TwitterFacebookGoogle

omhttp: HTTP Output Module

Module Name:omhttp
Module Type:contributed - not maintained by rsyslog core team
Current Mainainer:Gabriel Intrator
Original Author:Christian Tramnitz

Warning

This page is incomplete, if you want to contribute you can do this on github in the rsyslog-doc repository.

Purpose

This module provides the ability to send messages over an HTTP REST interface.

This module supports sending messages in individual requests (the default), and batching multiple messages into a single request. Support for retrying failed requests is available in both modes. GZIP compression is configurable with the compress parameter. TLS encryption is configurable with the useHttps parameter and associated tls parameters.

In the default mode, every message is sent in its own HTTP request and it is a drop-in replacement for any other output module. In batch mode, the module implements several batch formatting options that are configurable via the batch.format parameter. Some additional attention to message formatting and retry strategy is required in this mode.

See the Examples section for some configuration examples.

Notable Features

Configuration Parameters

Note

Parameter names are case-insensitive.

Action Parameters

Server

typedefaultmandatoryobsolete legacy directive
arraylocalhostnonone

The server address you want to connect to.

Serverport

typedefaultmandatoryobsolete legacy directive
integer443nonone

The port you want to connect to.

healthchecktimeout

typedefaultmandatoryobsolete legacy directive
integer3500nonone

The time after which the health check will time out in milliseconds.

httpcontenttype

typedefaultmandatoryobsolete legacy directive
wordapplication/json; charset=utf-8nonone

The HTTP “Content-Type” header sent with each request. This parameter will override other defaults. If a batching mode is specified, the correct content type is automatically configured.

httpheaderkey

typedefaultmandatoryobsolete legacy directive
wordnonenonone

The header key. Currently only a single additional header/key pair is configurable, further development is needed to support arbitrary header key/value lists.

httpheadervalue

typedefaultmandatoryobsolete legacy directive
wordnonenonone

The header value for httpheaderkey.

uid

typedefaultmandatoryobsolete legacy directive
wordnonenonone

The username for basic auth.

pwd

typedefaultmandatoryobsolete legacy directive
wordnonenonone

The password for the user for basic auth.

restpath

typedefaultmandatoryobsolete legacy directive
wordnonenonone

The rest path you want to use. Do not include the leading slash character. If the full path looks like “localhost:5000/my/path”, restpath should be “my/path”.

dynrestpath

typedefaultmandatoryobsolete legacy directive
binaryoffnonone

When this parameter is set to “on” you can specify a template name in the parameter restpath instead of the actual path. This way you will be able to use dynamic rest paths for your messages based on the template you are using.

checkpath

typedefaultmandatoryobsolete legacy directive
wordnonenonone

The health check path you want to use. Do not include the leading slash character. If the full path looks like “localhost:5000/my/path”, checkpath should be “my/path”. When this parameter is set, omhttp utilizes this path to determine if it is safe to resume (from suspend mode) and communicates this status back to rsyslog core. This parameter defaults to none, which implies that health checks are not needed, and it is always safe to resume from suspend mode.

Important - Note that it is highly recommended to set a valid health check path, as this allows omhttp to better determine whether it is safe to retry. See the rsyslog action queue documentation for more info regarding general rsyslog suspend and resume behavior.

batch

typedefaultmandatoryobsolete legacy directive
binaryoffnonone

Batch and bulkmode do the same thing, bulkmode included for backwards compatibility. See the Message Batching section for a detailed breakdown of how batching is implemented.

This parameter activates batching mode, which queues messages and sends them as a single request. There are several related parameters that specify the format and size of the batch: they are batch.format, batch.maxbytes, and batch.maxsize.

Note that rsyslog core is the ultimate authority on when a batch must be submitted, due to the way that batching is implemented. This plugin implements the output plugin transaction interface. There may be multiple batches in a single transaction, but a batch will never span multiple transactions. This means that if batch.maxsize or batch.maxbytes is set very large, you may never actually see batches hit this size. Additionally, the number of messages per transaction is determined by the size of the main, action, and ruleset queues as well.

