Log-based metrics in GCP - Step-by-Step Execution

Process Flow - Log-based metrics

Logs Generated by Services

↓

Logs Collected in Cloud Logging

↓

Define Log-based Metric Filter

↓

Metric Created and Updated

↓

Metric Used for Monitoring/Alerting

↓

Review Metric Data and Adjust Filter

Logs are collected, filtered by a metric definition, then counted or measured to create metrics used for monitoring.

Execution Sample

GCP

resource.type = "gce_instance"
severity >= ERROR

This filter selects logs from GCE instances with severity ERROR or higher to create a custom error count metric.

Process Table

Step	Action	Filter Evaluation	Metric Update	Result
1	Log generated: INFO from GCE instance	Does not match severity >= ERROR	No metric increment	Metric remains 0
2	Log generated: ERROR from GCE instance	Matches filter	Increment metric by 1	Metric is 1
3	Log generated: WARNING from GCE instance	Does not match severity >= ERROR	No metric increment	Metric remains 1
4	Log generated: ERROR from other resource	Resource type mismatch	No metric increment	Metric remains 1
5	Log generated: CRITICAL from GCE instance	Matches filter	Increment metric by 1	Metric is 2
6	Metric used in alerting policy	Metric value = 2	Alert condition evaluated	Alert triggered if threshold exceeded
7	Filter updated to include WARNING	Future logs with WARNING match	Metric increments on WARNING logs	Metric count increases faster
8	No new logs	No filter match	Metric remains unchanged	Metric stable
9	End of trace	-	-	-

💡 No new logs or filter changes; metric updates stop.

Status Tracker

Variable	Start	After 1	After 2	After 3	After 4	After 5	After 6	After 7	After 8	Final
Metric Count	0	0	1	1	1	2	2	3	3	3
Filter Severity Threshold	ERROR	ERROR	ERROR	ERROR	ERROR	ERROR	ERROR	WARNING	WARNING	WARNING

Key Moments - 3 Insights

Why does the metric not increment when a WARNING log appears before the filter update?

Because the filter only matches logs with severity ERROR or higher, WARNING is lower severity and does not match (see execution_table step 3).

What happens when the filter is updated to include WARNING logs?

Logs with WARNING severity now match the filter and increment the metric (see execution_table step 7 and variable_tracker showing filter change).

Why does a log from a different resource type not affect the metric?

The filter includes resource.type = "gce_instance", so logs from other resources do not match and do not increment the metric (see execution_table step 4).

Visual Quiz - 3 Questions

Look at the execution table at step 2. What is the metric count after processing the ERROR log?

A 0

B 2

C 1

D Error, metric not updated

💡 Hint

Check the 'Metric Update' and 'Result' columns at step 2 in the execution_table.

At which step does the filter change to include WARNING logs?

A Step 6

B Step 7

C Step 5

D Step 8

💡 Hint

Look at the 'Filter Severity Threshold' variable in variable_tracker and execution_table step descriptions.

If a CRITICAL log from a GCE instance is generated at step 10, what happens to the metric count?

A Metric count increments by 1

B Metric count decreases

C Metric count stays the same

D Metric resets to zero

💡 Hint

CRITICAL severity is higher than ERROR, so it matches the filter and increments the metric (see execution_table step 5).

Concept Snapshot

Log-based metrics count or measure logs matching a filter.
Define a filter on log fields like resource type and severity.
Matching logs increment the metric count.
Metrics can trigger alerts or dashboards.
Updating filters changes which logs count.
Useful for monitoring specific log events in GCP.

Full Transcript

Logs are generated by cloud services and collected in Cloud Logging. A log-based metric is created by defining a filter that selects logs based on criteria like resource type and severity. When logs match the filter, the metric increments or measures data accordingly. This metric can then be used for monitoring or alerting. For example, a filter selecting ERROR logs from GCE instances increments a custom metric each time such a log appears. If the filter is updated to include WARNING logs, those logs also increment the metric. Logs from other resource types do not affect the metric if the filter excludes them. This process helps track important events in logs and respond to them automatically.