes!namespace_name, $!kubernetes!pod_name, $!kubernetes!container_name, $!docker!id, $!kubernetes!master_url.

If mmkubernetes can extract the above fields from the input, the following fields will always be present. If they are not present, mmkubernetes failed to look up the namespace or pod in Kubernetes:

$!kubernetes!namespace_id, $!kubernetes!pod_id, $!kubernetes!creation_timestamp, $!kubernetes!host

The following fields may be present, depending on how the namespace and pod are defined in Kubernetes, and depending on the value of the directive annotation_match:

$!kubernetes!labels, $!kubernetes!annotations, $!kubernetes!namespace_labels, $!kubernetes!namespace_annotations

More fields may be added in the future.

Error Handling

If the plugin encounters a 404 Not Found in response to a request for namespace or pod metadata, that is, the pod or namespace is missing, the plugin will cache that result, and no metadata will be available for that pod or namespace forever. If the pod or namespace is recreated, you will need to restart rsyslog in order to clear the cache and allow it to find that metadata.

If the plugin gets a 429 Busy response, the plugin will _not_ cache that result, and will _not_ add the metadata to the record. This can happen in very large Kubernetes clusters when you run into the upper limit on the number of concurrent Kubernetes API service connections. You may have to increase that limit. In the meantime, you can control what the plugin does with those records using the busyretryinterval setting. If you want to continue to process the records, but with incomplete metadata, set busyretryinterval to a non-zero value, which will be the number of seconds after which mmkubernetes will retry the connection. The default value is 5, so by default, the plugin will retry the connection every 5 seconds. If the 429 error condition in the Kubernetes API server is brief and transient, this means you will have some (hopefully small) number of records without the metadata such as the uuids, labels, and annotations, but your pipeline will not stop. If the 429 error condition in the Kubernetes API server is persistent, it may require Kubernetes API server administrator intervention to address, and you may want to use the busyretryinterval value of “0”. This will cause the module to return a “hard” error (see below).

For other errors, the plugin will assume they are “hard” errors requiring admin intervention, return an error code, and rsyslog will suspend the plugin. Use the Statistic Counter to monitor for problems getting data from the Kubernetes API service.

Example

Assuming you have an imfile input reading from docker json-file container logs managed by Kubernetes, with addmetadata=”on” so that mmkubernetes can get the basic necessary Kubernetes metadata from the filename:

input(type="imfile" file="/var/log/containers/*.log"
      tag="kubernetes" addmetadata="on")

(Add reopenOnTruncate=”on” if using Docker, not required by CRI-O).

and/or an imjournal input for docker journald container logs annotated by Kubernetes:

input(type="imjournal")

Then mmkubernetes can be used to annotate log records like this:

module(load="mmkubernetes")

action(type="mmkubernetes")

After this, you should have log records with fields described in the Fields section above.

Credits

This work is based on https://github.com/fabric8io/fluent-plugin-kubernetes_metadata_filter and has many of the same features.

See also

Help with configuring/using Rsyslog:

See also

Contributing to Rsyslog:

Copyright 2008-2023 Rainer Gerhards (Großrinderfeld), and Others.