omelasticsearch: Elasticsearch Output Module

Module Name: omelasticsearch
Author: Rainer Gerhards <rgerhards@adiscon.com>

Purpose

This module provides native support for logging to Elasticsearch.

Notable Features

Configuration Parameters

Note

Parameter names are case-insensitive.

Action Parameters

Server

type default mandatory obsolete legacy directive
array none no none

An array of Elasticsearch servers in the specified format. If no scheme is specified, it will be chosen according to usehttps. If no port is specified, serverport will be used. Defaults to “localhost”.

Requests to Elasticsearch will be load-balanced between all servers in round-robin fashion.

Examples:
     server="localhost:9200"
     server=["elasticsearch1", "elasticsearch2"]

Serverport

type default mandatory obsolete legacy directive
integer 9200 no none

Default HTTP port to use to connect to Elasticsearch if none is specified on a server. Defaults to 9200

HealthCheckTimeout

type default mandatory obsolete legacy directive
integer 3500 no none

Specifies the number of milliseconds to wait for a successful health check on a server. Before trying to submit events to Elasticsearch, rsyslog will execute an HTTP HEAD to /_cat/health and expect an HTTP OK within this timeframe. Defaults to 3500.

Note, the health check is verifying connectivity only, not the state of the Elasticsearch cluster.

searchIndex

type default mandatory obsolete legacy directive
word none no none

Elasticsearch index to send your logs to. Defaults to “system”

dynSearchIndex

type default mandatory obsolete legacy directive
binary off no none

Whether the string provided for searchIndex should be taken as a rsyslog template. Defaults to “off”, which means the index name will be taken literally. Otherwise, it will look for a template with that name, and the resulting string will be the index name. For example, let’s assume you define a template named “date-days” containing “%timereported:1:10:date-rfc3339%”. Then, with dynSearchIndex=”on”, if you say searchIndex=”date-days”, each log will be sent to and index named after the first 10 characters of the timestamp, like “2013-03-22”.

searchType

type default mandatory obsolete legacy directive
word none no none

Elasticsearch type to send your index to. Defaults to “events”

dynSearchType

type default mandatory obsolete legacy directive
binary off no none

Like dynSearchIndex, it allows you to specify a rsyslog template for searchType, instead of a static string.

pipelineName

type default mandatory obsolete legacy directive
word none no none

The ingest node pipeline name to be inclued in the request. This allows pre processing of events bevor indexing them. By default, events are not send to a pipeline.

dynPipelineName

type default mandatory obsolete legacy directive
binary off no none

Like dynSearchIndex, it allows you to specify a rsyslog template for pipelineName, instead of a static string.

asyncrepl

type default mandatory obsolete legacy directive
binary off no none

No longer supported as ElasticSearch no longer supports it.

usehttps

type default mandatory obsolete legacy directive
binary off no none

Default scheme to use when sending events to Elasticsearch if none is specified on a server. Good for when you have Elasticsearch behind Apache or something else that can add HTTPS. Note that if you have a self-signed certificate, you’d need to install it first. This is done by copying the certificate to a trusted path and then running update-ca-certificates. That trusted path is typically /usr/local/share/ca-certificates but check the man page of update-ca-certificates for the default path of your distro

timeout

type default mandatory obsolete legacy directive
word 1m no none

How long Elasticsearch will wait for a primary shard to be available for indexing your log before sending back an error. Defaults to “1m”.

template

type default mandatory obsolete legacy directive
word see below no none

This is the JSON document that will be indexed in Elasticsearch. The resulting string needs to be a valid JSON, otherwise Elasticsearch will return an error. Defaults to:

$template JSONDefault, "{\"message\":\"%msg:::json%\",\"fromhost\":\"%HOSTNAME:::json%\",\"facility\":\"%syslogfacility-text%\",\"priority\":\"%syslogpriority-text%\",\"timereported\":\"%timereported:::date-rfc3339%\",\"timegenerated\":\"%timegenerated:::date-rfc3339%\"}"

Which will produce this sort of documents (pretty-printed here for readability):

{
    "message": " this is a test message",
    "fromhost": "test-host",
    "facility": "user",
    "priority": "info",
    "timereported": "2013-03-12T18:05:01.344864+02:00",
    "timegenerated": "2013-03-12T18:05:01.344864+02:00"
}

bulkmode

type default mandatory obsolete legacy directive
binary off no none

The default “off” setting means logs are shipped one by one. Each in its own HTTP request, using the Index API. Set it to “on” and it will use Elasticsearch’s Bulk API to send multiple logs in the same request. The maximum number of logs sent in a single bulk request depends on your maxbytes and queue settings - usually limited by the dequeue batch size. More information about queues can be found here.

maxbytes

type default mandatory obsolete legacy directive
word 100m no none

New in version 8.23.0.

