Skip to main content
Availability8.6.9 or later

What Is Monitored

Workers, traffic, requests

How It Works

Status monitor and log parser

Deploy

One wizard per server

KPIs

Workers, rate, throughput

Overview

GermainUX shows you how busy each Apache HTTP Server is and what it serves. You see busy and idle workers, request rate, throughput and Apache process memory next to every request from the access log, with its URL, status and client. When a server runs out of workers or a backend starts returning errors, the charts show it within a minute. Your team finds the saturated server without logging in to read server-status pages and log files one by one.

Worker Saturation at a Glance

Busy and idle workers per server, so you see a server reach its worker limit before users time out.

Every Request, Searchable

Method, URL, status, response size, client, user, referrer and browser for each access-log entry.

Failures Stand Out

Requests with status 400 or higher count as failed, so error spikes on a URL or a server show up right away.

Ready-Made Dashboard

The Apache HTTP dashboard charts workers, traffic, process memory and TCP retransmissions out of the box.

Questions It Answers

  • Is any Apache server close to running out of workers?
  • Which URLs return 5xx errors right now, and since when?
  • Is one server handling more traffic or failing more requests than the others?
  • Are clients or bots scanning for pages that don’t exist?
  • Is the network dropping packets between clients and the web server?
  • Did traffic really change, or did Apache restart?
  • Is every Apache server still sending data to GermainUX?

What Is Monitored

Analytics, Alerts and Automation

AI Analysis

Ask AI Explore questions such as “Which Apache URLs failed most this week?” and get the answer without building a report. Outlier detection flags unusual spikes or drops in busy workers, request rate and failures.

Dashboards and Reports

GermainUX ships the Apache HTTP dashboard. In the left menu, go to Dashboards > All, and then open Apache HTTP. It has one trend chart per KPI: Apache Busy Workers, Apache Idle Workers, Apache Process Count, Apache Request Rate, Apache Request Size, Apache Shared Memory, Apache Throughput and TCP Retransmissions. Busy workers that climb while idle workers fall to zero mean the server has reached its worker limit. New requests then wait in the queue or time out.
Apache HTTP dashboard over four hours, with busy workers, process count and shared memory rising at 10:00 and 11:00 while idle workers fall, next to request rate, request size, throughput and TCP retransmissions

Apache HTTP Dashboard - GermainUX

Build your own dashboards and reports on the access-log requests: volume and failure rate by server or URL, status-code distribution, and the busiest clients.

Analysis

Click a chart on the dashboard to open its KPI in the Analysis screen. The segmentation breakdown lists the servers and applications that changed most, and labels each one as a spike, an outlier or degraded.
Analysis screen for the Apache Busy Workers KPI, with average busy workers up 361.5% on the previous days, a peak on 16 November in the relationships chart, and request rate and TCP retransmissions as related metrics

Analysis of Apache Busy Workers - GermainUX

The relationships chart adds the KPIs related to the one you opened: Apache Idle Workers next to Apache Busy Workers, Apache Throughput next to Apache Request Rate and Apache Request Size, Apache Busy Workers next to Apache Process Count and TCP Retransmissions, and Inbound HTTP Request next to Apache Busy Workers and Apache Request Rate. The segmentation breakdown compares the servers, so you see which one moved.

Investigate Backend Errors Behind mod_proxy

When an application server behind Apache stops, mod_proxy answers 503 to every request it forwards there, and each of these requests counts as failed. Open the Inbound HTTP Request KPI in the Analysis screen, filtered to failed Apache requests: Success is false and Member is Apache HTTP. The top row counts the failed requests, compares them with the same hours on the previous days, and shows the hourly trend. The segmentation breakdown shows which server they come from. In the example below, app-02 stops for 45 minutes on the afternoon of 15 November. mod_proxy on web-02, which forwards /app/ and /api/ requests to app-02, answers 503 to each of them. The screen counts 46 failed requests between 11:00 and 16:00 UTC, up 411.1% on the previous days, all in one hour. The breakdown marks web-02 as a spike with 44 of them, so you know which backend to check. The failures stop when app-02 is back. An SLA on the failure rate per server, with an alert, tells you about the next outage as it starts.
Analysis screen for failed Apache requests from 11:00 to 16:00 UTC on 15 November, with 46 failed requests against 9 on the previous days, one tall bar in the hourly trend, and web-02 marked as a spike with 44 failed requests in the segmentation breakdown

