OnCallReadyStart learning

AWS III: CloudWatch, CloudTrail, Cost & Incidents: glossary

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

alarm action (alarm actions, --alarm-actions, --ok-actions)
What an alarm does when its state changes, once per change: usually notify an SNS topic, sometimes scale or stop an instance. Lesson 36.4
alarm history (describe-alarm-history, StateReasonData)
The record of every state change, configuration update and action an alarm ran, with the datapoints that decided it. Lesson 36.4
alarm state (ALARM, INSUFFICIENT_DATA, alarm states)
Where a CloudWatch alarm stands: OK (within the threshold), ALARM (breaching) or INSUFFICIENT_DATA (not enough data to decide, also the state of a new alarm). Lesson 36.4
AWS Budgets (Budgets, aws budgets, create-budget)
The AWS service that tracks spending (or usage) for a period against a limit, optionally filtered by tag or service, and emails or notifies a topic when a threshold is crossed. Lesson 36.22
AWS Health Dashboard (AWS Health, Health API, Personal Health Dashboard)
Where AWS reports problems with its services that affect your account. The Health API needs a Business support plan or higher. Lesson 36.32
aws logs tail (logs tail, --follow)
The AWS CLI command that prints a log group's newest events from all its streams and, with --follow, keeps printing new ones - tail -f for a fleet. Lesson 36.8
aws:RequestedRegion (RequestedRegion)
The condition key with the Region a request is sent to - the key Region guardrails test. Global services are called in us-east-1. Lesson 36.26
basic monitoring (detailed monitoring)
EC2 publishes its metrics every five minutes for free (basic monitoring); detailed monitoring, paid, publishes them every minute. Lesson 36.1
burst credits (burst balance, BurstBalance)
The allowance that lets a small gp2 volume run faster than its baseline for a while. When it runs out under sustained load, the disk drops to its slow baseline. Lesson 36.29
bytes scanned (bytesScanned, recordsScanned)
How much log data a Logs Insights query read - everything in its time window of the groups it names - and what it is billed on. Lesson 36.11
CloudWatch (Amazon CloudWatch, CloudWatch metrics)
The AWS monitoring service: every service publishes its numbers (metrics) into it, and it keeps them, draws them, and raises alarms on them. aws cloudwatch list-metrics --namespace AWS/ApplicationELB lists what the load balancers publish. Lesson 36.1
CloudWatch agent (CWAgent, amazon-cloudwatch-agent)
A program you install on an instance that publishes what EC2 cannot see from outside - memory and disk usage - into the CWAgent namespace, and ships log files to CloudWatch Logs. Lesson 36.1
CloudWatch alarm (CloudWatch alarms, metric alarm, put-metric-alarm)
A rule that watches one metric (or a formula) against a threshold over a number of periods and changes state when it is crossed, running actions such as notifying an SNS topic. Lesson 36.4
CloudWatch Logs (Amazon CloudWatch Logs)
The AWS service that stores log lines from programs, AWS services and agents, sorted into log groups, searchable and kept as long as the retention says. Lesson 36.8
composite alarm (composite alarms)
A CloudWatch alarm whose rule combines other alarms (A AND B, A OR NOT C), to page once for a pattern instead of once per symptom. Lesson 36.4
cost allocation tag (cost allocation tags, update-cost-allocation-tags-status, user:team)
A tag key switched on in Billing so costs can be grouped by it. It only groups costs from the activation on, unless you backfill; Cost Explorer writes user-defined ones as user:key. Lesson 36.22
Cost Anomaly Detection (anomaly monitor, anomaly monitors, cost anomaly)
The AWS Billing feature that learns each service's or team's normal daily spend and alerts when a day is unusual, without a fixed threshold. Lesson 36.22
Cost Explorer (AWS Cost Explorer, get-cost-and-usage, aws ce)
The AWS tool that breaks the bill down by service, usage type, Region, tag or day, from the console or the ce API (each API request costs one cent). Lesson 36.18
custom metric (custom metrics (CloudWatch), put-metric-data)
A metric your own code or script publishes into CloudWatch with put-metric-data, in a namespace of your own. Each distinct name and dimension set is billed as one metric. Lesson 36.1
DAILY (MONTHLY, --granularity)
The granularity of a Cost Explorer query: one row per day or per month (HOURLY needs opting in). Lesson 36.18
data transfer (AWS) (data transfer, cross-AZ traffic, cross-AZ)
What AWS charges for moving data: into AWS is free, out to the internet and between Regions costs per GB, and even between zones of one Region costs per GB in each direction. Lesson 36.18
