In the previous articles , I have explained , How you can
Monitor and auto-recover repaired EC2 instances
Monitor AWS resources using Cloudwatch Alarms
In this guide , We will learn How we can monitor the RDS instances using Cloudwatch.
By default , RDS instance send metrics to the Cloudwatch Metrics.
Using the cloudwatch metrics , We can monitor the RDS instances for parameters such as Read IOPS , Write IOPS , Free Storage , CPU utilization.
Monitoring RDS Instances
As said earlier , We are going to setup monitoring for
- CPU utilization
- Free Storage
- Read IOPS
- Write IOPS
To create cloudwatch alarms , Login to Cloudwatch console.
In the left navigation pane , Choose Metrics
From the lists of available metrics , Choose RDS
You can check the metrics of the RDS DB instance by
- Per-Database Metrics
- By Database class
- By Database Engine
- Across All Databases
For example , If you want to check the metrics for particular DB instance , Choose Per-Database Metrics
There you can find the List of the databases and their respective metric name such as Free Memory , Freestoragespace , WriteIOPS , ReadIOPS etc.
If you select the particular DBInstanceIdentifier and their Metric name
A graph will be plotted for the DB instance with the metrics.
If you want to check the metrics for a particular time duration , You can adjust it.
Now using these metrics lets setup a alarm , So that when a DB instance metric cross a threshold we set , For example , CPU usage is above 70/100 (in Percentage), Trigger the cloudwatch alarm and alert it to the medium (email or SMS).
Services Used for Monitoring and Alerting
- Cloudwatch Alarms
- SNS Topic with Subscriber
Creating Cloudwatch Alarms
Lets go ahead and create a Cloudwatch alarm for
CPU Utilization
Login to Cloudwatch console , In the left pane , Choose Alarms
Then Click Create alarm
Under Specify metric and conditions , Click Select metric
From the lists of available metrics , Choose RDS
and then click Per-Database Metrics
Select the DBInstanceIdentifier with the Metric Name CPUUtilization
Then choose Select metric
Under statistic Choose Minimum and Under Period Choose 1 minute
and Under Conditions
Threshold type should be static and whenever CPUUtillization is
Greater/Equal to 70 percent
Which means when the metric value goes above the threshold (70 percent) for 1 datapoint within 1 minute , The cloudwatch alarm will be triggered and will alert us.
Click Next , Here we are going to setup alerting for two cases.
- If the metric value goes above the threshold (In alarm)
- If the metric value goes below the threshold (OK)
Here we have to either use an existing SNS topic or We can create a new topic.
If you want to setup a SNS topic with a subscribers , Refer here
Once the SNS setup is completed , You can resume with the Notifications.
Or if you want to create SNS topic from cloudwatch console , Select Create new topic
Provide a name for the SNS topic and add the email addresses (who should receive the notifications) and then click Create topic
Once it is created , You should see as shown below.
We have now added for In alarm notification.
To get notifications for OK status , Choose Add notification and then select OK and select existing topic and we already have created a topic with a subscriber.
and then click Next
Provide a name for the alarm , For better identification , provide a name as follows,
Environment-DBname-Metricname
You can also add a short description for each alarm you create.
Click Next and then Choose Create Alarm.
Read IOPS
Lets create a alarm for Read IOPS.
We are going to following the same procedure , The only place that need modification is Specify metric and conditions
Here we are going to choose ReadIOPS
Under Conditions , For static threshold , Lets say your DB usually gets a ReadIOPS count of 150 , and anything above 150 should be considered as anomaly.
Click Next and Setup notifications for In alarm and OK state as discussed above.
You can use the same SNS topic.
Provide a name for the alarm and then click Create alarm
Write IOPS
Lets create a alarm for Write IOPS.
We are going to following the same procedure , The only place that need modification is Specify metric and conditions
Here we are going to choose WriteIOPS
Under Conditions , For static threshold , Lets say your DB usually gets a WriteIOPS count of 250 , and anything above 300 should be considered as anomaly.
Click Next and Setup notifications for In alarm and OK state as discussed above.
You can use the same SNS topic.
Provide a name for the alarm and then click Create alarm
Free Storage Space
Lets create a alarm for Free storage space
We are going to following the same procedure , The only place that need modification is Specify metric and conditions
Here we are going to choose FreeStorageSpace
Under Conditions , For static threshold , Lets say , The RDS DB instance has the storage size of 200 GB in total and If the free storage size goes less than 25 GB , It should trigger the alarm and alert us.
and while giving the threshold value , 25 GB should be converted into bytes.
Whenever FreeStorageSpace is Lower/Equal to 25 GB.
Click Next and Setup notifications for In alarm and OK state as discussed above.
You can use the same SNS topic.
Provide a name for the alarm and then click Create alarm.
Conclusion
We have successfully implemented a monitoring and alerting mechanism using Cloudwatch alarm and SNS topic.
Hope you find it helpful. Thanks for reading this article.
Please check my other articles.