IBM Cloud Associate SRE flashcards
IBM Cloud Associate SRE · Build recall with a focused review session.
32 cards. Read the term and the definition here. The deck at the top of this page is optional.
- Error budget
- The allowed amount of unreliability, computed as 1 minus the SLO over the same window. Exhausting it triggers the agreed release freeze or reliability investment.
- Service level indicator
- A quantitative measure of service behavior, such as the proportion of requests answered under 300 milliseconds, collected continuously and expressed as a good-to-valid ratio.
- Toil
- Manual, repetitive, automatable, tactical work that produces no enduring value and grows linearly with the service. SRE practice caps it near half of engineering time.
- MTTR
- Mean time to repair: total downtime divided by the number of incidents. Availability equals MTBF divided by MTBF plus MTTR, so shortening recovery raises reliability.
- Four golden signals
- Latency, traffic, errors, and saturation. Together they form a compact monitoring set for request-driven services and indicate what deserves a paging alert.
- Burn rate alerting
- Paging when the error budget is consumed faster than the SLO permits. Multi-window burn rates catch both fast outages and slow degradation while keeping false pages low.
- High cardinality labels
- Metric dimensions with effectively unbounded unique values, such as user or request identifiers. They multiply time series and cost, so they belong in logs or traces.
- IBM Cloud Monitoring
- A Sysdig-based observability service collecting platform and application metrics for IBM Cloud resources; Kubernetes clusters need its agent deployed as a DaemonSet.
- Incident commander
- The person who coordinates the response, owns decisions and the communication cadence, and deliberately does not debug. Separating coordination from hands-on work prevents chaos.
- Blameless postmortem
- An incident review that identifies systemic and process causes rather than individuals, assuming people acted reasonably given the information they had at the time.
- Severity levels
- Predefined incident categories that determine who gets paged, escalation paths, how often stakeholders receive updates, and whether a postmortem is mandatory.
- Postmortem action items
- Follow-up tasks from an incident review that are specific, owned by one named person, and dated. Their completion rate is tracked as a reliability metric.
- Readiness probe
- A Kubernetes check deciding whether a pod may receive traffic. Failing it removes the pod from service endpoints without restarting the container.
- Canary deployment
- Routing a small percentage of live traffic to a new version and comparing its error rate and latency against the baseline before widening the rollout.
- Immutable image tags
- Deploying containers by digest or unique version tag instead of a floating latest tag, so rollback and audit point at exactly the artifact that ran.
- GitOps
- Managing cluster and infrastructure state as files in Git, with a controller continuously reconciling the live environment to the committed desired state and recording drift.
- Resource group
- An IBM Cloud account-level container used to organize resources for access control and billing. Resource groups cannot be nested inside one another.
- Multizone region
- An IBM Cloud region containing three separate availability zones. Deploying workloads across all zones lets them survive the complete loss of one zone.
- IBM Cloud Code Engine
- A serverless platform for containers and batch jobs that scales to zero, removing the need to provision and patch Kubernetes worker nodes for suitable workloads.
- Object Storage classes
- Standard, Vault, and Cold Vault tiers that trade retrieval latency against storage cost; resiliency options are cross-region, regional, and single-site buckets.
- Security groups versus network ACLs
- Security groups are stateful and attach to instances; network ACLs are stateless and attach to subnets, are evaluated in order, and need explicit rules in both directions.
- IBM Cloud Internet Services
- A Cloudflare-based edge service providing global DNS, CDN, web application firewall, DDoS protection, and TLS termination in front of IBM Cloud or external origins.
- Transit Gateway
- A private routing hub connecting VPCs, classic infrastructure, and Direct Link attachments across accounts and regions without exposing traffic to the public internet.
- VPC public gateway
- A per-zone egress device letting subnet instances reach the internet. Inbound internet traffic instead requires a floating IP or a load balancer.
- Service ID
- A non-human IBM Cloud identity with its own API keys and IAM policies, used by applications and pipelines instead of embedding a person's credentials.
- Access group
- A collection of identities to which IAM policies are attached once, so adding or removing members never requires rewriting every individual policy.
- Activity Tracker Event Routing
- An IBM Cloud audit service recording account-level API, CLI, and console events with outcomes, used to prove who performed which action and when.
- Envelope encryption
- Encrypting data with a data encryption key that is itself wrapped by a root key held in Key Protect or Hyper Protect Crypto Services.
- Requests and limits
- Requests are the CPU and memory the scheduler reserves for a pod; limits are the ceiling it cannot exceed, causing throttling or OOM termination beyond that point.
- Horizontal Pod Autoscaler
- A Kubernetes controller that adds or removes pod replicas based on observed CPU, memory, or custom metrics compared with configured target values.
- OOMKilled
- The container termination reason when a process exceeds its memory limit. Raising the limit or fixing the leak is required; restarting alone will not resolve it.
- ImagePullBackOff
- A pod status meaning the kubelet cannot fetch the container image, usually caused by a wrong image reference or a missing registry pull secret.