Guidance for the security-side use of Azure Monitor and Log Analytics — designing the workspace strategy that feeds Microsoft Sentinel and Defender for Cloud, choosing analytics vs basic vs auxiliary log tiers, retention and archive, table-level transformations to drop noise pre-ingestion, Data Collection Rules (DCRs) and Azure Monitor Agent (AMA), workspace topology (single vs regional vs sovereign), commitment tiers and cost control, customer-managed keys for the workspace, RBAC including table-level RBAC, integration with Sentinel (workspace requirements), and decommissioning legacy MMA. WHEN: Log Analytics workspace design, Azure Monitor security data, Sentinel workspace, log tier basic auxiliary, AMA Azure Monitor Agent, DCR data collection rule, drop columns ingestion transform, log retention archive, workspace CMK, table-level RBAC, log ingestion cost control. DO NOT USE for Sentinel detection authoring (use sentinel-detection-engineering), App Insights instrumentation (use appinsights-instrumentation)
설치
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Guidance for the security-side use of Azure Monitor and Log Analytics — designing the workspace strategy that feeds Microsoft Sentinel and Defender for Cloud, choosing analytics vs basic vs auxiliary log tiers, retention and archive, table-level transformations to drop noise pre-ingestion, Data Collection Rules (DCRs) and Azure Monitor Agent (AMA), workspace topology (single vs regional vs sovereign), commitment tiers and cost control, customer-managed keys for the workspace, RBAC including table-level RBAC, integration with Sentinel (workspace requirements), and decommissioning legacy MMA. WHEN: Log Analytics workspace design, Azure Monitor security data, Sentinel workspace, log tier basic auxiliary, AMA Azure Monitor Agent, DCR data collection rule, drop columns ingestion transform, log retention archive, workspace CMK, table-level RBAC, log ingestion cost control. DO NOT USE for Sentinel detection authoring (use sentinel-detection-engineering), App Insights instrumentation (use appinsights-instrumentation), or non-security observability.
license
MIT
metadata
{"author":"Microsoft","version":"0.1.0"}
Azure Monitor & Log Analytics for Security
Microsoft Sentinel runs on a Log Analytics workspace. Defender for Cloud, Defender XDR
hunting tables, Activity logs, and Diagnostic settings also flow there. The workspace
strategy decides whether your security telemetry is complete, compliant, and
affordable — or two of three.
This skill covers the security-side decisions in workspace design: tiering, retention,
DCRs, RBAC, and cost.
When to use
Designing or refactoring the Log Analytics workspace(s) that back Sentinel and the
broader security telemetry estate.
Do not use this skill for Sentinel detection authoring
(sentinel-detection-engineering), App Insights instrumentation
(appinsights-instrumentation), or non-security observability.
Log table tiers — pick per table
Tier
Query latency
Retention max
Use case
Analytics
Sub-second
730 days interactive
Detections, hunting, dashboards
Basic
Seconds (KQL with limits)
30 days interactive + archive
High-volume "search if needed" logs
Auxiliary
Slower / async
12 months interactive
Low-touch compliance archive feeds
Archive (any tier)
Search-job recall
up to 12 years
Long-tail evidence, regulatory hold
Microsoft has been re-naming/re-shaping these — verify current naming and pricing in your
region before commit.
Approach
Workspace topology.
Default: one Sentinel workspace per tenant for security signal — keeps
correlation easy.
Regional / sovereign: separate workspaces where data residency or sovereignty
requires.
Multi-workspace Sentinel: only when business reasons (M&A boundary, MSSP
model) force it. Cross-workspace queries work but add complexity.
Don't split per-team; that fragments correlation.
Use Azure Monitor Agent (AMA) for everything new. The Microsoft Monitoring Agent
(MMA / Log Analytics agent) is retired. Migration paths via Defender for Cloud
auto-provisioning and Sentinel solutions.
Define DCRs per data class.
One DCR per data source family (Windows security events, Linux Syslog, custom
logs).
Use transformKql to drop noisy columns or whole event IDs at ingest. A 30%
pre-ingest reduction beats any post-ingest tuning.
