Erlang logo
User's Guide
Reference Manual
Release Notes
PDF
Top

Simple Network Management Protocol (SNMP)
User's Guide
Version 5.1.1


Expand All
Contract All

Chapters

14 Audit Trail Log

The chapter Audit Trail Log describes the audit trail logging.

Both the agent and the manager can be configured to log incoming and outgoing messages. It uses the Erlang standard log mechanism disk_log for logging. The size and location of the log files are configurable. A wrap log is used, which means that when the log has grown to a maximum size, it starts from the beginning of the log, overwriting existing log records.

The log can be either a read, write or a read_write.

14.1  Agent Logging

For the agent, a write, means that all set requests and their responses are stored. No get requests or traps are stored in a write. A read_write, all requests, responses and traps are stored.

The log uses a raw data format (basically the BER encoded message), in order to minimize the CPU load needed for the log mechanism. This means that the log is not human readable, but needs to be formatted off-line before it can be read. Use the function snmpa:log_to_txt for this purpose.

14.2  Manager Logging

For the manager, a write, means that all requests (set and get) and their responses are stored. No traps are stored in a write. A read_write, all requests, responses and traps are stored.

The log uses a raw data format (basically the BER encoded message), in order to minimize the CPU load needed for the log mechanism. This means that the log is not human readable, but needs to be formatted off-line before it can be read. Use the function snmpm:log_to_txt for this purpose.