Aggregate Analysis of Failed Apache Requests During a Backend Outage - GermainUX

Investigate a 404 Scan

A vulnerability scanner or a bot that probes for files that don’t exist causes a burst of 404 responses. Open the failed Apache requests in the Analysis screen, as for backend errors. The segmentation breakdown lists the servers, user agents and URLs that the failed requests have in common. In the example below, a scanner probes both servers for 50 minutes in the night of 16 November, asking for paths such as /.env, /.aws/credentials, /xmlrpc.php and /backup.tar.gz. The screen counts 843 failed requests between 23:00 and 04:00 UTC, against 2 at the same hours on the previous days. The breakdown splits them between web-01 (440) and web-02 (403), marks the Robot user agent device as new, and lists each probed path. None of these paths exists on the site, so the requests don’t come from real users.
Analysis screen for failed Apache requests from 23:00 on 15 November to 04:00 UTC on 16 November, with 843 failed requests against 2 on the previous days, web-01 and web-02 marked as spikes, the Robot user agent device marked as new, and the probed paths /backup.tar.gz, /.aws/credentials, /xmlrpc.php, /.env and /vendor/phpunit in the segmentation breakdown

Aggregate Analysis of Failed Apache Requests During a 404 Scan - GermainUX

Investigate Packet Loss on a Web Server

TCP Retransmissions holds the total number of TCP segments the Apache host has retransmitted since it started, so its line always climbs. Packet loss shows as a steeper climb, not as a spike. Click the TCP Retransmissions chart on the Apache HTTP dashboard to open it in the Analysis screen, and compare the slope of each server in the segmentation breakdown. In the example below, a faulty switch port drops packets on web-02 for 45 minutes in the morning of 15 November. The breakdown marks web-02 as degraded: its total climbs steeply during the fault, then levels off, and ends up 28.7% higher than on the previous days. The total of web-01 climbs slowly and evenly, 7.1% above the previous days.
Analysis screen for the TCP Retransmissions KPI from 06:00 to 12:00 UTC on 15 November, with a step in the relationships chart, and web-02 marked as degraded with a steep climb next to web-01 climbing slowly in the segmentation breakdown

Aggregate Analysis of TCP Retransmissions - GermainUX

Investigate a Request Rate Change After a Restart

Apache Request Rate, Apache Throughput and Apache Request Size are averages that mod_status computes over the time since Apache started. After days of uptime they barely move, so their lines look flat. When Apache restarts, mod_status starts again from zero. The first values cover only the few requests since the restart, and the line then follows the traffic of the hours since. Busy workers, idle workers and process count are current values, so a restart doesn’t move them this way. Before you read a jump in these three KPIs as a traffic change, check whether Apache restarted at that time. In the example below, Apache restarts on web-01 and web-02 at 05:00 UTC on 16 November to load a renewed certificate. Before the restart, the request rate is a flat line at about 0.046 requests per second. Right after it, the line dips, because few requests arrive at that hour, and then climbs with the day’s traffic to close to 0.1 requests per second in the afternoon. Both servers show the same shape in the segmentation breakdown. Traffic didn’t change; only the period that the average covers did.
Analysis screen for the Apache Request Rate KPI from 03:00 to 08:00 UTC on 16 November, with an average of 0.046 requests per second, a flat relationships chart that dips after the restart and then rises in the afternoon, and the same dip-and-climb line for web-01 and web-02 in the segmentation breakdown

Aggregate Analysis of Apache Request Rate After a Restart - GermainUX

SLAs and Alerts

Set SLAs on busy workers, request rate or the failure rate of requests, and get an alert by email, SMS, Slack or Teams when one is breached. An SLA on idle workers tells you when a server is about to run out of capacity.

Correlation with Other Technologies

Correlate Apache requests with browser sessions and with the application servers behind Apache, such as Apache Tomcat, to follow a failed request to its cause.

