OpenMetrics Explorer

OpenMetrics Explorer is a metric data query and visualization tool provided by WhaTap. You can query collected metrics with PromQL and visualize the results as a chart (Graph) and a table (Table).

Key features

Metric query with PromQL
Data visualization in Graph or Table view
When you select a time range, the time is automatically inserted into PromQL
Real-time metric monitoring

Screen layout

Top: time range picker, query input
Middle: Graph, Table, and Stacked Bar view switch buttons
Bottom: query results display (graph or table)

Why do we need PromQL?

OpenMetrics collects time series indicators for diverse systems and applications based on the Prometheus metric format. Prometheus is an open source monitoring tool that stores time based numeric data such as CPU usage, memory usage, HTTP request count, and error rate as metric indicators.

Most of these indicators are collected every few seconds. They keep changing over time and take the form of time series data.

Simple collection is not enough to answer questions like whether the service is healthy now, whether the error rate became higher than usual, or which API became slow.

This is why OpenMetrics Explorer needs a query language to analyze collected metrics. That role is performed by PromQL (Prometheus Query Language).

With PromQL you can run the following analysis:

Real-time monitoring: check current system status immediately (current CPU usage, active user count)
Trend analysis: identify change patterns over time (increase rate of requests per second during the last hour)
Comparison analysis: compare indicators across servers or services (error rate by server, response time by region)
Aggregation and calculation: perform complex math and statistics (overall error rate, average response time, top 10 slow APIs)
Alert condition setup: receive automatic alerts when thresholds are exceeded (error rate at least 5 percent, memory usage at least 90 percent)

PromQL User Guide

Query Execution Types

Query	Description
Instant Query	Retrieves data at a specific point in time. Used in the Table tab.
Range Query	Retrieves data at regular intervals between a start time and an end time.

Data Types

PromQL expressions evaluate to four types.

Type	Description
Instant Vector	A set of time series that each contain a single sample, all with the same timestamp
Range Vector	A set of time series that each contain data points over a time range
Scalar	A single numeric value
String	A single string value

Basic Usage

1. Metric Selection (Instant Vector Selectors)

Basic Metric Query

nginx_http_response_count_total

Label Filtering

nginx_http_response_count_total{status="200"}

Multiple Label Conditions

nginx_http_response_count_total{method="GET", status="200"}

2. Label Match Operators

Operator	Description
=	Label value matches exactly
!=	Label value does not match
=~	Label value matches a regular expression
!~	Label value does not match a regular expression

Regular Expression Examples

# Query only staging, testing, and development environments
nginx_http_response_count_total{environment=~"staging|testing|development"}

# All methods except GET
nginx_http_response_count_total{method!="GET"}

# Query replicas starting with "rep"
nginx_http_response_count_total{replica=~"rep.*"}

3. Range Vector Selectors

Selects data from a specific time window going back from the current moment.

# Data from the last 5 minutes
nginx_http_response_count_total[5m]

# Data from the last 1 hour
nginx_http_response_count_total{status="200"}[1h]

4. Offset Modifier

Retrieves data from a specific past point in time.

# Value 5 minutes ago
nginx_http_response_count_total offset 5m

# 5-minute rate from 1 week ago
rate(nginx_http_response_count_total[5m] offset 1w)

# Future comparison (negative offset)
rate(nginx_http_response_count_total[5m] offset -1w)

Practical Examples

Example 1: Requests by Status Code

nginx_http_response_count_total{instance="http://192.168.100.122:4040/metrics"}

Example 2: Specific Status Codes

# Only 200 responses
nginx_http_response_count_total{status="200"}

# Only 4xx errors (regex)
nginx_http_response_count_total{status=~"4.."}

# Only 5xx errors
nginx_http_response_count_total{status=~"5.."}

Example 3: Method and Status Code Combination

# GET requests with 200 responses
nginx_http_response_count_total{method="GET", status="200"}

# 404 responses among POST, PUT, DELETE requests
nginx_http_response_count_total{method=~"POST|PUT|DELETE", status="404"}

Example 4: Search by Metric Name

# All metrics that start with "job:"
{__name__=~"job:.*"}

# All metrics that start with "nginx"
{__name__=~"nginx.*"}

Example 5: Recent Time Range Data

# nginx responses from the last 5 minutes
nginx_http_response_count_total{status="200"}[5m]

# 404 errors from the last 1 hour
nginx_http_response_count_total{status="404"}[1h]

Example 6: Time Comparison

# Compare current value with value from 1 hour ago
nginx_http_response_count_total{status="200"}
nginx_http_response_count_total{status="200"} offset 1h

# Data from the same time yesterday
nginx_http_response_count_total offset 1d

Aggregation Functions

Aggregation functions combine multiple time series into a single result.

sum()

Adds all values across time series.