When shipping logs with bulkmode on, maxbytes specifies the maximum size of the request body sent to Elasticsearch. Logs are batched until either the buffer reaches maxbytes or the the dequeue batch size is reached. In order to ensure Elasticsearch does not reject requests due to content length, verify this value is set accoring to the http.max_content_length setting in Elasticsearch. Defaults to 100m.

parent

type default mandatory obsolete legacy directive
word none no none

Specifying a string here will index your logs with that string the parent ID of those logs. Please note that you need to define the parent field in your mapping for that to work. By default, logs are indexed without a parent.

dynParent

type default mandatory obsolete legacy directive
binary off no none

Using the same parent for all the logs sent in the same action is quite unlikely. So you’d probably want to turn this “on” and specify a rsyslog template that will provide meaningful parent IDs for your logs.

uid

type default mandatory obsolete legacy directive
word none no none

If you have basic HTTP authentication deployed (eg through the elasticsearch-basic plugin), you can specify your user-name here.

pwd

type default mandatory obsolete legacy directive
word none no none

Password for basic authentication.

errorFile

type default mandatory obsolete legacy directive
word none no none

If specified, records failed in bulk mode are written to this file, including their error cause. Rsyslog itself does not process the file any more, but the idea behind that mechanism is that the user can create a script to periodically inspect the error file and react appropriately. As the complete request is included, it is possible to simply resubmit messages from that script.

Please note: when rsyslog has problems connecting to elasticsearch, a general error is assumed and the submit is retried. However, if we receive negative responses during batch processing, we assume an error in the data itself (like a mandatory field is not filled in, a format error or something along those lines). Such errors cannot be solved by simpy resubmitting the record. As such, they are written to the error file so that the user (script) can examine them and act appropriately. Note that e.g. after search index reconfiguration (e.g. dropping the mandatory attribute) a resubmit may be succesful.

tls.cacert

type default mandatory obsolete legacy directive
word none no none

This is the full path and file name of the file containing the CA cert for the CA that issued the Elasticsearch server cert. This file is in PEM format. For example: /etc/rsyslog.d/es-ca.crt

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 Elasticsearch. This file is in PEM format. For example: /etc/rsyslog.d/es-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 Elasticsearch. This file is in PEM format, and must be unencrypted, so take care to secure it properly. For example: /etc/rsyslog.d/es-client-key.pem

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.

bulkid

type default mandatory obsolete legacy directive
word none no none

This is the unique id to assign to the record. The bulk part is misleading - this can be used in both bulk mode bulkmode or in index (record at a time) mode. Although you can specify a static value for this parameter, you will almost always want to specify a template for the value of this parameter, and set dynbulkid=”on” dynbulkid. NOTE: you must use bulkid and dynbulkid in order to use writeoperation=”create” writeoperation.

dynbulkid

type default mandatory obsolete legacy directive
binary off no none

If this parameter is set to “on”, then the bulkid parameter bulkid specifies a template to use to generate the unique id value to assign to the record. If using bulkid you will almost always want to set this parameter to “on” to assign a different unique id value to each record. NOTE: you must use bulkid and dynbulkid in order to use writeoperation=”create” writeoperation.

writeoperation

type default mandatory obsolete legacy directive
word index no none

The value of this parameter is either “index” (the default) or “create”. If “create” is used, this means the bulk action/operation will be create - create a document only if the document does not already exist. The record must have a unique id in order to use create. See bulkid and dynbulkid. See Example 3 for an example.

retryfailures

type default mandatory obsolete legacy directive
binary off no none

If this parameter is set to “on”, then the module will look for an “errors”:true in the bulk index response. If found, each element in the response will be parsed to look for errors, since a bulk request may have some records which are successful and some which are failures. Failed requests will be converted back into records and resubmitted back to rsyslog for reprocessing. Each failed request will be resubmitted with a local variable called $.omes. This is a hash consisting of the fields from the metadata header in the original request, and the fields from the response. If the same field is in the request and response, the value from the field in the request will be used, to facilitate retries that want to send the exact same request, and want to know exactly what was sent. See below Example 4 for an example of how retry processing works. NOTE The retried record will be resubmitted at the “top” of your processing pipeline. If your processing pipeline is not idempotent (that is, your processing pipeline expects “raw” records), then you can specify a ruleset to redirect retries to. See retryruleset below.

