﻿---
title: SNMP Integration Plugin v4.3.0
description: Plugin version: v4.3.0, Released on: 2026-05-19, Changelog. For other versions, see the overview list. To learn more about Logstash, see the Logstash...
url: https://docs-v3-preview.elastic.dev/elastic/logstash-docs-md/tree/main/vpr/v4-3-0-plugins-integrations-snmp
products:
  - Logstash
applies_to:
  - Elastic Stack: Generally available
---

# SNMP Integration Plugin v4.3.0
- Plugin version: v4.3.0
- Released on: 2026-05-19
- [Changelog](https://github.com/logstash-plugins/logstash-integration-snmp/blob/v4.3.0/CHANGELOG.md)

For other versions, see the [overview list](https://docs-v3-preview.elastic.dev/elastic/logstash-docs-md/tree/main/vpr/integration-snmp-index).
To learn more about Logstash, see the [Logstash Reference](https://www.elastic.co/guide/en/logstash/current/index.html).

## Getting help

For questions about the plugin, open a topic in the [Discuss](http://discuss.elastic.co) forums. For bugs or feature requests, open an issue in [Github](https://github.com/logstash-plugins/logstash-integration-snmp). For the list of Elastic supported plugins, please consult the [Elastic Support Matrix](https://www.elastic.co/support/matrix#matrix_logstash_plugins).
**Announcing the new SNMP integration plugin**
The new `logstash-integration-snmp` plugin is available and bundled with Logstash 8.15.0 (and later) by default. This plugin combines our classic `logstash-input-snmp` and `logstash-input-snmptrap` plugins into a single Ruby gem at v4.0.0 and later. Earlier versions of the stand-alone plugins that were bundled with Logstash by default will be replaced by the 4.0.0+ version contained in this new integration.
Before you upgrade to Logstash 8.15.0 that includes this new integration by default, be aware of [behavioral and mapping differences](https://www.elastic.co/guide/en/logstash/current/plugins-integrations-snmp.html#plugins-integrations-snmp-migration) between stand-alone plugins and the new versions included in `integration-snmp`. If you need to maintain current mappings for the `input-snmptrap` plugin, you have options to [preserve existing behavior](https://www.elastic.co/guide/en/logstash/current/plugins-integrations-snmp.html#plugins-integrations-snmp-input-snmptrap-compat).

## Description

The SNMP integration plugin includes:
- [SNMP input plugin](https://www.elastic.co/guide/en/logstash/current/plugins-inputs-snmp.html)
- [Snmptrap input plugin](https://www.elastic.co/guide/en/logstash/current/plugins-inputs-snmptrap.html)

The new `logstash-integration-snmp` plugin combines the `logstash-input-snmp` and `logstash-input-snmptrap` plugins into one integrated plugin that encompasses the capabilities of both. This integrated plugin package provides better alignment in snmp processing, better resource management, easier package maintenance, and a smaller installation footprint.
In this section, we’ll cover:
- [Migrating to `logstash-integration-snmp` from individual plugins](#v4.3.0-plugins-integrations-snmp-migration)
- [Importing MIBs](#v4.3.0-plugins-integrations-snmp-import-mibs)


## Migrating to `logstash-integration-snmp` from individual plugins

You’ll retain and expand the functionality of existing stand-alone plugins, but in a more compact, integrated package. In this section, we’ll note mapping and behavioral changes, and explain how to preserve current behavior if needed.

### Migration notes: `logstash-input-snmp`

As a component of the new `logstash-integration-snmp` plugin, the `logstash-input-snmp` plugin offers the same capabilities as the stand-alone [logstash-input-snmp](https://github.com/logstash-plugins/logstash-input-snmp).
You might need to address some behavior changes depending on the use-case and how the ingested data is being handled through the pipeline.

### Changes to mapping and error logging: `logstash-input-snmp`

- **No such instance errors** are mapped as `error: no such instance currently exists at this OID string` instead of `noSuchInstance`.
- **No such object errors** are mapped as `error: no such object currently exists at this OID string` instead of `noSuchObject`.
- **End of MIB view errors** are mapped as `error: end of MIB view` instead of `endOfMibView`.
- An **unknown variable type** falls back to the `string` representation instead of logging an error as it did in with the stand-alone `logstash-input-snmp`. This change should not affect existing pipelines, unless they have custom error handlers that rely on specific error messages.


### Migration notes: `logstash-input-snmptrap`

As a component of the new `logstash-integration-snmp` plugin, the `logstash-input-snmptrap` plugin offers *almost the same capabilities* as the stand-alone [logstash-input-snmptrap](https://github.com/logstash-plugins/logstash-input-snmptrap) plugin.
You might need to address some behavior changes depending on your use case and how the ingested data is being handled through the pipeline.

