The operation and maintenance automation technology craze in the end of the house

Original works, allow reprint, please be sure to use hyperlinks in the form of the original source of the article, author information and this statement. Otherwise, the legal liability will be investigated. http://caoyameng.blog.51cto.com/4975863/1359732

Operation and maintenance automation is the beginning of the 2010 speculation is very hot a concept, but also many engineers, employers blindly excited for a long time, I also follow the puppet and Python learning, job seekers often spend a lot of time in the interview Hingui Automation.

But calm down and think about it, the so-called automation, just let the training institutions make money gimmick only.

A word summary of operation and maintenance automation

"Operation and maintenance Automation" a few words too abstract easy to be subjective into a lot of concepts, on the Wikipedia search for IT operations Automation introduction is too detailed, hat too much.

If the operation and maintenance automation in a sentence is clear, the comparative official said: "Operation and maintenance automation is in the enterprise business more and more complex, more and more demanding IT personnel ... balabalabla ... Under the premise, by artificial can not meet the operation and maintenance of the demand, only rely on automation technology to solve the problem. ”

If the use of a more rough saying is "live many people less, operations do not want to rely on the heap of manpower to solve the cumbersome problems, can only rely on the operation and maintenance of automation to their burden." ”

Operation and maintenance automation theory is contrary to reality

It makes sense to look at these theories, but careful analysis is not the case at all. First of all, are we really busy?

I think the workload of operation and maintenance has not become larger with the increasing complexity of enterprise demand, even if the increase is not by automation can solve the physical life.

Operation and maintenance automation is for operation and maintenance, please OPS think, our daily work, these years have changed too much?

Primary operations in most of the time to do on-line and monitoring, advanced operations in the revision of the structure of the bug. For those repetitive jobs, cloud computing providers can do better than you do, cloud hosting, cloud monitoring, Cloud RDS, cloud storage and so on services are to the operation and maintenance of the burden.

Enterprise business needs more and more complex is true, in particular, the technical progress of enterprise requirements more and more tricky. Database requires the master-slave real-time synchronization, storage can not be used with NFS distributed, front-end business requirements seamless switching and so on. Did I talk about tricky, what does this have to do with operational automation? You are aware of the problem, we have new business requirements These years, not much can be solved by operation and maintenance automation, to solve these problems, but also rely on our own brains.

operation and maintenance Automation =shell Script

In fact we have been doing operations automation because we will use shell scripts.

We can say that as long as the needs of enterprises change, we need to take services, monitoring, to do all these things to write automated scripts. When you get excited about "Automation scripts", I want to ask, do you not write shell scripts?

After a service, an experienced and responsible operation, will naturally write the System optimization script, copy monitoring and control template. If we use puppet, with Python, we end up with the task of specifying the hostname.

We can use shell script to complete a variety of simulation operations operation, skilled use of shell script is also required for each operation, we need to learn python for a gimmick?

I have read the official documents of puppet, and the resources he can manage are listed as "files", "belonging to the main group" "Mount" "Package" "Service" "-exec use local shell", which in my opinion is "file" and "-exec".

In the Linux shell script, there are so many commands about operations "CP, SCP, NC, SSH, rsync, SVN, chmod, chown, service,/etc/init.d/", these commands are enough

When I use puppet, I just use him to monitor several important system configuration files frequently. On-line work is really cumbersome in order to take realserver from LB, and need to use manpower to determine whether can pick.

Specific pick-up equipment, new old code, restart the Java container, rollback code work I have written the script, so also because of the careless several times the high load or lost steps.

If you can automate things, you will be able to write shell scripts to do, if the shell scripting things, "operation and maintenance automatic-hang."

Operation and maintenance automation ≠ optimization

It's a cliché that OPS should be a bit more eye-opener, not always busy optimizing the job at hand, and having to work on it is not necessary.

A friend is sure to say that my work is not home, on-line incredibly need human judgment, I admit that this is the problem, but this problem in the framework is not operational. If manual intervention is not required on-line, why not just let development execute? Even further, let the application server periodically go to SVN to detect the new code? In the test environment we will also use Hudson and MAVEN to make the development of their own, but I am sure to do a system image to ensure that they play the system is bad and can quickly recover.

In the production environment, operations should not be a struggle to do the human flesh operation of the shell or Python to do, but to say good to push, can be more than a few servers, can reduce the coupling degree, do not let us manually stare at the line work.

I am not doing well in the background of the company, a lot of business-related SQL modifications should be developed to write the statement to the operation of the implementation. If this time I give MySQL install a phpadmin is putting the cart before the horse, write a script can automatically pass SQL past or put the cart before the horse, I actually do is to urge the company as soon as possible to make the enterprise management background can let operations and customer service personnel directly to change business data.

