OnCallReadyStart learning

AWS II: VPC, EC2, ELB & EKS: glossary

90 terms, in plain words, with where each one is taught.

access entry (access entries)
An EKS object that maps an IAM user or role to a Kubernetes identity in a cluster, optionally with access policies or Kubernetes groups. Lesson 35.29
access policy (EKS) (access policies, AmazonEKSClusterAdminPolicy, AmazonEKSEditPolicy)
An AWS-defined set of Kubernetes permissions (cluster admin, admin, edit, view) attached to an access entry for the whole cluster or chosen namespaces. Lesson 35.29
alias record (alias records, Route 53 alias)
A Route 53 record that points at an AWS resource such as a load balancer and is answered with its current addresses; it works at the zone apex. Lesson 35.19
AmazonSSMManagedInstanceCore
The AWS managed policy an instance role needs so the SSM agent can register and accept Session Manager sessions. Lesson 35.17
AMI (AMIs, Amazon Machine Image)
Amazon Machine Image: the disk image an EC2 instance starts from, with the operating system and whatever was baked into it. IDs differ per Region. Lesson 35.14
Application Load Balancer (ALB (AWS), ALBs)
The AWS HTTP load balancer: listeners with rules by host, path or header, target groups of instances, IPs or Lambda functions, health checks by status code. Lesson 35.19
AssumeRoleWithWebIdentity
The STS call that exchanges an OIDC token (a pod's IRSA token, a CI job's token) for a role's temporary credentials, if the role's trust policy accepts the token. Lesson 35.29
authentication mode (EKS) (API_AND_CONFIG_MAP, CONFIG_MAP)
Which mapping an EKS cluster accepts: only the aws-auth ConfigMap, access entries and the ConfigMap, or only access entries. Lesson 35.29
Auto Scaling group (ASG, ASGs, Auto Scaling groups)
A set of EC2 instances that AWS keeps at a desired number from a launch template, spread over zones, replacing unhealthy ones. Lesson 35.24
AWS Load Balancer Controller (LB Controller, aws-load-balancer-controller, LBC)
A controller in an EKS cluster that turns Ingresses into ALBs and LoadBalancer Services into NLBs, with targets that can be pod IPs. Lesson 35.33
aws provider (hashicorp/aws, Terraform aws provider)
The Terraform plugin that manages AWS resources by calling the AWS APIs with the same credentials the AWS CLI would use. Lesson 35.12
aws-auth ConfigMap (aws-auth)
The old way to map IAM roles and users into an EKS cluster: a ConfigMap in kube-system. One bad edit can lock everyone out. Lesson 35.29
blackhole route (blackhole)
A route whose target (for example a deleted NAT gateway) no longer exists. Traffic to it is dropped silently until the route is changed. Lesson 35.7
cloud-init
The program on cloud images that sets up an instance at first boot: hostname, SSH keys, packages, and the user data. Its output is in /var/log/cloud-init-output.log. Lesson 35.14
console output (get-console-output, serial console output)
What an instance printed to its serial console while booting, readable from outside with aws ec2 get-console-output - no access to the instance needed. Lesson 35.14
default VPC
The VPC every AWS Region comes with (172.31.0.0/16, a public subnet in every zone, an internet gateway attached). Handy for experiments, wrong for production. Lesson 35.1
default_tags
A Terraform aws provider setting that puts the same tags on every resource it creates; they show up in tags_all. Lesson 35.12
deregistration delay (connection draining)
How long a load balancer lets open requests finish on a target that is being removed, before cutting it off (300 seconds by default). Lesson 35.19
desired capacity (desired)
The number of instances an Auto Scaling group keeps running, between its minimum and maximum. Lesson 35.24
EBS volume (EBS volumes, EBS (volumes))
A network disk for an EC2 instance in the same Availability Zone. It can be grown online and snapshotted; the partition and file system inside must be grown too. Lesson 35.14
EC2 instance (EC2 instances, instance (EC2))
A virtual machine rented from AWS by the second, defined at launch by an image, an instance type, a subnet, security groups, a role and user data. Lesson 35.14
EKS (in depth) (Amazon EKS cluster, EKS cluster)
Amazon's managed Kubernetes: AWS runs the control plane across three zones and gives you an endpoint; you run the nodes and the workloads. Lesson 35.26
EKS add-on (EKS add-ons, add-on (EKS))
A cluster component EKS installs and upgrades for you, such as vpc-cni, coredns, kube-proxy or the pod identity agent. Lesson 35.26
