There is such a superstition in IT: "It works, don't touch it." This can be said about our monitoring system. At Southbridge, we use Zabbix - when we chose it, it was very cool. And, in fact, he had no alternatives.
Over time, our ecosystem has acquired instructions, additional bindings, and integration with redmine has appeared. Zabbix had a powerful competitor that was superior in many aspects: speed, HA almost out of the box, beautiful visualization, optimization of work in the kubernethes environment.
But we are not in a hurry to move on. We decided to take a peek at Zabbix and ask what features they plan to make in the next releases. We did not stand on ceremony and asked uncomfortable questions to Sergey Sorokin, Zabbix development director, and Vitaly Zhuravlev, Solution architect. What came of it, read on.
1. Tell us about the history of the company. How was the product idea born?
The history of the company began in 1997, when the founder and owner of the company Alexey Vladyshev worked as a database administrator in one of the banks. It seemed to Alexei inefficient to manage the bases without having data on the historical values of various parameters, without understanding the current and historical state of the environment.
At the same time, the monitoring solutions currently on the market are very expensive, clumsy, and resource-intensive. Therefore, Alexey begins to write various scripts that allow him to effectively monitor the part of the infrastructure entrusted to him. It's turning into a hobby. Aleksey changes jobs, but interest in the project remains. In 2000-2001, the project was rewritten from scratch - and Alexey thinks about how to enable other administrators to use the developments. At the same time, the question arose under which license to release the existing code. Alexey decided to release it under the GPLv2 license. The tool was immediately noticed in a professional environment. Over time, Alexey began to receive requests for support, training, and expanding the capabilities of the software. The number of such orders is constantly growing. So, naturally, the decision to create a company came. The company was founded on April 12, 2005
2. What are the key points in the history of Zabbix development?
There are several such points at the moment:
A. Alexey started working on scripts in 1997.
b. Publication of code under the GPLv2 license - 2001.
V. Zabbix was founded in 2005.
d. Conclusion of the first partnership agreements, creation of a partnership program - 2007.
e. Founding of Zabbix Japan LLC - 2012.
f. Founding of Zabbix LLC (USA) - 2015
and. Founding of Zabbix LLC - 2018
3. How many people do you employ?
At the moment, the Zabbix group of companies employs a little more than 70 employees: developers, testers, project managers, support engineers, consultants, salespeople, marketing employees.
4. How do you write a roadmap, do you collect feedback from users? How do you decide where to go next?
When creating a Roadmap for the next version of Zabbix, we focus on the following important factors, more precisely, we collect the Roadmap in the following categories:
A. Strategic improvements to Zabbix. Something that Zabbix itself considers very important. For example, Zabbix agent written in Go.
b. Things that Zabbix customers and partners want to see in Zabbix. And for which they are willing to pay.
V. Wishes/suggestions from the Zabbix community.
d. Technical debt. 🙂 Things that we released in previous versions, but did not provide full functionality, did not make it flexible enough, did not offer all the options.
5. Can you compare Zabbix and prometheus? What is better in Zabbix and what is worse?
The main difference, in our opinion, is that Prometheus is a system primarily for collecting metrics - and in order to assemble full-fledged monitoring in an enterprise, you need to add many other components to Prometheus, such as graphana for visualization, a separate long-term storage, and separate management somewhere problems, work with logs separately ...
There will be no typical monitoring templates in Prometheus, having received all the thousands of metrics from exporters, you will need to find problematic signals in them yourself. Prometheus setup - configuration files. Somewhere it is more convenient, somewhere not.
Zabbix is a universal platform for creating monitoring "from and to", we have our own visualization, correlation of problems and their display, distribution of access rights to the system, audit of actions, many options for collecting data through an agent, proxy, using completely different protocols, the ability to quickly extend the system with plugins, scripts, modules…
Or you can simply collect the data as is, for example, via the HTTP protocol, and then turn the responses into useful metrics using preprocessing functions such as JavaScript, JSONPath, XMLPath, CSV, and the like. Many users appreciate Zabbix for the ability to configure and manage the system through a web interface, for the ability to describe typical monitoring configurations in the form of templates that can be shared with each other, and containing not only metrics, but also detection rules, thresholds, graphs, descriptions - a complete set of objects for monitoring typical objects.
