Dynatrace review: Server & website monitoring made easy

For IT admins, they sometimes cause sleepless nights, and yet they are loved: Monitoring tools. Be it for the server infrastructure, application, website or the cluster in the cloud - Dynatrace wants to make life easier for admins. And act very differently than other monitoring tools. As? We'll clarify that review.

Lots of monitoring tools like Nagios & Co. do their job well, no question. They send timely notifications when the hard drive is full, the memory is full, the CPU is busy and much more. Also that the application, server or website is currently unavailable, loads slower than before or the server throws another 5xx error. The cause of this ProBut you always have to get to the bottom of problems yourself. Searching through logs for hours – maybe even at night – is pure madness. Dynatrace wants to improve exactly this point and look behind the scenes: What exactly is going on there? What is causing this error? Were there configuration changes?

To test whether the software can actually keep its promises, we threw Dynatrace on our server and just let it run. We will now clarify how the monitoring tool performs in everyday life - of course, put through its paces TechnikNews sysadmin offprobeers.

What we clarify in this test

Dynatrace can do a lot of things that are beyond the scope of this article. Since the TechnikNews Infrastructure is generally relatively simple, let's focus here on monitoring websites on a Linux server. We leave out the use of Dynatrace in Kubernetes clusters, AWS, Azure and more.

What normal monitoring tools do TechnikNews do

Let's take a concrete look at our test and let's talk about the status quo: We at TechnikNews use a combination of Nagios and own software in the background as a website crawler for service monitoring. With this setup we monitor all our servers and websites. If a service is not available, we can intervene quickly and that Profix the problem. This pushes using a webhook TechnikNews Monitoring the incidents directly on the smartphone via Telegram bot, automatically writes the error to ours via API Status page and also send me details by email.

If a certain service is not available, the search for the cause begins. The identification of ProThe problem mostly lies in looking at the relevant logs at the point in time when they occur. Or you can take it easy, restart the server and just hope it works again. You shouldn't make it that easy for yourself - after all, you have to find out what actually happened and how it could have happened. Is there an alternative? No - at least I thought so. Until I got to know Dynatrace.

What about Dynatrace TechnikNews silent

Let's get to the exciting part, what is Dynatrace actually doing now? Basically, regardless of the constellation in which you run the services (Kubernetes, cluster, virtualization) - Dynatrace should work for almost all use cases. Unless very outdated software is used on the servers, then the monitoring may not work as well as desired. It is generally recommended, also here at Dynatrace, to always run new software on the server in a dev environmentprobeer This is the only way to ensure smooth operation when switching to the Proproductive environment guaranteed.

Dynatrace wants to offer an all-in-one monitoring solution. (Image: Dynatrace)

Dynatrace is at TechnikNews meanwhile responsible for the entire monitoring. It runs on our Windows server, on our Linux systems and the TechnikNews Host system with the virtualization software Proxmox. With all of these environments, Dynatrace prohandle easily. Even if Dynatrace Procannot closely monitor xmox-specific functions (but Dynatrace supports VMware for example).

Nonetheless, Dynatrace becomes more intelligent during operation and tries to understand the system. For example, if the CPU load is high every day at 12 noon, Dynatrace learns and does not send any additional monitoring notifications. If the website loads a few seconds longer than usual or if a service takes longer to run, Dynatrace reports immediately via app push, e-mail or webhook. These messages are based on exceeding the “baseline” (the typical loading time / execution time), which Dynatrace should be aware of after several weeks of operation.

Synthetic monitoring

With what is known as “Synthetic Monitoring”, Dynatrace could completely replace our previous monitoring system as an all-in-one tool. Dynatrace acts like a normal client (desktop or mobile device) and accesses a specific URL from up to three different regions in a monitoring process. Settings for the client and user agent can be made individually - be it for example an accessing Apple iPhone X or a Samsung Galaxy S20. You can then choose from almost 100 different data centers. A maximum of three can be selected in one step. The specified URL is then accessed from these at adjustable intervals from five minutes to four hours.

If the loading time in a defined period (about 5 minutes) is exceeded by one second, for example, a Problem opened and you get a notification. What with Proproblem is meant exactly, we'll clarify right away - hold on.

dynaTrace Proproblem analysis

Via the <strong>integrated datalog</strong> the measuring values can be stored on the humimeter RH5 paper moisture meter and additional data can be added. You also have the possibility to use the Autolog function. This function automatically saves measuring values in adjustable time intervals. Problem viewer, in this case the longer loading time of the TechnikNews Identify the homepage right away. (Image: TechnikNews)

Should be Proproblem occurs, a new one will appear Proproblem in the form of a "ticket" in Dynatrace. At the same time, you can be notified by email, push notification with the Dynatrace app, Slack and many other services via webhook. Dynatrace can also generate a Jira ticket directly for IT, for example. The email contains a link to Proproblem view, which you can find in Dynatrace under "Proproblems” finds.