EKS Pod Identity (Pod Identity, pod identity association)
The EKS way to give pods an IAM role: an agent add-on on each node and an association of namespace and service account to a role trusted by pods.eks.amazonaws.com. Lesson 35.29
eks-pod-identity-agent
The EKS add-on (a DaemonSet) that serves pods their Pod Identity credentials at 169.254.170.23. Lesson 35.29
Elastic IP (Elastic IPs, EIP, EIPs)
A public IPv4 address you allocate to your account and attach to an instance or a NAT gateway; it stays yours until you release it. Lesson 35.7
enableDnsHostnames (DNS hostnames)
A VPC setting that gives instances DNS names (ip-10-20-0-5...). Off by default in a new VPC; interface endpoints need it for private DNS. Lesson 35.1
endpoint policy (endpoint policies)
A policy on a VPC endpoint that limits what can be reached through it, for example only your own S3 buckets. Lesson 35.7
ephemeral port (ephemeral ports)
The temporary source port a client picks for a connection, somewhere in 1024-65535. Replies go back to it, so a stateless firewall must allow that range. Lesson 35.4
extended support (EKS) (EKS extended support)
The 12 months after an EKS version's 14 months of standard support, during which the control plane costs six times more per hour. Lesson 35.26
fail open (ELB) (fail-open)
What an ALB does when every target in a group is unhealthy: it sends traffic to all of them anyway, assuming the health check is wrong. Lesson 35.19
gateway endpoint (gateway endpoints)
A VPC endpoint for S3 or DynamoDB: a route to the service's prefix list in the route tables you choose. No network interface, no charge. Lesson 35.7
golden image (golden images, golden AMI)
An image a team builds with its application and settings already installed, so new instances only need to boot instead of installing at boot. Lesson 35.14
gp3
The general-purpose EBS SSD volume type: 3,000 IOPS and 125 MB/s at any size, with more of either bought separately. Lesson 35.14
grace period (ASG) (health check grace period)
How long after launch an Auto Scaling group ignores failed load balancer health checks, so new instances have time to start. Lesson 35.24
Graviton
AWS's own arm64 processors, used by instance types with a g (t4g, m7g). Usually cheaper and faster per dollar, but images and binaries must be arm64. Lesson 35.14
growpart
A command that grows a partition to fill the free space on its disk: growpart /dev/nvme0n1 1 (disk, space, partition number). Lesson 35.14
health check (ELB) (ELB health check, target health)
The load balancer's regular probe of each target on a port and path; enough passes make it healthy, enough failures unhealthy. describe-target-health shows the state and a reason code such as Target.Timeout or Target.ResponseCodeMismatch. Lesson 35.19
hop limit (HttpPutResponseHopLimit)
The IP TTL on metadata responses. 1 lets only the instance itself reach IMDS; 2 also lets containers on it (such as EKS pods) reach the node's role. Lesson 35.17
hosted zone (hosted zones)
A DNS zone (for example shop.example.com) whose records Route 53 serves; a public zone for the internet or a private zone for VPCs. Lesson 35.19
IMDSv2 (IMDSv1)
The token version of the instance metadata service: a client first PUTs for a session token and sends it with every request. IMDSv1 answered any plain GET. HttpTokens=required turns v1 off; it blocks the SSRF credential theft that hit Capital One in 2019. Lesson 35.17
IngressClass alb (ingressClassName: alb)
The Ingress class the AWS Load Balancer Controller handles; an Ingress with it becomes an ALB. Lesson 35.33
IngressGroup (group.name, Ingress group)
Several Ingresses sharing one ALB, set with the alb.ingress.kubernetes.io/group.name annotation; avoids one load balancer per Ingress. Lesson 35.33
instance metadata service (IMDS, metadata service)
A service at 169.254.169.254 that only the instance itself can reach: its ID, Region, network details, user data and its role's temporary credentials. Lesson 35.17
instance profile (instance profiles)
The container that attaches an IAM role to an EC2 instance; the instance gets the role's temporary credentials through the metadata service. Lesson 35.17
instance refresh (instance refreshes)
An Auto Scaling operation that replaces the group's instances in batches with ones from the current launch template, keeping a minimum share healthy. Lesson 35.24
instance type (instance types)
The size and kind of an EC2 machine (t4g.small: family t, generation 4, Graviton, size small): CPUs, memory, network and architecture. Lesson 35.14