datapoint (datapoints, Datapoints)
One summarised value of a metric for one period, for example the Sum of requests between 09:00 and 09:01. get-metric-statistics returns them unordered; sort with --query 'sort_by(Datapoints, &Timestamp)'. Lesson 36.1
delegated administrator (delegated admin, list-delegated-administrators)
A member account allowed to run an Organizations feature (such as policy management) instead of the management account, so the management account can stay almost unused. Lesson 36.26
dimension (CloudWatch) (dimensions (CloudWatch), CloudWatch dimension, --dimensions)
A name=value label on a CloudWatch metric, such as LoadBalancer=app/shop-alb/... A query must give exactly the dimensions the metric was published with, or it finds nothing. Lesson 36.1
discovered field (discovered fields, @timestamp, @message)
A field Logs Insights finds by itself: @timestamp, @message and @logStream on every event, plus every key of JSON logs, CloudTrail records and VPC flow logs. Lesson 36.11
EBS snapshot (EBS snapshots, create-snapshot)
A point-in-time copy of an EBS volume stored by AWS, incremental after the first; the safety net before a risky change and the base of EBS backups. Lesson 36.29
EC2 - Other (EC2-Other)
The Cost Explorer service line that is not instances: EBS volumes and snapshots, NAT gateways, Elastic IPs and data transfer. Lesson 36.18
effective policy (describe-effective-policy, effective policies)
The result of all the tag (or other management) policies inherited along an account's path, merged - what actually applies to that account. Lesson 36.26
EFS (Amazon EFS, Elastic File System)
Amazon Elastic File System: a managed NFS file share many instances in several zones can mount at once. Lesson 36.32
ELB (Elastic Load Balancing, LCU, load balancer capacity unit)
Elastic Load Balancing, the AWS load balancer family (ALB, NLB). Its metrics are in AWS/ApplicationELB; its bill is hours plus LCUs, units of traffic and connections. Lesson 36.1
EstimatedCharges (AWS/Billing, billing alarm, billing alerts)
The CloudWatch metric (AWS/Billing, us-east-1 only) with the month's running total, published every few hours once billing alerts are enabled. An alarm on it is the simplest billing alert. Lesson 36.18
eventSource (eventName, eventTime, sourceIPAddress)
Fields of every CloudTrail record: the service endpoint called (iam.amazonaws.com), the API action, when, from which IP, and the error if the call failed. Lesson 36.14
extended statistic (extended statistics, --extended-statistics)
A percentile such as p99 or p95 asked from CloudWatch, the slowest values that an average hides. Lesson 36.1
Fault Injection Service (AWS FIS, FIS)
The AWS service that runs planned failure experiments - stopping instances, adding latency - with stop conditions tied to CloudWatch alarms. Lesson 36.32
FILL (FILL())
The metric math function that replaces missing datapoints with a value, FILL(m1, 0), so a sparse error count becomes a continuous series. Lesson 36.1
filter pattern (filter patterns, --filter-pattern)
The small search language of CloudWatch Logs used by filter-log-events, logs tail and metric filters: words, "exact phrases", ?OR, -exclude, JSON selectors like { $.status >= 500 }. Lesson 36.8
forecasted alert (FORECASTED, ACTUAL, actual alert)
A budget alert type: ACTUAL fires when spend so far crosses the threshold, FORECASTED when AWS projects the period will end above it - earlier, while there is time to act. Lesson 36.22
FullAWSAccess (p-FullAWSAccess)
The AWS-managed SCP that allows everything, attached by default to every root, OU and account. Every level of the path needs an allow, so detaching it denies all below. Lesson 36.26
gateway endpoint (gateway endpoints, S3 gateway endpoint, interface endpoint)
A free VPC endpoint for S3 or DynamoDB that adds a route to the service, so private subnets reach it without the NAT gateway. Other services need interface endpoints, which are billed. Lesson 36.18
gp3 (gp2, gp3 volume, gp2 volume)
The general-purpose SSD types of EBS: gp3 sets size, IOPS and throughput independently (3,000 IOPS and 125 MB/s included); gp2 ties performance to size and lets small volumes burst on credits. Lesson 36.29
growpart (growpart /dev/nvme0n1 1)
The command (from cloud-guest-utils) that grows a partition to fill a bigger disk. It takes the disk and the partition number as two arguments. Lesson 36.29
high-resolution metric (high-resolution metrics, --storage-resolution)
A custom metric kept per second instead of per minute, for alarms that must react within seconds. It costs more. Lesson 36.1
ingestion (logs) (log ingestion, ingested bytes, IncomingBytes)
