# MongoDB Diagnostics and Debugging (Version 3.4) <br />Ch 3: Slow Queries

###### tags: `MongoDB University M312`

## What Slow Queries Look Like in an Application

### Response Time

* End to end application request should be < 200ms
* \> 100ms is considered slow for MongoDB by default

## Response Time Degredation

### The Culprits

* Working set exceeding RAM
* Queries takes longer as data set grows
* Growing pool of clients
* Unbounded array growth
* Excessive number of indexes

:::info
Mongostat, mlogvis and the porfiler are here to help to resolve the slowness problem.
***# Make sure to use profiler only in development***
:::

## Throughput Drops

### Possible Causes for a Drop in Throughput

* Long-running queries
* Collection scan
* Poorly anchored regex
* Inefficient index usage
* Index Builds
* Write Contention

### If You Suspect an Increase in Long-Running Queries

* Server log
* db.currentOp()
* Turn on the profiler(But further slowdown)

## Impact of Application Changes

### Detect Application Changes

* Keep historical monitering data
* Watching out for spike in connections
* Watching out for spike in ops/s
* Spotting long lasting, non-indexed queries

### Connections Spikes

* Exhausting connection pools
* Incorrect connection to production environment
* Analytical workload and reporting tools

## Using Mtools to Find Slow Queries

with 3 tools
* mloginfo
* mplotqueries
* mlogvis

使用 mtools 快速啟動 Replica set

mlaunch init --replicaset --name m312RS --wiredTigerCacheSizeGB 0.3 --port 30000 --host localhost

## Fixing Missing Indexes

### Building an Index Options
* Build the index on the primary from the shell
* Build the index in the Background on the primary on the shell
* Build the index from the Compass
* Build the index on each member of the replica set
* Use Ops manager and Cloud manager to build the index

# check the log file to find if the index is building background 
mlogfilter $logfile --component INDEX
mlogfilter $logfile --component COMMAND