Additionally, due to some open issues with rsyslog and the transaction interface, batching requires some nuanced retry configuration.

batch.format

typedefaultmandatoryobsolete legacy directive
wordnewlinenonone

This parameter specifies how to combine multiple messages into a single batch. Valid options are newline (default), jsonarray, and kafkarest.

Each message on the “Inputs” line is the templated log line that is fed into the omhttp action, and the “Output” line describes the resulting payload sent to the configured HTTP server.

  1. newline - Concatenates each message into a single string joined by newline (“\n”) characters. This mode is default and places no restrictions on the structure of the input messages.
Inputs: "message 1" "message 2" "message 3"
Output: "message 1\nmessage2\nmessage3"
  1. jsonarray - Builds a JSON array containing all messages in the batch. This mode requires that each message is parseable JSON, since the plugin parses each message as JSON while building the array.
Inputs: {"msg": "message 1"} {"msg"": "message 2"} {"msg": "message 3"}
Output: [{"msg": "message 1"}, {"msg"": "message 2"}, {"msg": "message 3"}]
  1. kafkarest - Builds a JSON object that conforms to the Kafka Rest Proxy specification. This mode requires that each message is parseable JSON, since the plugin parses each message as JSON while building the batch object.
Inputs: {"msg": "message 1"} {"msg"": "message 2"} {"msg": "message 3"}
Output: {"records": [{"value": {"msg": "message 1"}}, {"value": {"msg": "message 2"}}, {"value": {"msg": "message 3"}}]}

batch.maxsize

typedefaultmandatoryobsolete legacy directive
Size100nonone

This parameter specifies the maximum number of messages that will be sent in each batch.

batch.maxbytes

typedefaultmandatoryobsolete legacy directive
Size10485760 (10MB)nonone

batch.maxbytes and maxbytes do the same thing, maxbytes included for backwards compatibility.

This parameter specifies the maximum size in bytes for each batch.

template

typedefaultmandatoryobsolete legacy directive
wordStdJSONFmtnonone

The template to be used for the messages.

Note that in batching mode, this describes the format of each individual message, not the format of the resulting batch. Some batch modes require that a template produces valid JSON.

retry

typedefaultmandatoryobsolete legacy directive
binaryoffnonone

This parameter specifies whether failed requests should be retried using the custom retry logic implemented in this plugin. Requests returning 5XX HTTP status codes are considered retriable. If retry is enabled, set retry.ruleset as well.

Note that retries are generally handled in rsyslog by setting action.resumeRetryCount=”-1” (or some other integer), and the plugin lets rsyslog know it should start retrying by suspending itself. This is still the recommended approach in the 2 cases enumerated below when using this plugin. In both of these cases, the output plugin transaction interface is not used. That is, from rsyslog core’s point of view, each message is contained in its own transaction.

  1. Batching is off (batch=”off”)
  2. Batching is on and the maximum batch size is 1 (batch=”on” batch.maxsize=”1”)

This custom retry behavior is the result of a bug in rsyslog’s handling of transaction commits. See this issue for full details. Essentially, if rsyslog hands omhttp 4 messages, and omhttp batches them up but the request fails, rsyslog will only retry the LAST message that it handed the plugin, instead of all 4, even if the plugin returns the correct “defer commit” statuses for messages 1, 2, and 3. This means that omhttp cannot rely on action.resumeRetryCount for any transaction that processes more than a single message, and explains why the 2 above cases do work correctly.

It looks promising that issue will be resolved at some point, so this behavior can be revisited at that time.

retry.ruleset

typedefaultmandatoryobsolete legacy directive
wordnonenonone

This parameter specifies the ruleset where this plugin should requeue failed messages if retry is on. This ruleset generally would contain another omhttp action instance.

Important - Note that the message that is queued on the retry ruleset is the templated output of the initial omhttp action. This means that no further templating should be done to messages inside this ruleset, unless retries should be templated differently than first-tries. An “echo template” does the trick here.

template(name="tpl_echo" type="string" string="%msg%")

This retry ruleset can recursively call itself as its own retry.ruleset to retry forever, but there is no timeout behavior currently implemented.