Sending log data into CloudWatch Logs, billed per GB - often the largest part of a logging bill, and the reason a debug log level in production is expensive. Lesson 36.8
IOPS (I/O operations per second)
Input/output operations per second: how many separate reads and writes a disk can do each second. Lesson 36.29
log group (log groups, /shop/checkout)
A folder of logs in CloudWatch Logs, usually one per service, holding its log streams and its settings: retention, metric filters, encryption. Lesson 36.8
log stream (CloudWatch) (CloudWatch log stream, log streams (CloudWatch))
One sequence of log events inside a log group, usually one per instance, container or Lambda run. Lesson 36.8
Logs Insights (CloudWatch Logs Insights, Insights query, Insights queries)
The query language and engine of CloudWatch Logs: commands joined with pipes (fields, filter, stats, sort, limit, parse) over one or more log groups, returning a table. Lesson 36.11
lookup-events (aws cloudtrail lookup-events, --lookup-attributes)
The CloudTrail call that searches the last 90 days of management events in one Region, by event name, user name, resource or event source. Lesson 36.14
M out of N (datapoints to alarm, --datapoints-to-alarm, evaluation periods)
An alarm setting: go to ALARM when at least M of the last N periods breach, so one noisy minute does not page anyone. Lesson 36.4
metric filter (metric filters, put-metric-filter)
A rule on a log group that turns matching log events into a CloudWatch metric, so you can alarm on a log message. It only counts events that arrive after it exists. Lesson 36.8
metric math (metric math expression, get-metric-data)
Formulas over CloudWatch metrics, such as 100 * errors / requests, computed by get-metric-data or an alarm. FILL replaces missing datapoints, for example with 0. Lesson 36.1
modify-volume (aws ec2 modify-volume, Elastic Volumes, optimizing)
The EC2 call that changes an attached EBS volume's size, type, IOPS or throughput online. It goes modifying -> optimizing -> completed, and the next change must wait six hours. Lesson 36.29
Multi-AZ (multi-AZ, Multi-AZ deployment)
Running a workload or database in more than one Availability Zone, with automatic failover, so the loss of one zone does not take it down. Lesson 36.32
namespace (CloudWatch) (CloudWatch namespace, AWS/ApplicationELB, AWS/EC2)
The folder a CloudWatch metric lives in, one per service (AWS/EC2, AWS/ApplicationELB) plus your own (Shop/Checkout). Names starting with AWS/ are reserved for AWS. Lesson 36.1
NAT gateway (AWS) (AWS NAT gateway, NAT gateway processing)
The managed AWS box that lets private subnets reach the internet. It is billed per hour and per GB processed, so heavy traffic through it - often to S3 - is a classic bill surprise. Lesson 36.18
NotAction (NotAction (policy))
The policy element meaning "every action except these". A Region guardrail uses it to deny everything outside approved Regions except the global services. Lesson 36.26
NVMe device name (nvme0n1, /dev/nvme0n1, nvme0n1p1)
How EBS volumes appear on current EC2 instances: /dev/nvme0n1 for the disk and nvme0n1p1 for its first partition, whatever device name the console shows. Lesson 36.29
parse (Logs Insights) (parse command)
The Logs Insights command that pulls fields out of plain-text log lines, with a glob pattern (* for each field) or a regular expression with named groups. Lesson 36.11
PendingConfirmation (pending confirmation, confirm-subscription)
The state of a new SNS email subscription until someone clicks the confirmation link. SNS delivers nothing to it meanwhile. Lesson 36.4
period (CloudWatch) (--period, CloudWatch period)
The length in seconds of each stretch of time CloudWatch summarises into one datapoint: 60 for one-minute data, 300 for five minutes. Lesson 36.1
pilot light (warm standby, backup and restore, multi-site active/active)
The four disaster-recovery strategies, from cheapest and slowest to dearest and fastest: backup and restore, pilot light (data copied, servers off), warm standby (a small copy running), multi-site active/active. Lesson 36.32
readOnly (CloudTrail) (readOnly, write events, read events)
The CloudTrail field that says whether a call only read (Describe, List, Get) or changed something. During an incident you usually want the write events. Lesson 36.14
resize2fs (xfs_growfs)
The command that grows an ext4 filesystem to fill its partition, online; XFS uses xfs_growfs on the mount point instead. Lesson 36.29
Resource Groups Tagging API (resourcegroupstaggingapi, get-resources, tag-resources)
The AWS API that lists and changes tags across services in one call - the way to find every untagged resource in a Region. Lesson 36.22