We write how many good Python script, rather than a stable to a single room power outage will not be downtime of the architecture; Is it good to use lucky dimension automation? Yes, it's just as good as using some kind of text editor.

Benefits driven by operational automation

In fact, many of the gurus who preach automation have developed two short-legged fish.

I have seen an operation Automation tutorial, the author is very serious to teach us, how to use some kind of automation tools to invoke the local shell, with the SED command crontab in the Ntpdata task time to change. See this paragraph, I was his persistent stupid cry, so-called automation incredibly is to use Ntpdate Update system time.

I have also seen a master written by the automation code, my friend told me that his Python level value 6k--even exception is not processed, I use Dabbler level carefully read his source I really laughed out, every few lines must be able to see a os.system ("Shell command").

In the working environment, I use "tar/var/aaa/bbb/ccc/*.jpg" this kind of wildcard characters to match out the target file, wrote a 10 line script, will be a master with Perl wrote more than 100 lines, but in fact is Find+tar script to replace.

I also write Python scripts when working with data because it's far more efficient than shell scripting. But operation and maintenance automation must use the Python script, update the file must use puppet, for the master this is the style, for the novice this is the follow-up.

Interested friends can help to check, from 2010 onwards, which training institutions have added operations Automation courses or Python operations courses, and who rely on these technologies to package themselves as masters.

The dilemma of operation and maintenance automation

Those high-end air-grade automation teachers, can never avoid my two problems:

1, in a 100 machine under the small company, engaged in operation and maintenance automation is not in their own project similarly? You write good operation and maintenance automation system, is not with the document written very fine very good, will not be the system upgrade on the operation of automatic hanging.

The smaller the company, the easier it is to have a single machine running multiple businesses, different machine environment variables completely different situations. Let's say you're a tech recruit, you don't have to automate a machine, you hang a bunch of machines with automation; If you're a tech whiz, you know the risks don't blindly trust a script.

2, in 500 machines and machines above the large companies, it is really necessary to operation and maintenance automation, otherwise, just hand animation network topology map and monitoring can be exhausted people.

But in this environment, the most important and the most valuable is the design and evolution of the system architecture, operation and maintenance automation is only the burden of work, which there are smart people put the bricks do not want to brick?

Do you think it's possible that this company dozens of technical master every day for uploading a JS file tired of dying, waiting for you an airborne to deploy automation system to rescue them?

Do operation and maintenance automation, it is inevitable that the internal server has a large number of increase to the point where you feel the manual operation is very tired, this time to do operation and maintenance automation is natural. But operation and maintenance automation is generally the work of enterprises within the existing operations to promote, this should not be used as a reason for hiring.

Operations Automation is not a simple script or deployment file Synchronization tool, it does not really shape the scheme, because it is to use the machine to simulate the operation of operations engineers. Each OPS team's work style is different, mechanically external automation solutions will only make us managernent difficult.

If you were asked to design a deployment Automation program in the first month of your career, it would only prove that the company does not have a talent for operations and that the company is idle and does not really need to be operational.

Re-examine operation and maintenance automation

The purpose of operation and maintenance automation, low-end point, is to solve the operation of manual error-prone problems, high-end point is to hope that operations ignore specific commands and pay more attention to the final results.

In the low-end domain, we can confidently say that the shell script is operation and maintenance automation, in the high-end area, must have set up the automation environment for us to observe learning and modification; If you are fortunate enough to be involved in large-scale automated deployments, it is a really interesting challenge; at a higher level, you will find things like System standardization, application of modular, unified certification system and so more valuable but no one hype technology. Operation and maintenance automation does not have to study specifically, automated "master" does not have to deliberately recruit.

The most handy tool is the best tool

It people who love a technology should be the master of a technology rather than a believer; I have repeatedly emphasized the usability of shell scripts in this article because shell scripts are the skills that each operation must master.

In this article I have cited a number of critical cases of some of the most popular automated operations tools. But these examples are not used to attack these technologies themselves. In fact, the puppet app QQ Group is the only Linux technology group I've ever been in, and I know my python level is laborious to see complex code. Just because these two technologies are the most popular in the wild wind, I use them to show that each of the large group of people blindly follow.

If you believe that a technology is particularly strong and angry at what I say, please think about what you can do with your tools, if I can get around in my environment, even if I can get rid of the shell script. I have no objection to your promotion of your proposal, but I think the "Loop invoke SSH command is an acceptable, workable solution."

We should reduce the blind obedience, pick up the most convenient tool to do a career, rather than the most beautiful props but lost the goal.

This article is from the "Make It Clearer" blog, so be sure to keep this source http://caoyameng.blog.51cto.com/4975863/1359732

The operation and maintenance automation technology craze in the end of the house