# Total number of requests across all instances
sum(nginx_http_response_count_total)

# Sum grouped by status code
sum by (status) (nginx_http_response_count_total)

# Sum grouped by instance
sum by (instance) (nginx_http_response_count_total)

avg()

Calculates the average value across all time series.

# Average number of responses across all instances
avg(nginx_http_response_count_total)

# Average by status code
avg by (status) (nginx_http_response_count_total)

count()

Counts the number of time series.

# Total number of time series
count(nginx_http_response_count_total)

# Number of time series grouped by status code
count by (status) (nginx_http_response_count_total)

max() / min()

Calculates the maximum and minimum values of time series.

# Maximum value
max(nginx_http_response_count_total)

# Minimum value
min(nginx_http_response_count_total)

# Maximum value by status code
max by (status) (nginx_http_response_count_total)

topk() / bottomk()

Returns the top or bottom K time series.

# Top 5 time series
topk(5, nginx_http_response_count_total)

# Bottom 3 time series
bottomk(3, nginx_http_response_count_total)

# Top 3 by status code
topk(3, sum by (status) (nginx_http_response_count_total))

Time Series Functions

rate()

The most commonly used function, rate() calculates the per-second average increase of a Counter metric.

# Per-second average number of requests over the last 5 minutes
rate(nginx_http_response_count_total[5m])

# Per-second average number of 200-status requests
rate(nginx_http_response_count_total{status="200"}[5m])

# Total requests per second across all instances
sum(rate(nginx_http_response_count_total[5m]))

Caution

Cautions

Use only with Counter metrics
The time range should be at least twice the scrape interval (for example, scrape interval 30 seconds → [1m] or longer)
Automatically handles counter resets (such as server restarts)

increase()

Calculates the total increase over a specified time range.

# Total number of requests in the last 5 minutes
increase(nginx_http_response_count_total[5m])

# Total number of errors in the last 1 hour
increase(nginx_http_response_count_total{status=~"5.."}[1h])

Relationship between rate() and increase()

increase(v[5m]) = rate(v[5m]) × 300 seconds

rate(): per-second rate
increase(): absolute increase

irate()

Calculates the instantaneous per-second rate of increase based on the last two data points.

# Instantaneous rate
irate(nginx_http_response_count_total[5m])

irate() vs rate()

irate(): suitable for fast-changing counters, useful for detecting short spikes
rate(): suitable for alerts and slowly changing counters, provides an average value

Time Range Functions

Aggregates values within a specific time window.

avg_over_time()

Calculates the average value over a given time range.

# Average value over the last 10 minutes
avg_over_time(nginx_http_response_count_total[10m])

max_over_time() / min_over_time()

Calculates the maximum or minimum value over a specified time range.

# Maximum value during the last 1 hour
max_over_time(nginx_http_response_count_total[1h])

# Minimum value during the last 1 hour
min_over_time(nginx_http_response_count_total[1h])

sum_over_time()

Calculates the sum of all values within a specified time range.

# Sum of all values in the last 5 minutes
sum_over_time(nginx_http_response_count_total[5m])

Important Rule: Rate then Sum

This rule applies when using time series functions such as rate(), increase(), and irate() with Counter metrics. It must be followed for all aggregation functions (sum, avg, max, min, count, etc.).

Caution

Always apply rate() first, then sum()

When aggregating Counter metrics, always maintain this order.

🔵 Correct Method

# Apply rate() first, then sum()
sum(rate(nginx_http_response_count_total[5m]))

❌ Incorrect Method

# Applying sum() first causes issues when counters reset
rate(sum(nginx_http_response_count_total[5m]))

If you apply sum() before rate(), counter resets on individual servers cannot be detected.
When one server restarts, the total sum suddenly decreases, and rate() miscalculates it as a negative increase or abnormal spike.
Applying rate() first ensures that counter resets are detected and handled correctly for each server.