Alternatively, the omhttp action in the retry ruleset could be configured to support action.resumeRetryCount as explained above in the retry parameter section. The benefit of this approach is that retried messages still hit the server in a batch format (though with a single message in it), and the ability to configure rsyslog to give up after some number of resume attempts so as to avoid resource exhaustion.

Or, if some data loss or high latency is acceptable, do not configure retries with the retry ruleset itself. A single retry from the original ruleset might catch most failures, and errors from the retry ruleset could still be logged using the errorfile parameter and sent later on via some other process.

ratelimit.interval

typedefaultmandatoryobsolete legacy directive
integer600nonone

This parameter sets the rate limiting behavior for the retry.ruleset. Specifies the interval in seconds onto which rate-limiting is to be applied. If more than ratelimit.burst messages are read during that interval, further messages up to the end of the interval are discarded. The number of messages discarded is emitted at the end of the interval (if there were any discards). Setting this to value zero turns off ratelimiting.

ratelimit.burst

typedefaultmandatoryobsolete legacy directive
integer20000nonone

This parameter sets the rate limiting behavior for the retry.ruleset. Specifies the maximum number of messages that can be emitted within the ratelimit.interval interval. For further information, see description there.

errorfile

typedefaultmandatoryobsolete legacy directive
wordnonenonone

Here you can set the name of a file where all errors will be written to. Any request that returns a 4XX or 5XX HTTP code is recorded in the error file. Each line is JSON formatted with “request” and “response” fields, example pretty-printed below.

{
    "request": {
        "url": "https://url.com:443/path",
        "postdata": "mypayload"
    },
    "response" : {
        "status": 400,
        "message": "error string"
    }
}

It is intended that a full replay of failed data is possible by processing this file.

compress

typedefaultmandatoryobsolete legacy directive
binaryoffnonone

When switched to “on” each message will be compressed as GZIP using zlib’s deflate compression algorithm.

A “Content-Encoding: gzip” HTTP header is added to each request when this feature is used. Set the compress.level for fine-grained control.

compress.level

typedefaultmandatoryobsolete legacy directive
integer-1nonone

Specify the zlib compression level if compress is enabled. Check the zlib manual for further documentation.

“-1” is the default value that strikes a balance between best speed and best compression. “0” disables compression. “1” results in the fastest compression. “9” results in the best compression.

useHttps

typedefaultmandatoryobsolete legacy directive
binaryoffnonone

When switched to “on” you will use https instead of http.

tls.cacert

typedefaultmandatoryobsolete legacy directive
wordnonenonone

This parameter sets the path to the Certificate Authority (CA) bundle. Expects .pem format.

tls.mycert

typedefaultmandatoryobsolete legacy directive
wordnonenonone

This parameter sets the path to the SSL client certificate. Expects .pem format.

tls.myprivkey

typedefaultmandatoryobsolete legacy directive
wordnonenonone

The parameters sets the path to the SSL private key. Expects .pem format.

allowunsignedcerts

typedefaultmandatoryobsolete legacy directive
booleanoffnonone

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

typedefaultmandatoryobsolete legacy directive
booleanoffnonone

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.

reloadonhup

typedefaultmandatoryobsolete legacy directive
binaryoffnonone

If this parameter is “on”, the plugin will close and reopen any libcurl handles on a HUP signal. This option is primarily intended to enable reloading short-lived certificates without restarting rsyslog.

Statistic Counter

This plugin maintains global statistics for omhttp that accumulates all action instances. The statistic origin is named “omhttp” with following counters:

  • messages.submitted - Number of messages submitted to omhttp. Messages resubmitted via a retry ruleset will be counted twice.
  • messages.success - Number of messages successfully sent.
  • messages.fail - Number of messages that omhttp failed to deliver for any reason.
  • messages.retry - Number of messages that omhttp resubmitted for retry via the retry ruleset.
  • request.count - Number of attempted HTTP requests.
  • request.success - Number of successful HTTP requests. A successful request can return any HTTP status code.
  • request.fail - Number of failed HTTP requests. A failed request is something like an invalid SSL handshake, or the server is not reachable. Requests returning 4XX or 5XX HTTP status codes are not failures.
  • request.status.success - Number of requests returning 1XX or 2XX HTTP status codes.
  • request.status.fail - Number of requests returning 3XX, 4XX, or 5XX HTTP status codes. If a requests fails (i.e. server not reachable) this counter will not be incremented.

Message Batching

See the batch.format section for some light examples of available batching formats.

Implementation

Here’s the pseudocode of the batching algorithm used by omhttp. This section of code would run once per transaction.

Q = Queue()

def submit(Q):                      # function to submit
    batch = serialize(Q)            # serialize according to configured batch.format
    result = post(batch)            # http post serialized batch to server
    checkFailureAndRetry(Q, result) # check if post failed and pushed failed messages to configured retry.ruleset
    Q.empty()                       # reset for next batch


while isActive(transaction):            # rsyslog manages the transaction
    message = receiveMessage()          # rsyslog sends us messages
    if wouldTriggerSubmit(Q, message):  # if this message puts us over maxbytes or maxsize
        submit(Q)                       # submit the current batch
    Q.push(message)                     # queue this message on the current batch

submit(Q)   # transaction is over, submit what is currently in the queue

Walkthrough

This is a run through of a file tailing to omhttp scenario. Suppose we have a file called /var/log/my.log with this content..

001 message
002 message
003 message
004 message
005 message
006 message
007 message
...

We are tailing this using imfile and defining a template to generate a JSON payload…

input(type="imfile" File="/var/log/my.log" ruleset="rs_omhttp" ... )

# Produces JSON formatted payload
template(name="tpl_omhttp_json" type="list") {
    constant(value="{")   property(name="msg"           outname="message"   format="jsonfr")
    constant(value=",")   property(name="hostname"      outname="host"      format="jsonfr")
    constant(value=",")   property(name="timereported"  outname="timestamp" format="jsonfr" dateFormat="rfc3339")
    constant(value="}")
}

Our omhttp ruleset is configured to batch using the jsonarray format with 5 messages per batch, and to use a retry ruleset.

module(load="omhttp")

ruleset(name="rs_omhttp") {
    action(
        type="omhttp"
        template="tpl_omhttp_json"
        batch="on"
        batch.format="jsonarray"
        batch.maxsize="5"
        retry="on"
        retry.ruleset="rs_omhttp_retry"
        ...
    )
}

call rs_omhttp

Each input message to this omhttp action is the output of tpl_omhttp_json with the following structure..

{"message": "001 message", "host": "localhost", "timestamp": "2018-12-28T21:14:13.840470+00:00"}

After 5 messages have been queued, and a batch submit is triggered, omhttp serializes the messages as a JSON array and attempts to post the batch to the server. At this point the payload on the wire looks like this..

[
    {"message": "001 message", "host": "localhost", "timestamp": "2018-12-28T21:14:13.000000+00:00"},
    {"message": "002 message", "host": "localhost", "timestamp": "2018-12-28T21:14:14.000000+00:00"},
    {"message": "003 message", "host": "localhost", "timestamp": "2018-12-28T21:14:15.000000+00:00"},
    {"message": "004 message", "host": "localhost", "timestamp": "2018-12-28T21:14:16.000000+00:00"},
    {"message": "005 message", "host": "localhost", "timestamp": "2018-12-28T21:14:17.000000+00:00"}
]

If the request fails, omhttp requeues each failed message onto the retry ruleset. However, recall that the inputs to the rs_omhttp ruleset are the rendered outputs of tpl_json_omhttp, and therefore we cannot use the same template (and therefore the same action instance) to produce the retry messages. At this point, the msg rsyslog property is {"message": "001 message", "host": "localhost", "timestamp": "2018-12-28T21:14:13.000000+00:00"} instead of the original 001 message, and tpl_json_omhttp would render incorrect payloads.