In this case, a "response time degradation" is the cause of the message. In the “impacted service” section we will also find further details on this. So our web server has to deliver the TechnikNews Homepage between 15:13 p.m. and 15:19 p.m. with 15,4 requests per minute instead of 40,5 milliseconds, this requires 206 milliseconds. A click on “Analyze response time degradation” also provides the culprit. This was determined by Davis - the Dynatrace KI - using the evaluation of 25.000 correlations on our system.

The Dynatrace KI "Davis" has identified the culprit of the longer loading time. (Image: TechnikNews)

Here Nginx itself was the deciding factor Proproblem, which led to longer loading times. Since we were updating Nginx at this point, this analysis is entirely correct and the culprit is immediately known to us. In another case, according to Dynatrace, slow hard drive access was the cause - which was quickly taken care of for us, since a backup was running at the same time.

Many services, many Proerror messages at the beginning

At the beginning there are hardly any Proerror messages because the tool needs to get to know the system better. Once this learning phase of a few weeks is complete, the monitoring tool learns to function properly, initially sends quite a few messages and opens Problem tickets. For example, if a service loads a few milliseconds longer, a TCP connect with a service is at 0 for a few seconds Procent sinks – all things that at a Procan happen in the productive system and have little or no influence on the services operated.

The threshold for notifications can be set for fewer notifications and e-mails. (Image: TechnikNews)

Of course you can find all these messages on the respective detail page of the Proswitch off zesses, but it would be good to have a wizard that lets you configure the notifications right from the start. If TechnikNews now loads a few milliseconds longer, doesn't really affect it. So we have adjusted the settings in the "Anomaly Detection". For example, we only want to be notified when TechnikNews loads a second longer, it affects at least 50 requests per minute and that Proproblem persists for at least five minutes.

Installation of Dynatrace in seconds

The installation on our Linux web server with Ubuntu was completed within a few seconds, then you have to restart the server, so everything is done within a few more seconds. This is necessary so that Dynatrace can communicate directly between the OS and the respective services/Pro(Apache, PHP, ...) can latch in. To do this, the so-called “OneAgent” (the Dynatrace Client) is simply installed on the server under the “Deploy Dynatrace” menu item. Common systems such as Windows, Linux, Kubernetes, AWS Lambda and many other environments are supported.

But not everything is perfect

Navigate

Like every tool, Dynatrace still has some room for improvement in some points. The structure of the settings and the sidebar should be revised. The familiarization with Dynatrace is very complicated due to the very nested menu structure. Although the developers have made the menu significantly better in the latest version 1.221, a few improvements will have to be made here.

Rights system

The rights system is not perfect either. A lot of settings are necessary to invite a second user to his account, who is only allowed to see one server of the infrastructure. Once the so-called "management zones" and "monitoring groups" have been created and assigned to the respective user, there are things like that Proproblem analysis and component relationships in part Proproblem with permissions. It would be practical to be able to assign a user directly to a server and all those belonging to the server Proalso release the user's data for display.

Conclusion

Whether Dynatrace is really suitable for hobby server operators is debatable. But you don't want to serve this target group either. It's more about large customers, significantly more complex infrastructure and services, which must be highly available. Here Dynatrace scores with the easy installation, endless setting options, the ProProblem analysis taking into account all events on the system, which becomes more and more intelligent over time, and the negligible resource consumption of the OneAgent.

In principle, one can say that Dynatrace is "only" a monitoring service. Yes, but Dynatrace can do a lot more. You don't just get notifications about Proproblem, but also the cause with the affected requests/services is displayed directly in a detailed overview. A usual monitoring would be about that Problem draw attention to the analysis of the Proproblems must be taken care of by the respective IT admin. What with a simple infrastructure like at TechnikNews Sometimes this is difficult, more complex systems are even more difficult. This is exactly where Dynatrace comes in. Even if the Dynatrace pricing May seem expensive at first glance, but the saved working time with overtime and lots of cups of coffee should justify the prices again.

Dynatrace will continue to monitor TechnikNews take over. In addition, the future development of the tool - especially in the area of Vulnerability detection (Monitoring messages about security gaps in the services used) - still pretty exciting.

Transparency Notice: Dynatrace was mine for TechnikNews made available for testing as part of a summer internship. However, this article only reflects my own opinion.

Recommendations for you

• Confluence with critical security gap: attackers can take over servers [+ demo]
• Microsoft Exchange Server security gaps: Current build numbers, overview of all updates and testing
• OVH: Hosting company fire destroys hundreds of servers in the data center
• Microsoft Exchange Server with huge security gap: Update urgently!