$.omes fields:

  • writeoperation - the operation used to submit the request - for rsyslog omelasticsearch this currently means either “index” or “create”
  • status - the HTTP status code - typically an error will have a 4xx or 5xx code - of particular note is 429 - this means Elasticsearch was unable to process this bulk record request due to a temporary condition e.g. the bulk index thread pool queue is full, and rsyslog should retry the operation.
  • _index, _type, _id, pipeline, _parent - the metadata associated with the request - not all of these fields will be present with every request - for example, if you do not use “pipelinename” or “dynpipelinename”, there will be no $.omes!pipeline field.
  • error - a hash containing one or more, possibly nested, fields containing more detailed information about a failure. Typically there will be fields $.omes!error!type (a keyword) and $.omes!error!reason (a longer string) with more detailed information about the rejection. NOTE: The format is apparently not described in great detail, so code must not make any assumption about the availability of error or any specific sub-field.

There may be other fields too - the code just copies everything in the response. Here is an example of a detailed error response, in JSON format, from Elasticsearch 5.6.9:

{"omes":
  {"writeoperation": "create",
   "_index": "rsyslog_testbench",
   "_type": "test-type",
   "_id": "92BE7AF79CD44305914C7658AF846A08",
   "status": 400,
   "error":
     {"type": "mapper_parsing_exception",
      "reason": "failed to parse [msgnum]",
      "caused_by":
        {"type": "number_format_exception",
         "reason": "For input string: \"x00000025\""}}}}

Reference: https://www.elastic.co/guide/en/elasticsearch/guide/current/bulk.html#bulk

retryruleset

type default mandatory obsolete legacy directive
word   no none

If retryfailures is not “on” (retryfailures) then this parameter has no effect. This parameter specifies the name of a ruleset to use to route retries. This is useful if you do not want retried messages to be processed starting from the top of your processing pipeline, or if you have multiple outputs but do not want to send retried Elasticsearch failures to all of your outputs, and you do not want to clutter your processing pipeline with a lot of conditionals. See below Example 4 for an example of how retry processing works.

ratelimit.interval

type default mandatory obsolete legacy directive
integer 600 no none

If retryfailures is not “on” (retryfailures) then this parameter has no effect. 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

type default mandatory obsolete legacy directive
integer 20000 no none

If retryfailures is not “on” (retryfailures) then this parameter has no effect. Specifies the maximum number of messages that can be emitted within the ratelimit.interval interval. For futher information, see description there.

rebindinterval

type default mandatory obsolete legacy directive
integer -1 no none

This parameter tells omelasticsearch to close the connection and reconnect to Elasticsearch after this many operations have been submitted. The default value -1 means that omelasticsearch will not reconnect. A value greater than -1 tells omelasticsearch, after this many operations have been submitted to Elasticsearch, to drop the connection and establish a new connection. This is useful when rsyslog connects to multiple Elasticsearch nodes through a router or load balancer, and you need to periodically drop and reestablish connections to help the router balance the connections. Use the counter rebinds to monitor the number of times this has happened.

Statistic Counter

This plugin maintains global statistics, which accumulate all action instances. The statistic is named “omelasticsearch”. Parameters are:

  • submitted - number of messages submitted for processing (with both success and error result)
  • fail.httprequests - the number of times a http request failed. Note that a single http request may be used to submit multiple messages, so this number may be (much) lower than fail.http.
  • fail.http - number of message failures due to connection like-problems (things like remote server down, broken link etc)
  • fail.es - number of failures due to elasticsearch error reply; Note that this counter does NOT count the number of failed messages but the number of times a failure occured (a potentially much smaller number). Counting messages would be quite performance-intense and is thus not done.

