peabody - back in time logging

Peabody is an extension for a few logging frameworks to allow back in time logging.

Usually you keep your logger to the INFO level, but as soon as something goes wrong you need to raise the log level to DEBUG just... before something went wrong.

Peabody does exactly this: detects an error, goes back in time and rise the log level for you.

Behind the scenes the trick is simple: it keeps a small buffer of recently logged messages that were below the current threshold. As soon as a message is logged with level ERROR or higher the whole buffer is written to the log file.

Here's a real world example. Without peabody:

2013-05-10 17:25:49,717 | INFO | BEGIN fetchTokenByMessageId
2013-05-10 17:25:55,999 | ERROR | Search failed. msgId: '[email protected]' - LD0999 - Error - it.zzz.messages.ricercaDocumento.result.Eccezione@24eecb5c

With peabody (in bold infos otherwise lost):

2013-05-10 17:25:49,717 | INFO | BEGIN fetchTokenByMessageId
2013-05-10 17:25:55,998 | ERROR | >>>>>>>>> FLASHBACK START >>>>>>>>>
2013-05-10 17:25:49,717 | INFO | BEGIN fetchTokenByMessageId
2013-05-10 17:25:49,718 | DEBUG | Start unmarshaling
2013-05-10 17:25:49,719 | DEBUG | Completed unmarshaling
2013-05-10 17:25:49,719 | DEBUG | Reading kdv doc 5666673
2013-05-10 17:25:49,818 | DEBUG | messageId '[email protected]'
2013-05-10 17:25:49,818 | DEBUG | Ref token null
2013-05-10 17:25:49,818 | DEBUG | Read legaldoc document for messageID '[email protected]', classeDocumentale: 'zzz-mail-search'
2013-05-10 17:25:55,999 | ERROR | <<<<<<<<< FLASHBACK END <<<<<<<<<
2013-05-10 17:25:55,999 | ERROR | Search failed. msgId: '[email protected]' - LD0999 - Error - it.zzz.messages.ricercaDocumento.result.Eccezione@24eecb5c

Supported frameworks:

• log4j through a custom LoggerFactory. Just declare a custom categoryFactory in your config file. See src/test/resources for examples.

Config properties (default value)

• triggerLevel: which level triggers the "flashback" (ERROR)
• bufferSize: how many messages to keep (100)
• loggerList: csv list of logger name prefixes that should be included (include all)
• discardOldEventsThreshold: on flashback do not log messages older than this time (millis, 10000)
• collectLocationInformation: enable the use of %C, %M, %L in PatternLayout for recorded messages (true)
• keepLoggedEvents: flashback also includes lines already logged for better readability (true)

Performance

For a typical application the overhead should be negligible. It accounts mostly to the Pattern formatting and time and location informations collection (if enabled) for all buffered messages. If you have a log line inside a tight loop you should probably consider to interely avoid the logger call by protecting the invocation checking on the current logger level (e.g. logger.isDebugEnabled()) anyway. If in doubt, check with a profiler.