How It Works

server-status and access.logEngineKPIsDashboardAlerts
The Apache HTTP Server wizard applies a server profile that deploys three components to the GermainUX Engine on the Apache host: The Engine runs the status commands and reads the log files locally, so it runs on the Apache host. See File and Log Monitoring for how GermainUX reads log files.

Before You Start

  • A GermainUX Engine is deployed on the Apache host, which runs Linux. See GermainUX Engine.
  • The Apache host is set up as a server in GermainUX, and the Engine on it is registered for that server.
  • mod_status is enabled, and http://APACHE_HOST/server-status answers on port 80 from the Apache host itself.
  • ExtendedStatus is On, so server-status reports request rate, throughput and request size.
  • The access log uses the combined format and is written to a file named access.log.
  • The Engine’s account can read the access log directory, for example /var/log/apache2.
  • /usr/bin/wget and /bin/netstat are installed on the Apache host.

Deploy

1

Check the Apache configuration

In the Apache configuration, check that server-status is enabled and that the access log uses the combined format:
apache2.conf
If you changed these lines, reload Apache.
2

Open the Apache HTTP Server wizard

In the left menu, go to Settings > Wizards. On the Apache HTTP Server card, click Add.
3

Select the Apache servers

In the Apply Apache HTTP Server dialog, select the Apache hosts in Servers, and then click Apply.
Apply Apache HTTP Server dialog over the Wizards screen, with one server selected in Servers and the Apply button

Apply Apache HTTP Server Dialog - GermainUX

4

Fill in the profile settings

Fill in the fields:Finish the wizard. GermainUX deploys the three components to the Engine on each selected server, and starts reading server-status and the access log.

Validate

1

Check server-status

On the Apache host, request the status page the Engine reads:
Replace APACHE_HOST with the host name of the Apache server. The output shows the number of busy and idle workers.
2

Send test requests

From any machine that can reach the Apache server, request a page and a path that doesn’t exist:
3

Find the data in GermainUX

After a few minutes, open the Apache HTTP dashboard and check that each chart shows the server. Then open the Inbound HTTP Request KPI and look for the requests to / and /germainux-test-missing. Check that the 404 request counts as failed.

Troubleshooting

Check that wget -qO- http://APACHE_HOST/server-status returns the status page on the Apache host, and that mod_status allows requests from the host itself. Check that the components run on the Engine of the Apache host: in the left menu, go to Settings > Wizards, and click Manage on the Apache HTTP Server card.
server-status reports these values only when ExtendedStatus is On. Turn it on and reload Apache.
Check that the access log file is named access.log, that it’s in the Access Log Directory you selected, and that the Engine’s account can read it. A log written to another file name, such as access_log or ssl_access.log, isn’t picked up.
GermainUX skips lines that don’t match the combined format. This happens when a CustomLog line uses another format, such as common or a custom one. Use the combined format for access.log.
The monitor reads TCP statistics with /bin/netstat. Install the package that provides it, such as net-tools.

Security and Privacy

Apache HTTP Server monitoring data includes personal data:
  • Identity data: the client address and the authenticated user name of each request
  • URLs: the request path and query string, and the referrer URL. Query strings can contain email addresses, IDs or tokens, so treat them as possibly containing personal or customer data.
  • Device data: the user agent of each request, with the browser and operating system
To mask, anonymize or exclude this data, see Data Privacy.

KPIs

The Apache KPIs have no default SLA. Add the thresholds you need. Inbound HTTP Request is a shared KPI that GermainUX uses for other web servers too. See KPIs and SLAs.

Apache Tomcat

Application servers behind Apache

File and Log Monitoring

How GermainUX reads log files

Automation

More Automation Features

Get Help

The GermainUX team can help you monitor Apache servers that use a custom log format, another log file name or a non-default status URL.

Contact GermainUX Support

The GermainUX team helps you get up and running faster, empower your team, turn data into actionable insights, and optimize performance to deliver measurable business outcomes.

Can't Find What You Need?

Search all GermainUX docs with Ctrl + K, or use the feedback buttons at the bottom of this page to tell us what is missing.

The content of this document is proprietary to GERMAIN SOFTWARE.