Monitoring and Diagnostics

• Checking Server Memory Usage
• collStats Command
• Database Profiler
• mongoperf
• Munin configuration examples
• serverStatus Command
• Http Interface
• mongostat
• mongosniff

• Admin UIs

Mongostat

mongostat is a great utility which exposes many internal MongoDB metrics. For any MongoDB related issues it is a good start for the analysis of performance issues.

Query Profiler

Use the Database Profiler to analyze slow queries.

db.currentOp() is another way to get a snapshot of what is currently happening.

Http Console

The mongod process includes a simple diagnostic screen at http://localhost:28017/. See the Http Interface docs for more information.

mongo Shell Diagnostic Commands

• db.serverStatus()
  • See the serverStatus Command page.
• db.stats()
  • Stats on the current database. Command takes some time to run, typically a few seconds unless the .ns file is very large (via use of --nssize). While running other operations may be blocked.
  • fileSize is the total size of all files allocated for the db.
• db.foo.find().explain()
  • explain plan
• help
  • db.help()
  • db.foo.help()

Trending/Monitoring Adaptors

• munin
  • Server stats: this will retrieve server stats (requires python; uses http interface)
  • Collection stats, this will display collection sizes, index sizes, and each (configured) collection count for one DB (requires python; uses driver to connect)
• Ganglia:
  • ganglia-gmond
  • mongodb-ganglia
• cacti
• Mikoomi provides a MongoDB plugin for Zabbix
• Nagios
• mtop - A top like utility for Mongo
• Mongo Live - A Chrome extension that provides a real-time server status view (uses the rest interface).

Chris Lea from (mt) Media Temple has made an easy to install Ubuntu package for the munin plugin.

Hosted Monitoring

• MongoDB Monitoring Service (MMS) is a free hosted monitoring tool for MongoDB provided by 10gen
• Server Density provides hosted monitoring for your hardware and software infrastructure, and supports a number of status checks for MongoDB.
• Cloudkick
• scout app slow queries
• AppFirst

Database Record/Replay (diagLogging command)

Recording database operations, and replaying them later, is sometimes a good way to reproduce certain problems in a controlled environment.

To enable logging:

db._adminCommand( { diagLogging : 1 } )

To disable:

db._adminCommand( { diagLogging : 0 } )

Values for diagLogging:

• 0 off. Also flushes any pending data to the file.
• 1 log writes
• 2 log reads
• 3 log both
  Note: if you log reads, it will record the findOnes above and if you replay them, that will have an effect!

Output is written to diaglog.bin_ in the /data/db/ directory (unless --dbpath is specified).

To replay the logged events:

nc ''database_server_ip'' 27017 < ''somelog.bin'' | hexdump -c