Operations

A cloud operations team receives thousands of alerts daily from their monitoring platform. On-call engineers report that most alerts are transient CPU spikes that self-resolve within a minute, and they have started ignoring notifications, which caused a genuine outage to go unnoticed for 20 minutes. Which action should the team take FIRST to reduce alert fatigue while preserving detection of real incidents?

A cloud operations team monitors a customer-facing API whose traffic varies significantly by time of day and day of week. Static threshold alerts on request latency generate misses during low-traffic periods (real spikes go unnoticed) and false alarms during known peak hours. Which monitoring approach best improves alert accuracy across these varying traffic patterns?

A cloud operations team must retain daily backups for 7 days, weekly backups for 5 weeks, and monthly backups for 12 months while minimizing storage cost. Backups are currently created manually, which has led to inconsistent retention and growing storage bills. Which approach best automates this requirement?

A financial services company runs nightly automated backups of its production database to a separate cloud region. During a quarterly audit, the compliance team asks for evidence that the backups can actually be used to recover the system within the documented RTO. The operations team confirms backup jobs complete successfully every night with no errors. What is the MOST important additional action the operations team should take to satisfy the auditor and reduce recovery risk?

A company performs a full backup of a 5 TB file server every night to cloud object storage, retaining 30 days of history. The daily data change rate is only about 2%. The cloud storage bill for backups has grown to an unsustainable level, and the backup window is beginning to exceed the overnight maintenance period. Which change will MOST effectively reduce both storage consumption and backup duration while preserving the 30-day recovery capability?

A cloud operations team monitors a fleet of Linux virtual machines running a memory-intensive analytics application. The provider's default monitoring dashboard shows CPU, disk, and network metrics, but the team keeps missing out-of-memory crashes because memory usage is not visible. What is the MOST appropriate way to gain visibility into memory consumption?

A cloud operations team receives the monthly bill and discovers that spend increased 40% over the prior month. Most of the increase comes from resources that carry no cost-allocation tags, so finance cannot attribute the charges to any team or project. The team wants to detect and investigate cost spikes as they happen rather than at month-end. Which action provides the FASTEST ongoing visibility into unexpected cost increases?

A cloud operations engineer configures a monitoring dashboard for a customer-facing API. The dashboard displays CPU utilization aggregated as a 5-minute average and shows steady values around 45%. However, users report intermittent timeouts, and support tickets indicate slow responses during brief periods. The underlying compute instances autoscale on CPU, but scaling rarely triggers. What is the MOST likely reason the dashboard fails to reveal the problem, and what should the engineer change?

A microservices application spread across dozens of containers is experiencing intermittent latency spikes. The operations team can see individual service metrics (CPU, memory, response time) in their monitoring dashboard, but they cannot determine which service in a request chain is responsible for the delay because each request passes through many services. Which observability capability should the team implement to identify the specific service causing the latency in a given request?

A cloud operations team receives complaints that a payment API intermittently returns HTTP 500 errors, but the existing CPU and memory alarms never trigger during these events. Application logs are already being shipped to the cloud provider's centralized logging service. The team wants automated notification the moment the 500 error rate crosses a defined threshold, without deploying additional agents or code changes. Which approach best meets this requirement?

A financial services company aggregates application logs into a centralized logging service. Compliance requires that logs be retained and searchable for 90 days, but auditors may request access to logs up to 7 years old on rare occasions. The team notices their monthly logging costs have tripled because all logs remain in the high-performance searchable tier indefinitely. Which approach best balances compliance, retrieval needs, and cost?

A cloud operations team notices their monthly log management bill has tripled over the past quarter. Investigation reveals that developers enabled DEBUG-level logging across all production application tiers to troubleshoot an issue three months ago, and the setting was never reverted. The logs are ingested into a centralized SaaS logging platform that charges per GB ingested and per GB stored. Compliance requires only that transaction audit events and error-level logs be retained for one year. Which action will most effectively reduce cost while preserving required data?

A cloud operations team is investigating intermittent latency complaints on a payment API. Their monitoring platform collects CPU and memory metrics at a 5-minute interval, but the reported slowdowns seem to last only 30 to 60 seconds. Dashboards show all resources well within normal ranges during the incident windows. What is the MOST effective change to help the team capture the cause of these transient events?

A cloud operations team monitors a customer-facing API. The dashboard shows the average (mean) response time is a healthy 120 ms, and the team receives no alerts. However, the support desk is flooded with complaints from users experiencing multi-second delays. Which monitoring change would BEST reveal the true impact on affected users?

A cloud operations team is asked by finance to break down monthly spending by department so each business unit can be charged for the resources it consumes. Currently all resources run in a single shared account and the bill arrives as one lump sum with no breakdown. Which approach should the team implement to enable accurate per-department cost attribution?

A cloud operations engineer reviews a month of monitoring data for a production application tier consisting of 12 general-purpose VMs. The dashboard shows average CPU utilization at 9%, peak CPU at 22%, and memory utilization consistently under 30%. The application meets all SLAs and has no seasonal spikes. Finance has asked the team to reduce spend without impacting performance. Which action should the engineer take FIRST?

A financial services company runs a transactional database in the cloud. Business stakeholders state that in a disaster the company can tolerate losing at most 15 minutes of committed transactions, and the application must be fully operational again within 2 hours. The current backup strategy takes a full snapshot once every 24 hours. Which change BEST aligns the backup and recovery approach with the stated business requirements?

A cloud operations team responds to the same disk-space-full alert on production web servers multiple times per week. Each time, an engineer manually logs in, clears temporary files, and archives old logs to object storage. Leadership wants to reduce manual toil and shorten mean time to resolution (MTTR) for this recurring, well-understood incident. Which approach BEST addresses this requirement?

A cloud operations team is reviewing monthly spend and notices that a fleet of development and QA virtual machines runs 24/7, even though engineers only use them during business hours (roughly 8 AM to 6 PM, Monday through Friday). The workloads are stateless and can tolerate being unavailable outside working hours. Management wants to reduce cost without deleting the environments or changing instance types. Which approach best achieves this goal?

A cloud operations team relies on real-user monitoring (RUM) to detect issues with a customer-facing checkout API. However, during low-traffic overnight hours, an outage went undetected for nearly two hours because no users were transacting, and the team only learned of it from morning complaints. Which monitoring approach should the team add to detect availability and performance problems regardless of actual user traffic?