DevOops 2017: Docker, Docker, Docker
10/20/2017 I attended the DevOps conference in St. Petersburg and described my impressions. Do not count on hardcore, real examples in the spirit of "how to implement DevOps in the company for 5 days" or endless hosanna Docker, under the cut they do not.
The organization was top notch: there was enough coffee-cookies for everyone. No queues for registration, signposts in prominent places. We were provided with everything we needed, even popcorn. The reports were distributed in three halls, only the first and the last were held in the same (first) hall. This is an important point, in the first act it will become clear why I paid attention to it, for now just remember.
The conference itself began with a report from Corey Quinn about Netflix, Google and AWS. The report was in English and it was obvious from the reaction of the audience to the joke that a good half of the people do not understand at all what this is about. This is one of the few mistakes of the organizers, it was not worth making the first report in English without translation. The report itself turned out to be interesting, although there were no technical details in it, Corey explained the general approach.
')
To summarize briefly, Netflix is ​​a company with classic DevOps, in which developers have root to production. Yes, yes, there are such companies, and they work well. There is no magic here, they simply have so much money that they can afford to hire only the best of the best. They basically have no junior.
Cory told an important story from the early days of his career. He was a junior (which means he did not work at Netflix), and one day he made a serious mistake (facac - I was forbidden to use in the article). The next day, when the consequences were eliminated, the manager summoned him and they had a dialogue:
- You broke production yesterday.
- Yes, I broke it, this is my mistake. I deleted very important data.
- Do you understand what was wrong? What has been done so that this does not happen again?
- Yes, the error was in ... Made ...
- Ok, go work.
Corey was very surprised that he was not fired or even punished. For many managers, at least in Russia, this line of conduct looks strange. They are always looking for the guilty and punishing him, although it is more logical to look for the causes of the problem (facac - I was forbidden to use in the article) and eliminate them. A case in point is the story of databaseremoval specialist and gitlab. By dismissing the culprit, you will lose a specialist who has just received the most valuable experience, how not to do it, and increase the turnover rate of personnel. And the problem will not be solved, and sooner or later it will happen again.
Another important thought that Corey expressed, and then many speakers repeated: “Docker, Docker, Docker”. Now there is a “universal” solution to any problem, and this is Docker. They shove him everywhere, everybody goes to him. The technology itself is good, but HYIP made it anti-advertising. They started shoving it into all holes, but it would be worth first to answer the simple question: “Do we need a Docker?” It is important to understand that there are no bad technologies, there is their inappropriate use. In modern IT, this is a big problem.
I chose the second report between k8s, How to properly blame things for causing latency and the story about the organization of work and duty. Without much thought, I went on duty. First, this topic is close to our company. Secondly, all colleagues went to the first two reports. In the third, materials about the technology is fully online.
There is not much to write, you have to watch the report. Baruch and Igolnik have powerful charisma and excellent style of presenting information. On the example of the junior Vasya on duty, an example of handling one night incident was considered. All in the best traditions, and logs in the mail in the Word file, and panic, and the lack of access to the right servers, and a disgruntled client. Classic. In reality, this is a big problem for many companies, and instead of solving, they are always looking for the guilty.
Although the hall was half empty, and many in the process were gone, the topic raised was very important. Fortunately, there were no revelations, most of the proposals we have already implemented or at the stage of implementation.
I chose between SmartMonitoring - monitoring business logic on Odnoklassniki, Extending k8s and Managing multiple clouds FTW. The choice fell on the first report, so among the areas that I develop, there is monitoring.
Served by Sergei Sharapov. His report could be hardcore, but ... left mixed impressions.
The report dealt with monitoring in Odnoklasniki. They have written their system, it is able to build a graph and allocate servers that have failed. She is able to look for anomalies by analyzing previous data. Everything is cool, but not enough technical details. I still can not help but note a question from the audience:
- Why did you make your product instead of taking the ready ones as a basis?
- Because we can.
Act Five: Continuously delivering infrastructure to the cloud, the round table “How to“ sell “DevOps”, Continuous Delivery in a Windows environment.
Where to go, it was clear right away: a conference about Devops means that I am going to a round table.
It was interesting. Some conclusions:
1. Devops approach will not free us from work, it will not be less, but rather the opposite. But the tasks will reach a qualitatively different level, they will become more difficult and more interesting.
2. At the moment, the devops are divided into two approaches more and more clearly, the google, netflix approach, when dev has root in production. And all the rest. The second approach is not one person who can do everything, this is staff development when dev appear with an understanding of ops and ops with an understanding of development. This is a close interaction and understanding of common problems and requests.