Many also like the ability to automate management and configuration through the Zabbix API. In general, I do not want to arrange a holivar. It seems to us that both systems are well suited for their tasks and can harmoniously complement each other, for example, Zabbix from version 4.2 can collect data from Prometheus exporters or from Prometheus itself.
6. Have you thought about making zabbix saas?
We thought about it and will do it in the future, but we want to make this solution as convenient as possible for customers. In this case, standard Zabbix should be offered along with communication tools, advanced data collection tools, and so on.
7. When can we expect zabbix ha? And wait?
Zabbix HA is definitely waiting. We really hope to see something in Zabbix 5.0 LTS, but the situation will become clearer in November 2019 when the Zabbix 5.0 Roadmap is fully confirmed.
8. Why does media type have such a poor selection out of the box? Do you plan to add Slack, telegram and so on? Does anyone else use jabber?
Jabber was removed in Zabbix 4.4, but Webhooks were added. Regarding media types, I would not like to make specific applications from the system, but standard messaging tools. It's no secret that a lot of these chats or service desks have an API via HTTP - so this year, with the release of 4.4, the situation will change.
With the advent of webhooks in Zabbix, you can expect all the most requested integrations out of the box in the near future. At the same time, integrations will be two-way, and not just simple one-way notifications. And those media types that we don’t get to will be made by our community - after all, now the entire media type can be exported to a configuration file, posted on share.zabbix.com or github. And other users will only need to import the file to start using this integration. In this case, you do not have to install any additional scripts!
9. Why is the direction of Virtual machine discovery not developing? There is only vmware. Many are waiting for integration with ec2, openstack.
No, the direction is developing. For example, in 4.4, datastore discovery appeared through the vm.datastore.discovery key. Very cool wmi.getall keys also appeared in 4.4 - we expect that through it, together with the perf_counter_en key, it will be possible to make good monitoring of Hyper-V. Well, there will be other important changes in this direction in Zabbix 5.0.
10. Have you thought about abandoning templates and doing like prometeus, when everything that is given is taken away?
Prometheus automatically takes all the metrics, it's convenient. And a template is more than just a set of metrics, it is a “container” that contains all the necessary typical monitoring configuration for a given type of resource or service. It already has a set of important triggers, graphs, detection rules, descriptions of metrics and thresholds that help the user understand what is being collected, and which thresholds are being checked and why. At the same time, templates are easy to share with other users - and they will get good monitoring of their system, even without necessarily being an expert in it.
11. Why are there so few metrics out of the box? This greatly complicates the setup in terms of operation.
If out of the box you mean ready-made templates, then right now we are working on expanding and improving our templates. Zabbix 4.4 came out with a new, improved set and better features.
For Zabbix, you can always find a ready-made template for almost any system on share.zabbix.com. But we decided that we should make basic templates ourselves, setting an example for the rest, and also freeing users from writing another template for some MySQL. Therefore, now in Zabbix there will only be more official templates with each version.
12. When it will be possible to build triggers that are not tied to hosts, but for example, based on labels. For example, we have a site being monitored from n different points, and we want a simple trigger that fires when the site is not available from 2 or more points.
In fact, this functionality has been in Zabbix for several years, written for one of the clients. The client is ICANN. Also, such checks can be done, for example, through aggregated items or using the Zabbix API. Now we are actively working to simplify the creation of such checks.
PS: At one of the Slurms, Zabbix developers asked us what we want to see in the product in order to monitor Kubernetes clusters using Zabbix, and not Prometheus.
It's great when developers go to meet customers, and do not remain a thing for themselves. And now we meet each release with sincere interest - the good news is that more and more features that we talked about are taking on flesh and blood.
As long as the developers do not withdraw into themselves, but are interested in the needs of customers, the product lives and develops. We will follow new Zabbix releases.
P.P.S: We will launch an online monitoring course in a few months. If you're interested, subscribe so you don't miss the announcement. In the meantime, you can go through our .
Source: habr.com