interface endpoint (interface endpoints, PrivateLink)
A VPC endpoint that puts a network interface with a private IP into your subnets for one AWS service, with a security group and optional private DNS. Billed per hour and per GB. Lesson 35.7
internet gateway (IGW, IGWs, internet gateways)
The VPC's door to the internet. A route to it plus a public IP on an instance makes the instance reachable from outside and able to reach out. Lesson 35.1
IRSA (IAM roles for service accounts)
IAM roles for service accounts: pods get a role by presenting a token from the cluster's OIDC issuer, which the role's trust policy accepts for one exact service account. Lesson 35.29
launch template (launch templates)
A saved, versioned set of instance launch settings (image, type, groups, role, user data, metadata options) used to launch instances and by Auto Scaling groups. Lesson 35.14
listener (ELB) (listener, listeners)
The port and protocol a load balancer accepts connections on, with rules that decide where each request goes and a default action. Lesson 35.19
local route
The route every AWS route table starts with: the whole VPC range goes to local, so all subnets of the VPC can reach each other. It cannot be removed. Lesson 35.1
main route table
The route table created with a VPC. Every subnet that is not explicitly associated with another table uses it, so it should stay local-only. Lesson 35.1
managed node group (managed node groups, node group (EKS))
EKS worker nodes created and upgraded by EKS as an Auto Scaling group of EKS-optimized images. Lesson 35.26
max pods (maximum pods per node)
How many pods an EKS node can run with the VPC CNI, set by its instance type's network interfaces and addresses per interface (17 for a t4g.medium). Lesson 35.26
MinHealthyPercentage
The share of an Auto Scaling group's capacity that must stay in service during an instance refresh; it decides the batch size. Lesson 35.24
NAT gateway (NAT gateways)
A managed AWS service in a public subnet that lets instances in private subnets start connections to the internet while nothing on the internet can connect to them. It needs an Elastic IP and lives in one Availability Zone; it is billed per hour and per GB. Lesson 35.7
network ACL (NACL, NACLs, network ACLs)
A stateless firewall on a subnet: numbered allow and deny rules, checked lowest number first, both directions separately. Replies need their own rules. Lesson 35.4
Network Load Balancer (NLB (AWS), NLBs)
The AWS TCP/UDP load balancer: static addresses per zone, the client's IP preserved, very high connection rates, no HTTP features. Lesson 35.19
node role fallback (node role)
What a pod without its own AWS identity gets: the node's role through the metadata service, shared by every pod on that node (unless the hop limit is 1). Lesson 35.29
OIDC issuer (EKS) (OIDC provider (EKS), cluster OIDC issuer)
The URL that signs an EKS cluster's service account tokens; registered in IAM as an identity provider, it lets IRSA roles trust the cluster's tokens. Lesson 35.29
prefix list (prefix lists)
A named set of address ranges, such as all S3 addresses in a Region, that route tables and security group rules can refer to as one entry (pl-...). Lesson 35.7
private DNS (endpoints) (private DNS)
An interface endpoint option that makes the service's normal name (ssm.eu-central-1.amazonaws.com) resolve to the endpoint's private addresses inside the VPC. Lesson 35.7
private subnet (private subnets)
A subnet with no route to an internet gateway. Its instances cannot be reached from the internet; they reach out through a NAT gateway or VPC endpoints, or not at all. Lesson 35.1
public subnet (public subnets)
A subnet whose route table sends 0.0.0.0/0 to an internet gateway. Nothing else makes a subnet public; instances there also need a public IP to be reachable. Lesson 35.1
Reachability Analyzer
An AWS tool that works out from the configuration alone whether one resource can reach another on a port, and which rule blocks it. Lesson 35.4
reserved addresses (AWS) (five reserved addresses)
The five addresses AWS takes in every subnet: the network address, .1 for the router, .2 for DNS, .3 for later use, and the broadcast address. Lesson 35.1
route table (AWS) (route tables (AWS), AWS route table)
The list of rules a subnet uses to decide where traffic for each destination goes: local inside the VPC, an internet gateway, a NAT gateway, an endpoint. The most specific (longest prefix) route wins. Each subnet uses exactly one route table. Lesson 35.1