3. Now the business is developing so rapidly that the only way is DevOps. But everyone has his own way, he should always be considered in context.
Act Six: Troubleshooting & debugging production applications in Kubernetes, “Success Story or“ Dev + DevOps + Ops ””, “One-cloud: OS-level data center in Odnoklassniki”.
The choice fell on Troubleshooting & debugging, it was necessary to at least one report to hear about Kubernetes. Speakers: Baruch Sadogursky, Ray Tsang.
The report was about JFrog and google cloud. An example of an unsuccessful release and hot fix on the fly for JavaScript was considered. Something difficult to distinguish was just curious.
Led her Baruch Sadogursky, Leonid Igolnik. What can I say, cool. It began with speakers in Greek togas and beer. Everything was presented in the manner peculiar to them, with humor, excellent pictures and gif-kami. They told about the development of a startup from three people to a large company. About the problems that the company had to face at every stage of development.
This report, in my opinion, has drawn a logical line. He showed that it is important to choose the right technologies at every moment of development. That there are no bad technologies, that there is their inappropriate use. That DevOps is not a set of specific rules that can be applied, and everything will be cool. That this is a certain approach, we need to change our approach to work, then it will be cool.
DevOps conference was a success. I want to come back here.
PS: In the breaks, many participants complained that there was little hardcore and real cases. In my opinion, they should not be there. This conference is not about that. It is about the approach to work, to the development of products and about the fact that Devops is not a magic pill, but a set of technologies and methodologies that, if used properly, will enable businesses to meet modern requirements. Many are not ready for this, they seek to use new technologies, not understanding whether they need it. No matter, you are junior or senior, the conference was useful for everyone. After all, even junior can come to the leadership with a proposal to improve the work of the company, and if he has a good justification, they will listen to him. And if you are not even listened to, then run away from such a company, there will never be DevOps in it, its ceiling is Docker :)
PSS: As you noticed, Docker is often found in the text. Simply, Baruch said that this is a trump card; the more mentions Docker are, the steeper the article.
UPD: here you can read "how it was" from the point of view of the organizers: habrahabr.ru/post/341020
Thanks to the organizers, speakers and all involved!
The organization was top notch: there was enough coffee-cookies for everyone. No queues for registration, signposts in prominent places. We were provided with everything we needed, even popcorn. The reports were distributed in three halls, only the first and the last were held in the same (first) hall. This is an important point, in the first act it will become clear why I paid attention to it, for now just remember.
Act one: Corey Quinn, Docker and root in production
The conference itself began with a report from Corey Quinn about Netflix, Google and AWS. The report was in English and it was obvious from the reaction of the audience to the joke that a good half of the people do not understand at all what this is about. This is one of the few mistakes of the organizers, it was not worth making the first report in English without translation. The report itself turned out to be interesting, although there were no technical details in it, Corey explained the general approach.
')
To summarize briefly, Netflix is ​​a company with classic DevOps, in which developers have root to production. Yes, yes, there are such companies, and they work well. There is no magic here, they simply have so much money that they can afford to hire only the best of the best. They basically have no junior.
Cory told an important story from the early days of his career. He was a junior (which means he did not work at Netflix), and one day he made a serious mistake (facac - I was forbidden to use in the article). The next day, when the consequences were eliminated, the manager summoned him and they had a dialogue:
- You broke production yesterday.
- Yes, I broke it, this is my mistake. I deleted very important data.
- Do you understand what was wrong? What has been done so that this does not happen again?
- Yes, the error was in ... Made ...
- Ok, go work.
Corey was very surprised that he was not fired or even punished. For many managers, at least in Russia, this line of conduct looks strange. They are always looking for the guilty and punishing him, although it is more logical to look for the causes of the problem (facac - I was forbidden to use in the article) and eliminate them. A case in point is the story of databaseremoval specialist and gitlab. By dismissing the culprit, you will lose a specialist who has just received the most valuable experience, how not to do it, and increase the turnover rate of personnel. And the problem will not be solved, and sooner or later it will happen again.
Another important thought that Corey expressed, and then many speakers repeated: “Docker, Docker, Docker”. Now there is a “universal” solution to any problem, and this is Docker. They shove him everywhere, everybody goes to him. The technology itself is good, but HYIP made it anti-advertising. They started shoving it into all holes, but it would be worth first to answer the simple question: “Do we need a Docker?” It is important to understand that there are no bad technologies, there is their inappropriate use. In modern IT, this is a big problem.