The following counters are available when retryfailures=”on” is used:

  • response.success - number of records successfully sent in bulk index requests - counts the number of successful responses
  • response.bad - number of times omelasticsearch received a response in a bulk index response that was unrecognized or unable to be parsed. This may indicate that omelasticsearch is attempting to communicate with a version of Elasticsearch that is incompatible, or is otherwise sending back data in the response that cannot be handled
  • response.duplicate - number of records in the bulk index request that were duplicates of already existing records - this will only be reported if using writeoperation=”create” and bulkid to assign each record a unique ID
  • response.badargument - number of times omelasticsearch received a response that had a status indicating omelasticsearch sent bad data to Elasticsearch. For example, status 400 and an error message indicating omelasticsearch attempted to store a non-numeric string value in a numeric field.
  • response.bulkrejection - number of times omelasticsearch received a response that had a status indicating Elasticsearch was unable to process the record at this time - status 429. The record can be retried.
  • response.other - number of times omelasticsearch received a response not recognized as one of the above responses, typically some other 4xx or 5xx http status.
  • rebinds - if using rebindinterval this will be the number of times omelasticsearch has reconnected to Elasticsearch

The fail.httprequests and fail.http counters reflect only failures that omelasticsearch detected. Once it detects problems, it (usually, depends on circumstances) tell the rsyslog core that it wants to be suspended until the situation clears (this is a requirement for rsyslog output modules). Once it is suspended, it does NOT receive any further messages. Depending on the user configuration, messages will be lost during this period. Those lost messages will NOT be counted by impstats (as it does not see them).

Note that some previous (pre 7.4.5) versions of this plugin had different counters. These were experimental and confusing. The only ones really used were “submits”, which were the number of successfully processed messages and “connfail” which were equivalent to “failed.http”.

How Retries Are Handled

When using retryfailures=”on” (retryfailures), the original Message object (that is, the original smsg_t *msg object) is not available. This means none of the metadata associated with that object, such as various timestamps, hosts/ip addresses, etc. are not available for the retry operation. The only thing available are the metadata header (_index, _type, _id, pipeline, _parent) and original JSON string sent in the original request, and whatever data is returned in the error response. All of these are made available in the $.omes fields. If the same field name exists in the request metadata and the response, the field from the request will be used, in order to facilitate retrying the exact same request. For the message to retry, the code will take the original JSON string and parse it back into an internal Message object. This means you may need to use a different template to output messages for your retry ruleset. For example, if you used the following template to format the Elasticsearch message for the initial submission:

template(name="es_output_template"
         type="list"
         option.json="on") {
           constant(value="{")
             constant(value="\"timestamp\":\"")      property(name="timereported" dateFormat="rfc3339")
             constant(value="\",\"message\":\"")     property(name="msg")
             constant(value="\",\"host\":\"")        property(name="hostname")
             constant(value="\",\"severity\":\"")    property(name="syslogseverity-text")
             constant(value="\",\"facility\":\"")    property(name="syslogfacility-text")
             constant(value="\",\"syslogtag\":\"")   property(name="syslogtag")
           constant(value="\"}")
         }

You would have to use a different template for the retry, since none of the timereported, msg, etc. fields will have the same values for the retry as for the initial try.

Same with the other omelasticsearch parameters which can be constructed with templates, such as “dynpipelinename”, “dynsearchindex”, “dynsearchtype”, “dynparent”, and “dynbulkid”. For example, if you generate the _id to use in the request, you will want to reuse the same _id for each subsequent retry:

template(name="id-template" type="string" string="%$.es_msg_id%")
if strlen($.omes!_id) > 0 then {
    set $.es_msg_id = $.omes!_id;
} else {
    # NOTE: depends on rsyslog being compiled with --enable-uuid
    set $.es_msg_id = $uuid;
}
action(type="omelasticsearch" bulkid="id-template" ...)

That is, if this is a retry, $.omes!_id will be set, so use that value for the bulk id for this record, otherwise, generate a new one with $uuid. Note that the template uses the temporary variable $.es_msg_id which must be set each time, to either $.omes!_id or $uuid.