Instead we define a simple template that echos its input..

template(name="tpl_echo" type="string" string="%msg%")

And assign it to the retry template..

ruleset(name="rs_omhttp_retry") {
    action(
        type="omhttp"
        template="tpl_echo"
        batch="on"
        batch.format="jsonarray"
        batch.maxsize="5"
        ...
    )
}

And the destination is none the wiser! The newline, jsonarray, and kafkarest formats all behave in the same way with respect to their batching and retry behavior, and differ only in the format of the on-the-wire payload. The formats themselves are described in the batch.format section.

Examples

Example 1

The following example is a basic usage, first the module is loaded and then the action is used with a standard retry strategy.

module(load="omhttp")
template(name="tpl1" type="string" string="{\"type\":\"syslog\", \"host\":\"%HOSTNAME%\"}")
action(
    type="omhttp"
    server="127.0.0.1"
    serverport="8080"
    restpath="events"
    template="tpl1"
    action.resumeRetryCount="3"
)

Example 2

The following example is a basic batch usage with no retry processing.

module(load="omhttp")
template(name="tpl1" type="string" string="{\"type\":\"syslog\", \"host\":\"%HOSTNAME%\"}")
action(
    type="omhttp"
    server="127.0.0.1"
    serverport="8080"
    restpath="events"
    template="tpl1"
    batch="on"
    batch.format="jsonarray"
    batch.maxsize="10"
)

Example 3

The following example is a batch usage with a retry ruleset that retries forever

module(load="omhttp")

template(name="tpl_echo" type="string" string="%msg%")
ruleset(name="rs_retry_forever") {
    action(
        type="omhttp"
        server="127.0.0.1"
        serverport="8080"
        restpath="events"
        template="tpl_echo"

        batch="on"
        batch.format="jsonarray"
        batch.maxsize="10"

        retry="on"
        retry.ruleset="rs_retry_forever"
    )
}

template(name="tpl1" type="string" string="{\"type\":\"syslog\", \"host\":\"%HOSTNAME%\"}")
action(
    type="omhttp"
    server="127.0.0.1"
    serverport="8080"
    restpath="events"
    template="tpl1"

    batch="on"
    batch.format="jsonarray"
    batch.maxsize="10"

    retry="on"
    retry.ruleset="rs_retry_forever"
)

Example 4

The following example is a batch usage with a couple retry options

module(load="omhttp")

template(name="tpl_echo" type="string" string="%msg%")

# This retry ruleset tries to send batches once then logs failures.
# Error log could be tailed by rsyslog itself or processed by some
# other program.
ruleset(name="rs_retry_once_errorfile") {
    action(
        type="omhttp"
        server="127.0.0.1"
        serverport="8080"
        restpath="events"
        template="tpl_echo"

        batch="on"
        batch.format="jsonarray"
        batch.maxsize="10"

        retry="off"
        errorfile="/var/log/rsyslog/omhttp_errors.log"
    )
}

# This retry ruleset gives up trying to batch messages and instead always
# uses a batch size of 1, relying on the suspend/resume mechanism to do
# further retries if needed.
ruleset(name="rs_retry_batchsize_1") {
    action(
        type="omhttp"
        server="127.0.0.1"
        serverport="8080"
        restpath="events"
        template="tpl_echo"

        batch="on"
        batch.format="jsonarray"
        batch.maxsize="1"
        action.resumeRetryCount="-1"
    )
}

template(name="tpl1" type="string" string="{\"type\":\"syslog\", \"host\":\"%HOSTNAME%\"}")
action(
    type="omhttp"
    template="tpl1"

    ...

    retry="on"
    retry.ruleset="<some_retry_ruleset>"
)

See also

Help with configuring/using Rsyslog:

  • Mailing list - best route for general questions
  • GitHub: rsyslog source project - detailed questions, reporting issues that are believed to be bugs with Rsyslog
  • Stack Exchange (View, Ask) - experimental support from rsyslog community

See also

Contributing to Rsyslog:

© 2008-2019, `Rainer Gerhards and Others. This site uses the “better” theme for Sphinx.