Act two: Baruch Sadogursky, Leonid Igolnik and Vasya on duty
I chose the second report between k8s, How to properly blame things for causing latency and the story about the organization of work and duty. Without much thought, I went on duty. First, this topic is close to our company. Secondly, all colleagues went to the first two reports. In the third, materials about the technology is fully online.
There is not much to write, you have to watch the report. Baruch and Igolnik have powerful charisma and excellent style of presenting information. On the example of the junior Vasya on duty, an example of handling one night incident was considered. All in the best traditions, and logs in the mail in the Word file, and panic, and the lack of access to the right servers, and a disgruntled client. Classic. In reality, this is a big problem for many companies, and instead of solving, they are always looking for the guilty.
Although the hall was half empty, and many in the process were gone, the topic raised was very important. Fortunately, there were no revelations, most of the proposals we have already implemented or at the stage of implementation.
Act Three: Monitoring
I chose between SmartMonitoring - monitoring business logic on Odnoklassniki, Extending k8s and Managing multiple clouds FTW. The choice fell on the first report, so among the areas that I develop, there is monitoring.
Served by Sergei Sharapov. His report could be hardcore, but ... left mixed impressions.
The report dealt with monitoring in Odnoklasniki. They have written their system, it is able to build a graph and allocate servers that have failed. She is able to look for anomalies by analyzing previous data. Everything is cool, but not enough technical details. I still can not help but note a question from the audience:
- Why did you make your product instead of taking the ready ones as a basis?
- Because we can.
Act Four did not remember anything, so we omit it.
Act Five: Continuously delivering infrastructure to the cloud, the round table “How to“ sell “DevOps”, Continuous Delivery in a Windows environment.
Where to go, it was clear right away: a conference about Devops means that I am going to a round table.
It was interesting. Some conclusions:
1. Devops approach will not free us from work, it will not be less, but rather the opposite. But the tasks will reach a qualitatively different level, they will become more difficult and more interesting.
2. At the moment, the devops are divided into two approaches more and more clearly, the google, netflix approach, when dev has root in production. And all the rest. The second approach is not one person who can do everything, this is staff development when dev appear with an understanding of ops and ops with an understanding of development. This is a close interaction and understanding of common problems and requests.
3. Now the business is developing so rapidly that the only way is DevOps. But everyone has his own way, he should always be considered in context.
Act Six: Troubleshooting & debugging production applications in Kubernetes, “Success Story or“ Dev + DevOps + Ops ””, “One-cloud: OS-level data center in Odnoklassniki”.
The choice fell on Troubleshooting & debugging, it was necessary to at least one report to hear about Kubernetes. Speakers: Baruch Sadogursky, Ray Tsang.
The report was about JFrog and google cloud. An example of an unsuccessful release and hot fix on the fly for JavaScript was considered. Something difficult to distinguish was just curious.
The culmination was DevOps in scale: the Greek tragedy in three acts.
Led her Baruch Sadogursky, Leonid Igolnik. What can I say, cool. It began with speakers in Greek togas and beer. Everything was presented in the manner peculiar to them, with humor, excellent pictures and gif-kami. They told about the development of a startup from three people to a large company. About the problems that the company had to face at every stage of development.
This report, in my opinion, has drawn a logical line. He showed that it is important to choose the right technologies at every moment of development. That there are no bad technologies, that there is their inappropriate use. That DevOps is not a set of specific rules that can be applied, and everything will be cool. That this is a certain approach, we need to change our approach to work, then it will be cool.
Epilogue
DevOps conference was a success. I want to come back here.
PS: In the breaks, many participants complained that there was little hardcore and real cases. In my opinion, they should not be there. This conference is not about that. It is about the approach to work, to the development of products and about the fact that Devops is not a magic pill, but a set of technologies and methodologies that, if used properly, will enable businesses to meet modern requirements. Many are not ready for this, they seek to use new technologies, not understanding whether they need it. No matter, you are junior or senior, the conference was useful for everyone. After all, even junior can come to the leadership with a proposal to improve the work of the company, and if he has a good justification, they will listen to him. And if you are not even listened to, then run away from such a company, there will never be DevOps in it, its ceiling is Docker :)
PSS: As you noticed, Docker is often found in the text. Simply, Baruch said that this is a trump card; the more mentions Docker are, the steeper the article.
UPD: here you can read "how it was" from the point of view of the organizers: habrahabr.ru/post/341020
Thanks to the organizers, speakers and all involved!
Source: https://habr.com/ru/post/340738/
All Articles