The rawmsg field is a special case. If the original request had a field called message, then when constructing the new message from the original to retry, the rawmsg message property will be set to the value of the message field. Otherwise, the rawmsg property value will be set to the entire original request - the data part, not the metadata. In previous versions, without the message field, the rawmsg property was set to the value of the data plus the Elasticsearch metadata, which caused problems with retries. See rsyslog issue 3573

Examples

Example 1

The following sample does the following:

  • loads the omelasticsearch module
  • outputs all logs to Elasticsearch using the default settings
module(load="omelasticsearch")
*.*     action(type="omelasticsearch")

Example 2

The following sample does the following:

  • loads the omelasticsearch module
  • defines a template that will make the JSON contain the following properties
    • RFC-3339 timestamp when the event was generated
    • the message part of the event
    • hostname of the system that generated the message
    • severity of the event, as a string
    • facility, as a string
    • the tag of the event
  • outputs to Elasticsearch with the following settings
    • host name of the server is myserver.local
    • port is 9200
    • JSON docs will look as defined in the template above
    • index will be “test-index”
    • type will be “test-type”
    • activate bulk mode. For that to work effectively, we use an in-memory queue that can hold up to 5000 events. The maximum bulk size will be 300
    • retry indefinitely if the HTTP request failed (eg: if the target server is down)
module(load="omelasticsearch")
template(name="testTemplate"
         type="list"
         option.json="on") {
           constant(value="{")
             constant(value="\"timestamp\":\"")      property(name="timereported" dateFormat="rfc3339")
             constant(value="\",\"message\":\"")     property(name="msg")
             constant(value="\",\"host\":\"")        property(name="hostname")
             constant(value="\",\"severity\":\"")    property(name="syslogseverity-text")
             constant(value="\",\"facility\":\"")    property(name="syslogfacility-text")
             constant(value="\",\"syslogtag\":\"")   property(name="syslogtag")
           constant(value="\"}")
         }
action(type="omelasticsearch"
       server="myserver.local"
       serverport="9200"
       template="testTemplate"
       searchIndex="test-index"
       searchType="test-type"
       bulkmode="on"
       maxbytes="100m"
       queue.type="linkedlist"
       queue.size="5000"
       queue.dequeuebatchsize="300"
       action.resumeretrycount="-1")

Example 3

The following sample shows how to use writeoperation with dynbulkid and bulkid. For simplicity, it assumes rsyslog has been built with –enable-libuuid which provides the uuid property for each record:

module(load="omelasticsearch")
set $!es_record_id = $uuid;
template(name="bulkid-template" type="list") { property(name="$!es_record_id") }
action(type="omelasticsearch"
       ...
       bulkmode="on"
       bulkid="bulkid-template"
       dynbulkid="on"
       writeoperation="create")

Example 4

The following sample shows how to use retryfailures to process, discard, or retry failed operations. This uses writeoperation=”create” with a unique bulkid so that we can check for and discard duplicate messages as successful. The try_es ruleset is used both for the initial attempt and any subsequent retries. The code in the ruleset assumes that if $.omes!status is set and is non-zero, this is a retry for a previously failed operation. If the status was successful, or Elasticsearch said this was a duplicate, the record is already in Elasticsearch, so we can drop the record. If there was some error processing the response e.g. Elasticsearch sent a response formatted in some way that we did not know how to process, then submit the record to the error_es ruleset. If the response was a “hard” error like 400, then submit the record to the error_es ruleset. In any other case, such as a status 429 or 5xx, the record will be resubmitted to Elasticsearch. In the example, the error_es ruleset just dumps the records to a file.

module(load="omelasticsearch")
module(load="omfile")
template(name="bulkid-template" type="list") { property(name="$.es_record_id") }

ruleset(name="error_es") {
        action(type="omfile" template="RSYSLOG_DebugFormat" file="es-bulk-errors.log")
}

ruleset(name="try_es") {
    if strlen($.omes!status) > 0 then {
        # retry case
        if ($.omes!status == 200) or ($.omes!status == 201) or (($.omes!status == 409) and ($.omes!writeoperation == "create")) then {
            stop # successful
        }
        if ($.omes!writeoperation == "unknown") or (strlen($.omes!error!type) == 0) or (strlen($.omes!error!reason) == 0) then {
            call error_es
            stop
        }
        if ($.omes!status == 400) or ($.omes!status < 200) then {
            call error_es
            stop
        }
        # else fall through to retry operation
    }
    if strlen($.omes!_id) > 0 then {
        set $.es_record_id = $.omes!_id;
    } else {
        # NOTE: depends on rsyslog being compiled with --enable-uuid
        set $.es_record_id = $uuid;
    }
    action(type="omelasticsearch"
              ...
              bulkmode="on"
              bulkid="bulkid-template"
              dynbulkid="on"
              writeoperation="create"
              retryfailures="on"
              retryruleset="try_es")
}
call try_es

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: