First 10 Minutes: High Latency¶

When function execution latency is elevated or users report slow responses, use this checklist to narrow down the cause within the first 10 minutes.

Prerequisites¶

Azure CLI access to the production subscription.
Access to Application Insights and Log Analytics.
Health endpoint implemented at GET /api/health.

Set shared variables:

RG="rg-myapp-prod"
APP_NAME="func-myapp-prod"
SUBSCRIPTION_ID="<subscription-id>"
APP_INSIGHTS_NAME="appi-myapp-prod"
WORKSPACE_ID="xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"

flowchart TD
    A[High latency reported] --> B["Check regional/platform status"]
    B --> C[Latency trend by function]
    C --> D{Cold start correlation?}
    D -->|Yes| E[Plan-specific cold-start mitigation]
    D -->|No| F[Dependency latency analysis]
    F --> G{Timeout signatures present?}
    G -->|Yes| H["Apply timeout/architecture mitigation"]
    G -->|No| I[Check recent deployments]

1) Check Azure status and regional incidents¶

Rule out platform-wide latency degradation.

Check in Portal¶

Azure portal → Service Health → Health advisories.

Filter for the production region and services: Azure Functions, Storage, Azure Monitor.

Check with Azure CLI¶

az account set --subscription "$SUBSCRIPTION_ID"
az rest --method get \
  --url "https://management.azure.com/subscriptions/$SUBSCRIPTION_ID/providers/Microsoft.ResourceHealth/events?api-version=2022-10-01&\$filter=eventType eq 'ServiceIssue' and status eq 'Active'"

How to Read This¶

Signal	Interpretation	Action
No active service issues	Latency is app or dependency-level	Continue to Step 2
Active incident on Storage or Networking	Platform dependency degraded	Monitor advisory, apply mitigations

2) Check function execution latency trend¶

Determine whether latency is broad or isolated to specific functions.

Check with KQL¶

let appName = "func-myapp-prod";
requests
| where timestamp > ago(1h)
| where cloud_RoleName =~ appName
| where operation_Name startswith "Functions."
| summarize
    P50Ms = percentile(duration, 50),
    P95Ms = percentile(duration, 95),
    P99Ms = percentile(duration, 99),
    Invocations = count()
  by FunctionName = operation_Name, bin(timestamp, 5m)
| order by timestamp desc

Example Output¶

FunctionName              timestamp               P50Ms    P95Ms    P99Ms    Invocations
------------------------  ----------------------  -------  -------  -------  -----------
Functions.HttpTrigger     2026-04-04T11:30:00Z    120      450      890      85
Functions.HttpTrigger     2026-04-04T11:25:00Z    115      420      850      92
Functions.QueueProcessor  2026-04-04T11:30:00Z    45       180      350      210
Functions.ExternalDep     2026-04-04T11:30:00Z    2100     8500     12000    30

How to Read This¶

Pattern	Interpretation	Action
One function slow, others normal	Function-specific issue (dependency or code)	Focus on that function's dependencies
All functions slow	Platform or shared dependency issue	Check cold starts and storage health
P95 >> P50	Tail latency — likely cold starts or intermittent dependency	Check cold start frequency
Latency rising over time bins	Progressive degradation	Check memory and dependency trends

3) Check cold start impact¶

Cold starts are the most common cause of tail latency in Azure Functions.

Check with KQL¶

let appName = "func-myapp-prod";
traces
| where timestamp > ago(1h)
| where cloud_RoleName =~ appName
| where message has "Host started"
| summarize StartupCount = count() by bin(timestamp, 5m)
| join kind=leftouter (
    requests
    | where timestamp > ago(1h)
    | where cloud_RoleName =~ appName
    | summarize P95Ms = percentile(duration, 95) by bin(timestamp, 5m)
) on timestamp
| order by timestamp desc

How to Read This¶

Pattern	Interpretation	Action
Startup count high + P95 elevated	Cold starts driving tail latency	Pre-warm instances or upgrade plan
Startup count normal + P95 elevated	Not cold start related	Check dependencies
FC1 with many startups + low latency	Normal Flex Consumption scaling	No action needed

Plan-specific cold start behavior

Consumption (Y1): Cold starts after idle periods are expected, and first-hit latency is commonly in the seconds range.
Flex Consumption (FC1): Cold-start impact is generally reduced versus Y1, but startup-related tail latency can still appear under bursts.
Premium (EP): Always-ready/prewarmed capacity reduces cold-start risk. If cold starts appear, review always-ready and prewarmed instance configuration.
Dedicated: No cold starts unless app restarts.

4) Check dependency latency¶

Dependency bottlenecks are the second most common cause of function latency.

Check with KQL¶

let appName = "func-myapp-prod";
dependencies
| where timestamp > ago(1h)
| where cloud_RoleName =~ appName
| summarize
    P95Ms = percentile(duration, 95),
    FailureRate = round(100.0 * countif(success == false) / count(), 2),
    Calls = count()
  by target, type
| order by P95Ms desc

Example Output¶

target                  type     P95Ms    FailureRate  Calls
----------------------  -------  -------  -----------  -----
api.partner.internal    HTTP     8500     2.40         120
stmyappprod.blob.core   Azure    45       0.00         890
stmyappprod.queue.core  Azure    12       0.00         1240

How to Read This¶

Pattern	Interpretation	Action
One dependency P95 >> others	That dependency is the bottleneck	Investigate downstream service
Storage dependency slow	Storage account may be throttled	Check storage metrics and region
All dependencies slow	Network-level issue	Check VNet, DNS, private endpoint

5) Check for timeout errors¶

Azure Functions have hard execution timeouts that vary by plan.

Timeout limits by plan¶

Plan	Default Timeout	Maximum Timeout
Consumption (Y1)	5 minutes	10 minutes
Flex Consumption (FC1)	30 minutes	Up to 4 hours
Premium (EP)	30 minutes	Unlimited
Dedicated	30 minutes	Unlimited

Check with KQL¶

let appName = "func-myapp-prod";
traces
| where timestamp > ago(1h)
| where cloud_RoleName =~ appName
| where message has_any ("timeout", "exceeded", "Timeout value of", "cancellation", "Task was cancelled")
| project timestamp, severityLevel, message
| order by timestamp desc

How to Read This¶

Signal	Interpretation	Action
`Timeout value of 00:05:00 exceeded`	Consumption plan timeout hit	Reduce execution time or upgrade plan
`Task was cancelled`	CancellationToken triggered	Check if function handles graceful shutdown
No timeout messages	Latency is not from timeouts	Focus on dependency and cold start causes

HTTP trigger 230-second limit

HTTP-triggered functions have an additional 230-second timeout imposed by the Azure load balancer, regardless of the functionTimeout setting in host.json. For long-running HTTP work, use the Durable Functions async HTTP pattern.

6) Check recent deployments¶

az monitor activity-log list \
  --resource-group "$RG" \
  --offset 2h \
  --status Succeeded \
  --output table

Correlate deployment timestamps with latency onset.

Fast routing after triage¶

What you see	Likely area	Next action
Cold starts driving P95	Scaling/plan	Use Cold Start lab guide
Single dependency bottleneck	Downstream service	Investigate target service health
Timeout errors in logs	Execution limits	Use Timeout / Execution Limit playbook
All functions slow after deploy	Regression	Roll back, then follow Methodology
Memory pressure signals	Resource exhaustion	Use Out of Memory playbook

First 10 Minutes: High Latency¶

Prerequisites¶

1) Check Azure status and regional incidents¶

Check in Portal¶

Check with Azure CLI¶

How to Read This¶

2) Check function execution latency trend¶

Check with KQL¶

Example Output¶

How to Read This¶

3) Check cold start impact¶

Check with KQL¶

How to Read This¶

4) Check dependency latency¶

Check with KQL¶

Example Output¶

How to Read This¶

5) Check for timeout errors¶

Timeout limits by plan¶

Check with KQL¶

How to Read This¶

6) Check recent deployments¶

Fast routing after triage¶

See Also¶

Sources¶