S3 backend (backend "s3")
Terraform state kept as an object in an S3 bucket, shared by everyone who works on the configuration. Lesson 35.12
scaling activity (scaling activities)
One action of an Auto Scaling group (launching or terminating an instance) with its status, description and cause. Lesson 35.24
security group (AWS) (security groups (AWS), SG, SGs)
A stateful firewall attached to network interfaces: a list of allowed inbound and outbound traffic. Replies to an allowed connection always pass; there are no deny rules. A rule's source can be another security group: "from whatever carries the load balancer's group". Lesson 35.4
Session Manager (SSM Session Manager, aws ssm start-session)
The AWS Systems Manager feature that gives a shell on an instance through the SSM agent, with no open port and no SSH key, controlled by IAM and logged. Lesson 35.17
Session Manager plugin (session-manager-plugin)
The program on your own machine that aws ssm start-session hands the connection to. Installed separately, from AWS's package. Lesson 35.17
SSM agent (amazon-ssm-agent)
The program on an instance that connects out to Systems Manager so it can be managed and reached with Session Manager. It uses the instance role. Lesson 35.17
SSRF (server-side request forgery)
An attack that tricks a server into making a request for the attacker, for example to the metadata service, and returning the answer. Lesson 35.17
stateful firewall (stateful)
A firewall that remembers connections, so the reply to an allowed request is allowed without a rule of its own. AWS security groups are stateful. Lesson 35.4
stateless firewall (stateless)
A firewall that judges every packet on its own, with no memory of connections, so replies need explicit rules. AWS network ACLs are stateless. Lesson 35.4
subnet (AWS) (AWS subnet, subnets (AWS))
A slice of a VPC's address range that lives in exactly one Availability Zone. Instances, load balancer nodes and EKS pods get their addresses from a subnet. AWS keeps five addresses in every subnet: a /24 has 251 usable. Lesson 35.1
subnet discovery tags (kubernetes.io/role/elb, kubernetes.io/role/internal-elb)
Subnet tags the Load Balancer Controller uses to choose subnets: kubernetes.io/role/elb for internet-facing, internal-elb for internal load balancers. Lesson 35.33
target group (target groups)
A set of targets (instances, IPs or Lambda functions) a load balancer sends traffic to, with its own port, protocol and health check. Lesson 35.19
target tracking (target tracking policy)
An Auto Scaling policy that changes desired capacity to keep a metric, such as average CPU, near a target value. Lesson 35.24
target type (ELB) (target-type ip, target-type instance, IP targets)
How a target group addresses its targets: instance (by instance ID, on a NodePort for Kubernetes) or ip (by address, such as pod IPs). Lesson 35.33
update-kubeconfig (aws eks update-kubeconfig)
The AWS CLI command that writes a kubeconfig context for an EKS cluster, with an exec plugin that gets a token from your AWS credentials. Lesson 35.26
use_lockfile (.tflock, tflock, S3 native locking)
The S3 backend option (Terraform 1.10+) that locks the state with a .tflock object written by a conditional PUT; a second run gets PreconditionFailed. Lesson 35.12
user data
A script or cloud-config document passed to an instance at launch; cloud-init runs it once, as root, at the first boot. Lesson 35.14
VPC (AWS) (VPCs, virtual private cloud)
Your own private network in one AWS Region: an address range (a CIDR such as 10.20.0.0/16) that you cut into subnets, with route tables, gateways and firewalls you control. It spans the Region's Availability Zones, but each of its subnets lives in one zone. The Azure equivalent is a VNet. Lesson 35.1
VPC CNI (aws-node, amazon-vpc-cni)
The EKS networking plugin that gives every pod an IP address from the VPC's subnets, so pods are real members of the network. Lesson 35.26
VPC endpoint (VPC endpoints, endpoint (VPC))
A way for resources in a VPC to reach an AWS service without the internet: a gateway endpoint (a route) or an interface endpoint (a network interface). Lesson 35.7
VPC flow logs (flow log, flow logs)
Records of the IP traffic of a VPC, subnet or network interface: source, destination, ports, protocol, packets, bytes and whether it was accepted or rejected. Delivered to CloudWatch Logs or S3 (gzip), aggregated over 60 or 600 seconds. They never contain payloads. Lesson 35.4