Monitoring Metrics Alerts

Here are some metrics to watch for, in order to avoid a critical situation leading to a lose in performance.

Page Faults

• Normal : 30 - 45% Req

• Worrying : 50 - 60% Req

• Critical : > 75% Req

  • Add Shard.
  • Add RAM.

-> Working set does not fit in RAM.

Replication Lag

• Normal : Ideal 0s

• Worrying : > 60s

• Critical : > 240s

  • Bandwidth insufficient (Among others).

Disk Space

• N : < 80%

• W : > 85%

• C : > 94%

  • Add storage.

Slow Query

• N : < 250ms

• W : +50%

• C : +70%

  • Improve Indexes & Queries.

-> Bad Queries, Bad Indexes.

CPU

• N : ...

• W : +50%

• C : *2

  • Add Java/Mongo node.
  • Add CPU.

NetOut

• N : <80% Bandwidth

• W : >85% Bandwidth

• C : >95% Bandwidth

  • Add Bandwidth.

-> Not enough Bandwidth.

? Number of connections ?

• N : ...

• W : +50%

• C : *2

  • Add Shard.

-> More connections to Mongo.

DB Lock %

This indicator can produce many false positive.

• N : Depends on usage, >60% if write regularly, <10% if read majority.

• W : *2

• C : > 80%

  • Add Shard.
  • Improve Indexes.