Sysadmins are divided into two categories - those who already use monitoring, and those who do not yet.
Joke of humor.
The need for monitoring comes in different ways. Some were lucky and monitoring came from the parent company. Everything is simple here, everything has already been thought about for you - what, what and how to monitor. And even for sure they have already written the necessary manuals and explanations. Others come to this need themselves and the initiative comes, as a rule, from the IT department. The downside is that you will have to collect all the bumps and go through the rake on your own experience. There are also pluses - you can choose any monitoring system and monitor only what is necessary, as well as come up with your own principles for responding to problems. At different times I worked in different companies, but where I was close to monitoring, I went the second way.
A brief excursion into the past
The first "experience" was in the distant past. It was one of the local providers where I was suddenly a storekeeper. Managed equipment was expensive back then, and so thefts and breaks were tracked using Friendly Pinger by pinging several clients that were constantly or almost constantly online. Worked so-so, but the best was not.
Then, at another local provider, admins used Nagios. By and large, I did not have access there, so I was not able to assess its capabilities. However, managed equipment was used at each site and monitoring was likely to be an effective tool.
Then I got into a company that is a backbone provider and offered home Internet as a subsidiary service. Zenoss was used here in all its glory. I didnβt get deep into it, but I was able to feel all its power and benefits - only the magic of regexp is worth it ... Thoughtful professionals assembled, configured the system and wrote regulations.
And at the next place of work, I came to the need to find out about problems before some chief accountant talks about it. Considering that there was time for creative experiments, I went to see what the folk industry offers us.
Flour choice
In fact, the choice turned out to be surprisingly simple. Of course, all felt-tip pens are different in taste and color, so my criteria and views that are relevant at that time may not be suitable for you. I remembered several systems and I will briefly describe my thoughts in relation to them.
As a Windows admin, the first thing that came to my mind was the Custer Center in all its glory. The first and main advantage is its integration into the Microsft environment, and without the tambourine, but Native. The second advantage is an integrated approach. Let's be honest, System Center is never a purely monitoring system - it's still an infrastructure maintenance system. However, this is the first disadvantage. Deploying this monster just for the sake of monitoring does not make any sense. Now, if all sorts of backups and the deployment of a million VDS were needed ... And the cost of implementation is not encouraging, because you will have to go broke twice - first on licenses, and then on the servers where it will live.
Next, let's turn to the past in the face of Nagios. The system was dropped immediately, since manually configuring the system through configuration files makes the system unattended. I do not blame people who like to flip through fifteen hundred lines of the same type in order to correct a single parameter, but I donβt want to do this myself either.
Zenoss. Great system! Everything is there, everything can be configured with an acceptable level of complexity, but it is a bit heavy. We didnβt have those scales, we never used any nested groups. And the engine itself turned out to be too demanding on resources. For what? Refused.
Zabbix is ββour choice. Attracted by fairly low system requirements and ease of launch. In fact, it took several minutes to launch. Download the image for VMWare and click the "enable virtual machine" button. All! Iβll tell you more, for our needs this βstarting imageβ would be quite enough, although we soon deployed everything as it should.
There was also Cacti on the original list, but it just didn't get there. Well, what's the point if Zabbix took off from the first kick and everyone immediately liked it? Therefore, I canβt say anything about Cacti.
After written
The company in which I implemented Zabbix safely died a natural death. The owner said βI'm tired of everything, I'm closing the businessβ, so there's nothing to talk about monitoring there. We looked after the servers, the Internet and tunnels at all sites and collected counters from printers.
Then PRTG was in my life for a short time. For my taste, it works great with Windows systems, uses a curious agent mechanism and costs obscene money. This is a rather sad ideology of access to version updates.
The company I currently work for uses Zabbix. It was not my choice, but I am happy with it and support it completely. Given the state of the monitoring system prior to my arrival, I almost recreated everything from scratch. The understanding that "we are doing something wrong" was. And even a new server with Zabbix was deployed, but there was no person who would take on this task and bring it to the end. We have not yet reached full enlightenment in monitoring, but we want to believe that we know the direction. The process of bringing monitoring to the ideal is endless, although I have already formulated the main theses for myself.
Source: www.habr.com