High Availability Flashcards
What are the polling methods in SQS?
short polling: Call might not return a message
long polling: Waits until a message is present and then returns it
Is SQS inter-regional?
No, SQS is present in each region and they cannot share messages
What is the format of SQS messages?
XML, JSON or Text - up to 256kb
Auto Scaling Group: What are the required / optional components?
Required: Launch configuration & Launch Group
Optional: Scaling policy & Scheduled actions
What is Auto Recovery on EC2?
Automatic recovery when an instance fails due to hardware failure.
This needs a CloudWatch Alarm.
An recovered instance has the same ID, IP and metadata
What is Fault Tolerance?
And how does it differ from High Availability?
System will continue to function without degradation of performance, even when one compontent failed.
High Availability on the other hand can have some performance issues
When can a queue be automatically deleted? (w/o notification)
30 days after no action has been performed on the queue
How to create a Highly Available Bastion Host?
Create an autoscaling group with multiple AZs and set the min./max. isntances to 1
DR: What are two typical features of a “pilot light”?
The typical Pilot Light scenario is to have your database server mirrored in your cloud environment. Another feature of Pilot Light is to have Amazon Machine Images of your servers that can be used to create cloud servers in minutes.
What is the difference between elasticity and scalability?
Examples for EC2
elasticity: Reacting on short term changes in demand (using autoscaling on demand instances)
scalability: Bulding infrastructure for long term use
EC2:
e: adding more instances
s: provisioning larger instances
What services have maintenance windows?
RDS DynamoDB DAX DocumentDB Elasticache Redshift Neptune
SQS: Can you delete an SQS queue with messages still in it?
Yes
SQS: What are the standard retention rates of a message?
Default 4 days
Up to 14 days
SQS: What is the Visibility Timeout?
Duration of an in-flight message that is invisible to other worker instances
SQS: Is a message available in all regions?
No
SQS: What is the format / size structure of a message?
Text, XML, JSON
up to 256kb
SNS: What is the format / size structure of a message?
Text, XML, JSON
up to 256kb
cannot be deleted after published
How can you achieve HA für private subnet internet access?
Put a NAT Gateway in more than one AZ, instances in other AZs can use the NAT
Auto Scaling: What is the Auto Scaling Policy?
Conditions to launch / terminate instances.
If condition is met, an Alarm is st
Auto Scaling: What are Scheduled Actions?
Defines the min. / max. / desired capacity for a point in time in the future
RDS: What are the limits for Read Replicas?
mysql, postgres: 5
aurora: 15
DynamoDB: How to replicate data between regions?
Using DynamoDB Streams replicates the data if modified from the one table to the other
Route53: What is the difference between Active-Active Failover and Active-Passive failover?
A-A: All instances are active and get traffic, if one set goes down they will just not receive anymore traffic
A-P: A secondary set of instances is on standby in case of a failure
Route53: What are the three HealthChecks?
- Monitoring an endpoint
- Monitoring an Alarm
- Monitoring the status of other health checks
What is the considered state of an instance that has no health check?
Healthy
How would a failover scenario for two regions be structured?
Set up a latency rule for the regions and a weighted rule for the instances inside the region.
With the Evaluated Target Health settings create a tree that goes from bottom to top if instances fail.
VPN: How to create a HA VPN connection?
Create multiple Customer Gateways (Virtual Private Gateway hast two connections per default)
VPN: How to create a private HA VPN connection?
Create multiple Direct Connect connection, if possible with multiple providers
VPN: How to create a cost effective (private) VPN connection
Create a Direct Connect connection as primary and a public VPN connection through the internet
What is RTO/RPO?
Recovery Time Object: How long can it take until the system is up again
Recovery Point Object: How much data can be lost
What is the Pilot Light Method?
(Cheapest DR method)
Minimal setting always running as backup, only the most critical processes
What is the Warm-Standby Method?
Scaled down version of the system running to satisfy the customers.
Instances should be scaled up in case of a DR.
What is the Multi Site Solution method?
Two complete identical sites both running an serving traffic.
COuld be on.-prem/cloud or cloud/cloud
What is the Auto Scaling Policy?
Set of instructions sent by Amazon CloudWatch to the Auto Scaling group is
referred to as a policy. The policy defines what the Auto Scaling group should do with
the alarm it receives from Amazon CloudWatch.
What is predictive scaling?
Predictive scaling is a feature of AWS Auto Scaling that can look back at previous
activity and use that to schedule the needed scaling changes based on both daily and
weekly patterns.
How much data does predictive scaling need?
About two weeks
Who can be the target of SNS?
. There are five different ways to subscribe to an SNS topic. They are AWS Lambda,
Amazon Simple Queue Service (SQS), HTTP and HTTPS, email, and SMS text
How to prevent an ELB to be deleted?
Enable deletion protection.
not available for classic LBs