### Changes to mapping and error logging: `logstash-input-snmptrap`

- The **PDU variable bindings** are mapped into the Logstash event using the defined data type. By default, the stand-alone `logstash-input-snmptrap` plugin converts all of the data to `string`, ignoring the original type. If this behavior is not what you want, you can use a filter to retain the original type.
- **SNMP `TimeTicks` variables** are mapped as `Long` timestamps instead of formatted date string (`%d days, %02d:%02d:%02d.%02d`).
- **`null` variables values** are mapped using the string `null` instead of `Null` (upper-case N).
- **No such instance errors** are mapped as `error: no such instance currently exists at this OID string` instead of `noSuchInstance`.
- **No such object errors** are mapped as `error: no such object currently exists at this OID string` instead of `noSuchObject`.
- **End of MIB view errors** are mapped as `error: end of MIB view` instead of `endOfMibView`.
- The previous generation (stand-alone) input-snmptrap plugin formatted the **`message` field** as a ruby-snmp `SNMP::SNMPv1_Trap` object representation.
  ```
  <SNMP::SNMPv1_Trap:0x6f1a7a4 @varbind_list=[#<SNMP::VarBind:0x2d7bcd8f @value="teststring", @name=[1.11.12.13.14.15]>], @timestamp=#<SNMP::TimeTicks:0x1af47e9d @value=55>, @generic_trap=6,  @enterprise=[1.2.3.4.5.6], @source_ip="127.0.0.1", @agent_addr=#<SNMP::IpAddress:0x29a4833e @value="test">, @specific_trap=99>
  ```
  The new integrated `input-snmptrap` plugin uses JSON to format **`message` field**.
  ```
  {"error_index":0, "variable_bindings":{"1.3.6.1.6.3.1.1.4.1.0":"SNMPv2-MIB::coldStart", "1.3.6.1.2.1.1.3.0":0}, "error_status":0, "type":"TRAP", "error_status_text":"Success", "community":"public", "version":"2c", "request_id":1436216872}
  ```


### Maintain maximum compatibility with previous implementation

If needed, you can configure the new `logstash-integration-snmp` plugin to maintain maximum compatibility with the previous (stand-alone) version of the [input-snmp](https://github.com/logstash-plugins/logstash-input-snmp) plugin.
```
input {
   snmptrap {
    use_provided_mibs => false
    oid_mapping_format => 'ruby_snmp'
    oid_map_field_values => true
   }
}
```


## Importing MIBs

The SNMP plugins already include the IETF MIBs (management information bases) and these do not need to be imported. To disable the bundled MIBs set the `use_provided_mibs` option to `false`.
Any other MIB will need to be manually imported to provide mapping of the numeric OIDs to MIB field names in the resulting event.
To import a MIB, the OSS [libsmi library](https://www.ibr.cs.tu-bs.de/projects/libsmi/) is required. libsmi is available and installable on most operating systems.
To import a MIB, you need to first convert the ASN.1 MIB file into a `.dic` file using the libsmi `smidump` command line utility.
**Example (using `RFC1213-MIB` file)**
```
$ smidump --level=1 -k -f python RFC1213-MIB > RFC1213-MIB.dic
```

Note that the resulting file as output by `smidump` must have the `.dic` extension.
When `mib_paths` is configured, note that the complete set of all MIBs required must be included here. Any conflicting MIBs with defaults will prefer the default, not the MIB configured on the `mib_paths`. The tightest control and maximal compatability can be acheived by ensuring that `use_provided_mibs` is set to `false` and the complete set of MIBs are included in files listed in `mib_paths`. The defaults from the old implementation can be found in the [ruby-snmp github repository](https://github.com/ruby-snmp/ruby-snmp/tree/master/data/ruby/snmp/mibs).

### Preventing a `failed to locate MIB module` error

The `smidump` function looks for MIB dependencies in its pre-configured paths list. To avoid the `failed to locate MIB module` error, you may need to provide the MIBs locations in your particular environment.
The recommended ways to provide the additional path configuration are:
- an environment variable, or
- a config file to provide the additional path configuration.

See the "MODULE LOCATIONS" section of the [smi_config documentation](https://www.ibr.cs.tu-bs.de/projects/libsmi/smi_config.html#MODULE%20LOCATIONS) for more information.

### Option 1: Use an environment variable

Set the `SMIPATH` env var with the path to your MIBs. Be sure to include a prepended colon (`:`) for the path.
```
$ SMIPATH=":/path/to/mibs/" smidump -k -f python CISCO-PROCESS-MIB.mib > CISCO-PROCESS-MIB_my.dic 
```


### Option 2: Provide a configuration file

The other approach is to create a configuration file with the `path` option. For example, you could create a file called `smi.conf`:
```
path :/path/to/mibs/
```

And use the config with smidump:
```
$ smidump -c smi.conf -k -f python CISCO-PROCESS-MIB.mib > CISCO-PROCESS-MIB_my.dic
```