Skip to content

Latest commit

 

History

History
676 lines (458 loc) · 19.6 KB

README.rst

File metadata and controls

676 lines (458 loc) · 19.6 KB

journalpump BuildStatus

https://codecov.io/gh/aiven/journalpump/branch/master/graph/badge.svg?token=nLr7M7hvCx

journalpump is a daemon that takes log messages from journald and pumps them to a given output. Currently supported outputs are Elasticsearch, Apache Kafka®, logplex, rsyslog, websocket and AWS CloudWatch. It reads messages from journald and optionally checks if they match a config rule and forwards them as JSON messages to the desired output.

Building

To build an installation package for your distribution, go to the root directory of a journalpump Git checkout and then run:

Debian:

make deb

This will produce a .deb package into the parent directory of the Git checkout.

Fedora:

make rpm

This will produce an RPM in rpm/RPMS/noarch/.

Other:

python3 setup.py bdist_egg

This will produce an egg file into a dist directory within the same folder.

For a source install the dependency python-systemd has to be installed through your distribution's package manager (The PyPI systemd package is not the same!).

journalpump requires Python 3.4 or newer.

Installation

To install it run as root:

Debian:

dpkg -i ../journalpump*.deb

Fedora:

su -c 'dnf install rpm/RPMS/noarch/*'

On Fedora it is recommended to simply run journalpump under systemd:

systemctl enable journalpump.service

and eventually after the setup section, you can just run:

systemctl start journalpump.service

Other:

python3 setup.py install

On systems without systemd it is recommended that you run journalpump within supervisord or similar process control system.

Setup

After installation you need to create a suitable JSON configuration file for your installation.

General notes

If correctly installed, journalpump comes with a single executable, journalpump that takes as an argument the path to journalpump's JSON configuration file.

journalpump is the main process that should be run under systemd or supervisord.

While journalpump is running it may be useful to read the JSON state file that will be created as journalpump_state.json to the current working directory. The JSON state file is human readable and should give an understandable description of the current state of the journalpump.

Top level configuration

Example:

{
    "log_level": "INFO",
    "field_filters": {
       ...
    },
    "unit_log_levels": {
       ...
    },
    "json_state_file_path": "/var/lib/journalpump/journalpump_state.json",
    "readers": {
       ...
    },
    "statsd":   {
        "host": "127.0.0.1",
        "port": 12345,
        "prefix": "user-",
        "tags": {
            "sometag": "somevalue"
        }
    }
}

json_state_file_path (default "journalpump_state.json")

Location of a JSON state file which describes the state of the journalpump process.

statsd (default null)

Enables metrics sending to a statsd daemon that supports the influxdb-statsd / telegraf syntax with tags.

The tags setting can be used to enter optional tag values for the metrics.

The prefix setting can be used to enter an optional prefix for all metric names.

Metrics sending follows the Telegraf spec.

log_level (default "INFO")

Determines log level of journalpump. Available log levels.

Field filter configuration

Field filters can be used to restrict the journald fields that journalpump sends forward. Field filter configuration structure:

{
    "field_filters": {
        "filter_name": {
            "type": "whitelist|blacklist",
            "fields": ["field1", "field2"]
        }
    }
}

filter_name

Name of the filter. The filters can be configured per sender and depending on the use case the filters for different senders may vary.

type (default whitelist)

Specifies whether the listed fields will be included (whitelist) or excluded (blacklist).

fields

The actual fields to include or exclude. Field name matching is case insensitive and underscores in the beginning of the fields are trimmed.

Unit log levels configuration

Unit log levels can be used to specify which log levels you want to set on a per unit basis. Matching supports glob patterns. For example, to only process messsages for a systemd-unit called test-unit with severity WARNING or higher, your config could look like this:

{
    "unit_log_levels": {
        "log_level_name": [
            {
                "service_glob": "test-unit*",
                "log_level": "WARNING"
            },
            {
                "service_glob": "*-unit",
                "log_level": "INFO"
            }
        ]
    }
}

Note that if your unit would match multiple patterns (like "test-unit" would in the example above), the first match will get used, i.e "WARNING" in this case.

log_level_name

Name of the log level configuration. This can be configured per sender and depending on the use case the settings for different senders may vary.

Reader configuration

Reader configuration structure:

{
    "readers": {
        "some_reader": {
            "senders": {
                "some_log": {
                    ...
                },
                "another_log": {
                    ...
                }
            }
        },
        "another_reader": {
            "senders": {
                "some_kafka": {
                    ...
                }
            }
        }
    }
}

Example configuration for a single reader:

{
    "field_filters": {
        "drop_process_id": {
            "fields": ["process_id"],
            "type": "blacklist"
        }
    },
    "unit_log_levels": {
        "drop_everything_below_warning": [
            {
                "service_glob": "*",
                "log_level": "WARNING"
            }
        ]
    },
    "journal_path": "/var/lib/machines/container1/var/log/journal/b09ffd62229f4bd0829e883c6bb12c4e",
    "senders": {
        "k1": {
            "output_type": "kafka",
            "field_filter": "drop_process_id",
            "unit_log_level": "drop_everything_below_warning",
            "ca": "/etc/journalpump/ca-bundle.crt",
            "certfile": "/etc/journalpump/node.crt",
            "kafka_address": "kafka.somewhere.com:12345",
            "kafka_topic": "journals",
            "keyfile": "/etc/journalpump/node.key",
            "ssl": true
        },
    },
    "searches": [
        {
            "fields": {
                "MESSAGE": "kernel: Out of memory: Kill process .+ \\((?P<process>[^ ]+)\\)"
            },
            "name": "journal.oom_killer"
        }
    ],
    "secret_filter_metrics": true,
    "secret_filters": [
      {
        "pattern": "SENSITIVE",
        "replacement": "[REDACTED]"
      }],
    "threshold_for_metric_emit": 10
    "tags": {
        "type": "container"
    }
}

initial_position (default head)

Controls where the readers starts when the journalpump is launched for the first time:

  • head: First entry in the journal
  • tail: Last entry in the journal
  • <integer>: Seconds from current boot session

match_key (default null)

If you want to match against a single journald field, this configuration key defines the key to match against.

match_value (default null)

If you want to match against a single journald field, this configuration key defines the value to match against. Currently only equality is allowed. Note this means if you specify match_key and not match_value, then the reader will match all entries that do not contain the match_key.

msg_buffer_max_length (default 50000)

How many journal entries to read at most into a memory buffer from which the journalpump feeds the configured logsender.

journal_path (default null)

Path to the directory containing journal files if you want to override the default one.

journal_namespace (default null - read from default systemd namespace)

Journal namespace to read logs from. This feature requires latest version of python-systemd with namespace support

units_to_match (default [])

Require that the logs message matches only against certain _SYSTEMD_UNITs. If not set, we allow log events from all units.

flags (default LOCAL_ONLY)

"LOCAL_ONLY" opens journal on local machine only; "RUNTIME_ONLY" opens only volatile journal files; and "SYSTEM" opens journal files of system services and the kernel, "CURRENT_USER" opens files of the current user; and "OS_ROOT" is used to open the journal from directories relative to the specified directory path or file descriptor. Multiple flags can be OR'ed together using a list: ["LOCAL_ONLY", "CURRENT_USER"].

secret_filters (default [])

Secret filters can be used to redact sensitive data which matches known patterns in logs before forwarding the message along to it's final destination. To use: add a number of filters following the pattern below to the reader config. The pattern is a standard python regex, and the matching substring will be subbed with replacement. Patterns are compiled at runtime.

Simple pattern example:

This simple pattern should be used for most cases. It will replace SECRET with [REDACTED] but will leave the rest of the message intact.

"secret_filters": [
{
"pattern": "SECRET", "replacement": "[REDACTED]"

}

]

Complex pattern example:

For more complex requirements, a python regex with capture groups can be provided, and the contents of the message restructured using backrefs. This example will only replace SENSITIVE with [REDACTED] as long as foo and bar are also part of the pattern.

"secret_filters": [
{
"pattern": "(bar)(SENSITIVE)(foo)", "replacement": "\1[REDACTED]\3",

}

]

Using backrefs, the message can also be restructured into a new format. "secret_filters": [

{
"pattern": "(bar)(SENSITIVE)(foo)", "replacement": "\1\3 pattern was [REDACTED]",

}

]

Secret filters and searches can be made to use re2 as a regex engine by running journalpump with the environment "USE_RE2=yes". Make sure that the PyPI package "google_re2" is installed with at least version 1.1

secret_filter_metrics ( default: false) Change this setting to true to emit metrics to the metrics host whenever a secret pattern is matched. This matching happens before other filtering to help catch secrets being leaked to disk.

threshold_for_metric_emit ( default: 10) For the regex searches in journalpump, if search takes longer than this value, default 10 seconds, a metric will be emitted. type: int unit: second

Sender Configuration

output_type (default null)

Output to write journal events to. Options are elasticsearch, kafka, file, websocket and logplex.

field_filter (default null)

Name of the field filter to apply for this sender, if any.

File Sender Configuration

Writes journal entries as JSON to a text file, one entry per line.

file_output sets the path to the output file.

Elasticsearch Sender Configuration

ca (default null)

Elasticsearch Certificate Authority path, needed when you're using Elasticsearch with self-signed certificates.

elasticsearch_index_days_max (default 3)

Maximum number of days of logs to keep in Elasticsearch. Relevant when using output_type elasticsearch.

elasticsearch_index_prefix (default journalpump)

Elasticsearch index name to use when Maximum number of days of logs to keep in Elasticsearch. Relevant when using output_type elasticsearch.

elasticsearch_timeout (default 10.0)

Elasticsearch request timeout limit. The default should work for most people but you might need to increase it in case you have a large latency to server or the server is very congested. Required when using output_type elasticsearch.

elasticsearch_url (default null)

Fully qualified elasticsearch url of the form https://username:[email protected]:port. Required when using output_type elasticsearch.

Apache Kafka Sender Configuration

ca (default null)

Apache Kafka Certificate Authority path, needed when you're using Kafka with SSL authentication.

certfile (default null)

Apache Kafka client certificate path, needed when you're using Kafka with SSL authentication.

kafka_api_version (default 0.9)

Which Apache Kafka server API version to use.

kafka_topic (default null)

Which Kafka topic do you want the journalpump to write to. Required when using output_type kafka.

kafka_topic_config (default null)

If this key is present, its value must be another mapping with the default configuration used to create the topic, if it does not exist yet.

The mapping must have these values:

{
    "num_partitions": 3,
    "replication_factor": 3
}

kafka_address (default null)

The address of the Kafka server which to write to. Required when using output_type kafka.

kafka_msg_key (default null)

The key to use when writing messages into Kafka. Can be used for partition selection.

keyfile (default null)

Kafka client key path, needed when you're using Kafka with SSL authentication.

socks5_proxy (default null)

Defined socks5 proxy to use for Kafka connections. This feature is currently only supported in Aiven fork of kafka-python library.

AWS CloudWatch Logs Sender Configuration

aws_cloudwatch_log_group

The log group used in AWS CloudWatch.

aws_cloudwatch_log_stream

The log stream used in AWS CloudWatch.

aws_region (default null)

AWS region used.

aws_access_key_id (default null)

AWS access key id used.

aws_secret_access_key (default null)

AWS secret access key used.

The AWS credentials and region are optional. In case they are not included credentials are configured automatically by the botocore module.

The AWS credentials that are used need the following permissions: logs:CreateLogGroup, logs:CreateLogStream, logs:PutLogEvents and logs:DescribeLogStreams.

Google Cloud Logging Sender Configuration

google_cloud_logging_project_id

The GCP project id to which logs will be sent.

google_cloud_logging_log_id

The log id to be used for this particular sender.

google_cloud_logging_resource_labels

A dictionary containing the labels added to the monitored resource. Find the allowed labels from https://cloud.google.com/monitoring/api/resources#tag_generic_node.

google_service_account_credentials

The service account credentials to be used for this sender. If not defined, the sender will try to find credentials from the system.

Rsyslog Sender Configuration

rsyslog_server (default null)

Address of the remote syslog server.

rsyslog_port (default 514)

Port used by the remote syslog server.

default_facility (default 1)

Facility for the syslog message if not provided by the entry being relayed. (see RFC5424 for list of facilities.)

default_severity (default 6)

Severity for the syslog message if not provided by the entry being relayed. (see RFC5424 for list of priorities.)

format (default rfc5424)

Log format to use. Can be rfc3164, rfc5424 or custom.

logline (default null)

Custom logline format (ignored unless format is set to custom). The format is a limited version of the formatting used by rsyslog. Supported tags are pri, procotol-version, timestamp, timestamp:::date-rfc3339, HOSTNAME, app-name, procid, msgid, msg and structured-data.

For example the rfc3164 log format would be defined as <%pri%>%timestamp% %HOSTNAME% %app-name%[%procid%]: %msg%

structured_data (default null)

Content of structured data section (optional, required by some services to identify the sender).

ssl (default false)

Require encrypted connection.

ca_certs (default null)

CA path. Note! setting ca will automatically also set ssl to True

client_cert (default null)

Client certificate path, required if remote syslog requires SSL authentication.

client_key (default null)

Client key path, required if remote syslog requires SSL authentication.

format (default rfc5424)

Format message according to rfc5424 or rfc3164

Websocket Sender Configuration

websocket_uri (default null)

Which Websocket URI do you want the journalpump to write to. Required when using output_type websocket.

ca (default null)

Websocket Certificate Authority path, needed when you're using SSL authentication.

certfile (default null)

Websocket client certificate path, needed when you're using SSL authentication.

keyfile (default null)

Websocket client key path, needed when you're using SSL authentication.

socks5_proxy (default null)

Defined socks5 proxy to use for Websocket connections.

max_batch_size (default 1048576)

Adjust message batch size, set to 0 to disable batching. When batching is enabled, multiple journal messages are sent in a single websocket message, separated by a single NUL byte.

compression (default "snappy")

Compress messages on application level using the specified algorithm. Decompression is done by an application behind the websocket server, allowing end-to-end compression. When batching is enabled, compression is done on complete batches. Supported values: "snappy", "none".

websocket_compression (default "none")

Enable compression of websocket messages using the permessage-deflate extension. The messages will be decompressed by the websocket server. When batching is enabled, compression is done on complete batches. Supported values: "deflate", "none".

License

journalpump is licensed under the Apache License, Version 2.0. Full license text is available in the LICENSE file and at http://www.apache.org/licenses/LICENSE-2.0.txt

Credits

journalpump was created by Hannu Valtonen <[email protected]> and is now maintained by Aiven hackers <[email protected]>.

Recent contributors are listed on the project's GitHub contributors page.

Trademark

Apache Kafka is either registered trademark or trademark of the Apache Software Foundation in the United States and/or other countries. Elasticsearch, AWS CloudWatch, logplex and rsyslog are trademarks and property of their respective owners. All product and service names used in this website are for identification purposes only and do not imply endorsement.

Contact

Bug reports and patches are very welcome, please post them as GitHub issues and pull requests at https://github.com/aiven/journalpump . Any possible vulnerabilities or other serious issues should be reported directly to the maintainers <[email protected]>.