Aurora Flashcards
What is Aurora and how does it relate to MySQL and Postgres?
How does the storage grow?
How many replicas can it have and what is the replication speed?
How does failover work?
What about cost?
Cloud native relational database with 5x performance improvements over MySql and 3x Performance increase over Postgres. Compatible with Postgres or MySQL depending on the driver you use.
Automatically growing storage in increments of 10GB up to 64TB
Can have between 5 and 15 replicas (MySQL has and the replication process is faster - sub 10 ms replica lag)
Failover in Aurora is Instantaneous (~30seconds) (It’s HA Native)
20% more cost
Walk me through the high availability and failover setup of Aurora by drawing a diagram and talking through it. ?
How many AZs?
How many write replicas?
How many read replicas can aurora have?
How does scaling work?
Is there support for cross region replication?
,,,
Aurora stores 6 copies of your data across 3AZ
1 write node (Master)
5-15 Read replicas
Read replicas can be auto scaling
Yes, there is support for cross Region Replication
In terms of Aurora High Availability, how many nodes are needed for reads and writes not to fail?
How is the data stored? Is it just one volume?
Anything else special about the data?
How long does failover take from any read replica to master?
4 copies out of 6 needed for writes (this means 2 can fail)
3 copies out of 6 needed for reads (this means 3 can fail)
Storage is striped across 100s of volumes
Will self heal if there is data corruption
Failover from any read replica to master is less than 30 seconds
What types of endpoints does an Aurora cluster have?
How does Aurora ensure that you always have the correct number of read replicas?
How does aurora know when to scale the read replicas?
1 Writer Endpoint - DNS Name Pointing to Master
1 Reader Endpoint - That will direct connection to on specific node
Read replicas can be auto-scaling
Aurora scales the number of read replicas based on metrics like CPU usage.
How much downtime is there for patching in Aurora?
Zero down time
What is Backtrack?
Restore data at any point of time without using backups
Why would you want one replica to be larger than another?
How would you access that one replica?
For large analytical queries aka reporting
These special replicas would then be accessed via a custom endpoint
How does aurora security differ from RDS Security?
It doesn’t, they’re exactly the same.
What is Aurora serverless and how does it work?
What’s the difference between serverless and autoscaling
When would you use something like this?
What do clients connect to?
How does pricing work for Aurora serverless?
On-demand pay-per second DB instances that scale based on usage.
No Capacity planning, potentially more cost effective
Good for intermittent or unpredictable workloads
Client connects to Proxy fleet
You pay per second
What is aurora multi-master and what is it used for?
Aurora multi-master in a configuration in which all nodes are used for reads and writes.
You would use it in cases where you want immediate failover for write nodes.
What is aurora cross region replication? What is a use-case for cross-region replication?
Useful for disaster recovery
Easier than setting up Aurora Global Database
What is an aurora global database and what are some of its uses?
How many primary regions does it have?
How many secondary regions does it have? What is the cross region replication time
How many read replicas can it have within a region
How long does it take to promote a secondary region to a primary region?
AKA what is the Recovery Time Objective (RTO)
…
1 Primary region read/write
Up to 5 secondary read-only regions. Replication lag is < 1s
Up to 16 read replicas per secondary region
Promoting another region for disaster recovery has an
Recovery Time Objective (RTO) of < 1min
Why is it useful to have replicas in multiple regions
To reduce latency
Why is Aurora integrated with Machine learning
To add machine learning based predictions to your application via SQL
What are the supported services what what are they used for?
What are some additional usecases for machine learning with Aurora
SageMaker - Allows you to use any kind of machine learning model on the back end
Amazon Comprehend - For sentiment analysis
Fraud detections, ads targeting, sentiment analysis & product recommendations.