retention (CloudWatch Logs) (retention policy, put-retention-policy, Never expire)
How long a log group keeps its events. A new group keeps them forever ("Never expire") until you set one. Lesson 36.8
role session name (session name, RoleSessionName, --role-session-name)
The free-text name given when a role is assumed. It appears in the session's ARN and in CloudTrail, so it is how you tell which person or pipeline run used a shared role. Lesson 36.14
RPO (recovery point objective)
Recovery point objective: how much data you may lose, measured as the time since the last good copy. Lesson 36.32
RTO (recovery time objective)
Recovery time objective: how long a system may be down after a failure before it must be back. Lesson 36.32
set-alarm-state (aws cloudwatch set-alarm-state)
The CloudWatch call that puts an alarm into a state by hand, so its real actions run - the standard way to test that an alarm pages. The next evaluation puts the true state back. Lesson 36.4
SIEM (security information and event management)
Security information and event management: a system the security team uses to collect logs from everywhere and search them for attacks. Lesson 36.8
SNS topic (SNS topics, topic (SNS), TopicArn)
A named mailbox in Amazon SNS: whatever is published to it is delivered to every subscriber - email, SMS, an HTTPS endpoint, a queue. Lesson 36.4
sparse metric (sparse metrics)
A metric that is only published when something happens, such as a count of 5xx errors that has no datapoint in a minute without errors. Lesson 36.1
start-query (get-query-results, queryId, query ID)
The two CLI calls behind a Logs Insights query: start-query returns a query ID, get-query-results returns the rows once its status is Complete. Lesson 36.11
static stability (statically stable)
Running with enough capacity already in place that losing a zone needs no new launches, because during a big outage the service that launches replacements is busy too. Lesson 36.32
statistic (CloudWatch) (CloudWatch statistic, SampleCount, --statistics)
How CloudWatch summarises the samples in each period: Sum, Average, Minimum, Maximum or SampleCount (how many samples there were). Lesson 36.1
stats (Logs Insights) (stats ... by, stats count(*))
The Logs Insights command that aggregates - count(*), avg(), pct(), sum() - optionally per group with by, like KQL's summarize. Lesson 36.11
StopLogging (stop-logging, DeleteTrail, get-trail-status)
The CloudTrail call that switches a trail off. Usually an attacker's first move, so it should page someone; get-trail-status shows IsLogging and when it stopped. Lesson 36.14
subscription filter (subscription filters)
A rule that streams a log group's matching events to another service as they arrive - a Lambda function, a Kinesis stream or a SIEM. Lesson 36.8
tag backfill (start-cost-allocation-tag-backfill, backfill (cost allocation))
Asking Billing to re-process up to twelve months of cost data with your activated tags, so old costs can be grouped by them too. Lesson 36.22
tag policy (tag policies, TAG_POLICY)
An Organizations policy that defines the allowed spelling and values of tags and reports resources that break them. A resource without the tag still counts as compliant. Lesson 36.22
treat missing data (TreatMissingData, --treat-missing-data, notBreaching)
What an alarm assumes for a period with no datapoint: missing (the default), notBreaching (fine), breaching (bad) or ignore. Sparse metrics need notBreaching. Lesson 36.4
unblended cost (UnblendedCost, amortized cost, AmortizedCost)
Cost Explorer metrics: unblended = what each usage was charged when it happened; amortized spreads upfront payments over the period they cover. Lesson 36.18
usage type (USAGE_TYPE, usage types, NatGateway-Bytes)
The billing name of exactly what was used, with a Region prefix: EUC1-NatGateway-Bytes, EUC1-DataTransfer-Regional-Bytes, EUC1-BoxUsage:t3.medium. Lesson 36.18
USD
US dollars, the currency of AWS prices and bills (Currency=USD in the billing metric). Lesson 36.18
userAgent (CloudTrail) (userAgent, user agent (CloudTrail))
The field of a CloudTrail record that names the client that made the call: the console, aws-cli/2.38.1, a Terraform provider, an SDK. Lesson 36.14
userIdentity (userIdentity.arn, sessionContext, sessionIssuer)
The part of a CloudTrail record that says who made the call: the principal's ARN, its type, and for a role session the role that issued it. Lesson 36.14
Well-Architected Framework (AWS Well-Architected, Well-Architected, pillar)
AWS's guide to good designs in six pillars: operational excellence, security, reliability, performance efficiency, cost optimization and sustainability. Lesson 36.32