Validate transformations against detection KQL — don't drop a column a Sentinel
analytic relies on.
Basic: AzureDiagnostics for chatty resources, NSG flow logs, web server access
logs (where you need them only for incident search).
Auxiliary: long-retention compliance feeds (e.g., raw firewall logs for 12
months).
Tier choice changes ingestion + retention costs significantly.
Retention and archive.
Interactive retention: 90 days default — extend to align with detection lookback
windows (often 180–730 days).
Archive (cold) for compliance and IR replay (12-year max for some scenarios).
Search jobs rehydrate archived data when needed.
Cost control.
Commitment tier at the workspace (100 GB/day → 5,000 GB/day) cuts per-GB
price ~15–30%.
Identify and cap top sources (DNS, NSG flow, Linux Syslog) — usually 80% of cost.
Daily cap as a guardrail (drops data above cap; alert on hit).
Review monthly: top tables, top resources, ingest trend.
Customer-managed keys (CMK) on the workspace if your data classification
requires. Plan for the dedicated cluster requirement and the operational complexity.
Table-level RBAC: scope analyst access to specific tables (e.g., HR mailbox
audit visible only to insider-risk team).
Sentinel: separate Sentinel Reader / Responder / Contributor roles.
Avoid Owner on the workspace for SOC analysts.
Cross-workspace and cross-tenant.
Sentinel workspace manager (preview/GA per region) for MSSP-style multi-tenant.
Cross-workspace queries via union for federated estates.
Guardrails
Default 90-day retention is rarely enough for security investigations. APT
dwell-time medians exceed that. Plan 180–730 days for primary detection sources.
Don't ingest everything in Analytics tier "to be safe." Cost spirals; add Basic
for "store and search-if-needed."
Transformations are powerful and dangerous. A column drop today breaks a
detection tomorrow. Source-control DCRs and review changes.
Daily cap as a cost control drops data — including security data. Pair with
alerting; never set a cap below high-water mark for security tables.
Don't deploy MMA in 2025+. Use AMA. Existing MMA workloads need a migration plan.
CMK on workspace requires dedicated cluster and ≥500 GB/day commitment. Big
operational and cost step.
Multi-workspace Sentinel = harder hunting. Justify per business reason, not per
team preference.
Archive recall is search jobs, not native query. SLA assumptions for IR change.
Diagnostic settings are per resource; per-subscription policy is what scales.
Enforce via Azure Policy.
Common anti-patterns
"Per-team workspace because each team wants 'their data'" — correlation breaks,
cost balloons.
"Analytics tier for everything" — 5x cost vs Basic for tables you query
reactively.
"NSG flow logs full-fidelity to Sentinel forever" — top cost driver. Sample,
transform, or Basic-tier.
"30-day retention to keep cost down" — incidents discovered in week 6 have no
data.
"Daily cap below normal volume" — silent data loss, missed detections.
"DCRs hand-edited in the portal across 50 subs" — drift, no audit. Bicep + Azure
Policy.
"Sentinel + Defender XDR in two separate workspaces" — cross-product hunting
pain. Unify.
"Upgraded to dedicated cluster for CMK without volume justification" — cost
surprise.
Example prompts
Design the Log Analytics workspace strategy for a 100,000-employee multi-region Sentinel deployment with sovereignty constraints in EU and APAC.
Migrate 6,000 servers from MMA to AMA via Defender for Cloud auto-provisioning; identify gaps in detection coverage during cutover.
Author DCRs that drop high-noise columns from SecurityEvent and Syslog with 35% expected ingest reduction.
Tier plan: which tables go Analytics vs Basic vs Auxiliary for a Sentinel + DfC workload, with retention.
Cost optimization: identify top 10 cost drivers in our 12 TB/day workspace and propose targeted reductions.
Configure table-level RBAC so insider-risk analysts see EmailEvents but not DeviceProcessEvents.
Customer-managed key rollout for the security workspace — dedicated cluster, key rotation, recovery runbook.
Daily cap and ingestion alerting strategy that protects budget without dropping detection-critical tables.