Practical Examples

Example 1. Requests Per Second (RPS) Calculation

# Total RPS
sum(rate(nginx_http_response_count_total[5m]))

# RPS by status code
sum by (status) (rate(nginx_http_response_count_total[5m]))

# RPS by method
sum by (method) (rate(nginx_http_response_count_total[5m]))

Info

Explanation of the PromQL for Total Requests Per Second (RPS)

Why does sum(rate(nginx_http_response_count_total[5m])) represent RPS?

What is nginx_http_response_count_total?
- A Counter metric: the cumulative number of responses since the server started
- Example: 1000 → 1050 → 1120 → 1180 → 1250...
What does rate(...[5m]) do?
- It looks at data from the last 5 minutes and calculates the per-second average increase.
- Example:
  5 minutes ago: 1000
  Now: 1300
  Difference: 300 (5 minutes = 300 seconds)
  rate = 300 ÷ 300 seconds = 1.0 requests/second
→ The result of rate() = requests per second (RPS)
Why is sum() needed?
- It adds the rate values from multiple servers or labels.
- Example:
  server1, status=200 → 10 req/s
  server1, status=404 → 2 req/s
  server2, status=200 → 15 req/s
  server2, status=404 → 3 req/s
  Total = 30 req/s

Conclusion:
sum(rate(nginx_http_response_count_total[5m])) = total requests per second (RPS) across all servers.

Tip: To calculate the total number of requests over 5 minutes, use increase() instead.
sum(increase(nginx_http_response_count_total[5m])) = total number of requests in 5 minutes.

Example 2. Error Rate Calculation

# Percentage of 5xx errors out of all requests
sum(rate(nginx_http_response_count_total{status=~"5.."}[5m])) / sum(rate(nginx_http_response_count_total[5m])) * 100

# Ratio of 4xx errors out of all requests
sum(rate(nginx_http_response_count_total{status=~"4.."}[5m])) / sum(rate(nginx_http_response_count_total[5m]))

Example 3. Find Top N Instances

# Top 5 instances with the highest number of requests
topk(5, sum by (instance) (rate(nginx_http_response_count_total[5m])))

# Top 3 instances with the most errors
topk(3, sum by (instance) (rate(nginx_http_response_count_total{status=~"5.."}[5m])))

Example 4. Time-based Comparison

# Current RPS
sum(rate(nginx_http_response_count_total[5m]))

# RPS from 1 hour ago
sum(rate(nginx_http_response_count_total[5m] offset 1h))

# RPS from the same time yesterday
sum(rate(nginx_http_response_count_total[5m] offset 1d))

Example 5. Average Responses

# Average responses per instance
avg by (instance) (rate(nginx_http_response_count_total[5m]))

# Average by status code
avg by (status) (rate(nginx_http_response_count_total[5m]))

Math Operators

Arithmetic Operations

# Addition
sum(rate(nginx_http_response_count_total{status="200"}[5m]))
+
sum(rate(nginx_http_response_count_total{status="201"}[5m]))

# Subtraction
sum(rate(nginx_http_response_count_total[5m])) - sum(rate(nginx_http_response_count_total{status=~"5.."}[5m]))

# Multiplication (convert bytes to MB)
nginx_http_response_size_bytes / 1024 / 1024

# Division (calculate ratio)
sum(rate(nginx_http_response_count_total{status=~"5.."}[5m])) / sum(rate(nginx_http_response_count_total[5m]))

Comparison Operations

# Values greater than 100
nginx_http_response_count_total > 100

# Values equal to a specific number
nginx_http_response_count_total == 200

# Range condition
nginx_http_response_count_total > 100 and nginx_http_response_count_total < 1000

Practical Tips

1. Choosing Time Range Based on Scrape Interval

Scrape Interval	Minimum Range	Recommended Range	Reason
15s	[30s]	[1m] or more	Ensures 2~4 data points
30s	[1m]	[2m] or more	Ensures 2~4 data points
1m	[2m]	[5m] or more	Ensures 2~4 data points

2. Choosing Grouping Labels

# Too granular (too many time series)
sum by (instance, method, status, path) (rate(nginx_http_response_count_total[5m]))

# Appropriate grouping
sum by (status) (rate(nginx_http_response_count_total[5m]))

3. Writing Alert Rules

# When the error rate exceeds 5% during the last 5 minutes
sum(rate(nginx_http_response_count_total{status=~"5.."}[5m])) / sum(rate(nginx_http_response_count_total[5m]))
> 0.05

4. Function Combination Order

# 1. rate() first
# 2. Aggregation function (sum, avg, etc.)
# 3. Arithmetic operation

#Correct order
sum(rate(nginx_http_response_count_total[5m])) * 100

Other Functions

abs()

Returns the absolute value of a number.

abs(rate(nginx_http_response_count_total[5m]) - 100)

round()

Rounds a value to the nearest integer or a specified unit.

