MongoDB profiler and database performance problem diagnosis and identification

MongoDB profiler is an internal tool that enables DBAs and developers to collect information about executed MongoDB requests.

Profiler has three levels:

0 – profile is off

1 – profiler collects information on slow requests

2 – profiler collects all information

For production purposes level 1 is what DBA needs. In general, it is best to set it to some reasonable SLA required threshold, like 1,000 milliseconds (1 second), etc.

Level 2 is best for development purposes to get a complete log of executed requests.

Below is the list of profile related commands:

# set profiling to level 1 capture requests with duration over 500 ms
db.setProfilingLevel(1,500)
#get current profiler level
db.getProfilingLevel()
# get current profiler settings
MongoDB Enterprise > db.getProfilingStatus()
{ "was" : 1, "slowms" : 500 }

To return operations slower than 500 milliseconds, run a query similar to the following:

db.system.profile.find( { millis : { $gt : 50 } } ).pretty()

Enteros UpBeat High Load Capture continuously captures history of executed mongodb requests across all collection as well as other mongo db, OS, SAN and other operational statistics across muliple systems.

When the spike is identified, it can be cross-correlated across MongoDB nodes and correlated with MongoDB requests. A detailed report is generated to pinpoint what requests are were active during the problem time.

Also, please check my presentation on MongoDB performance tuning