# Round to the nearest integer
round(sum(rate(nginx_http_response_count_total[5m])))

# Round to the nearest 10
round(sum(rate(nginx_http_response_count_total[5m])), 10)

clamp_max() / clamp_min()

Restricts values so that they do not exceed the specified maximum or fall below the minimum.

# Limit the maximum to 1000
clamp_max(nginx_http_response_count_total, 1000)

# Limit the minimum to 0
clamp_min(nginx_http_response_count_total, 0)

Notes

Matching Empty Label Values

When you match an empty label value, all series that either lack the label or have an empty value are also selected.

# Select all series without the "environment" label or with an empty value
nginx_http_response_count_total{environment=""}

Required Selectors

A vector selector must specify at least one metric name or a label matcher that does not match an empty string.

# ❌ Incorrect example
{job=~".*"}

# ✅ Correct examples
{job=~".+"}
{job=~".*", method="get"}

Performance Considerations

Tips for Efficient Queries

If you use only the metric name, thousands of time series may be selected.

Always use appropriate label filters to narrow down the results.
Start with the Table view to check the output, then filter appropriately before switching to the Chart view.
Ideally, the result should contain fewer than a few hundred series.

Staleness (Outdated Data)

If a time series is no longer being scraped, it becomes marked as "stale."

Once marked stale, it is excluded from query results.
By default, if no data is collected for 5 minutes, it is considered stale.

Regular Expressions

Prometheus uses RE2 syntax.

All regular expressions are fully anchored (Prometheus automatically adds ^ and $).
env=~"foo" is equivalent to env=~"^foo$".

Comments

# This is a comment
nginx_http_response_count_total{status="200"} # End-of-line comments are also supported

Tip

Recommended Learning Order

Basic: rate(), sum(), avg()
Intermediate: increase(), irate(), topk()
Advanced: complex ratio calculations, combining multiple functions

Note

References

Key features​

Screen layout​

Why do we need PromQL?​

PromQL User Guide​

Query Execution Types​

Data Types​

Basic Usage​

1. Metric Selection (Instant Vector Selectors)​

2. Label Match Operators​

3. Range Vector Selectors​

4. Offset Modifier​

Practical Examples​

Aggregation Functions​

sum()​

avg()​

count()​

max() / min()​

topk() / bottomk()​

Time Series Functions​

rate()​

increase()​

irate()​

Time Range Functions​

avg_over_time()​

max_over_time() / min_over_time()​

sum_over_time()​

Important Rule: Rate then Sum​

Practical Examples​

Math Operators​

Arithmetic Operations​

Comparison Operations​

Practical Tips​

Other Functions​

abs()​

round()​

clamp_max() / clamp_min()​

Notes​

Matching Empty Label Values​

Required Selectors​

Performance Considerations​

Staleness (Outdated Data)​

Regular Expressions​

Comments​

Key features

Screen layout

Why do we need PromQL?

PromQL User Guide

Query Execution Types

Data Types

Basic Usage

1. Metric Selection (Instant Vector Selectors)

2. Label Match Operators

3. Range Vector Selectors

4. Offset Modifier

Practical Examples

Aggregation Functions

sum()

avg()

count()

max() / min()

topk() / bottomk()

Time Series Functions

rate()

increase()

irate()

Time Range Functions

avg_over_time()

max_over_time() / min_over_time()

sum_over_time()

Important Rule: Rate then Sum

Practical Examples

Math Operators

Arithmetic Operations

Comparison Operations

Practical Tips

Other Functions

abs()

round()

clamp_max() / clamp_min()

Notes

Matching Empty Label Values

Required Selectors

Performance Considerations

Staleness (Outdated Data)

Regular Expressions

Comments