When doing the cluster, either the ARR Nginx, Microsoft can achieve cluster, but if the update node code file? For example: there are 100 machines, to update the code of the files in it now, that is to update the one hundred times, according to is impossible, how can you do batch updates, [] Windows platform,, every brother please answer.

CodePudding user response:

A variety of solutions, the first thing you how to deploy deployed 100 machines? Running a manual installation system? We are using ghost one-time installation, each machine as long as a few minutes, there are actually (at the bottom of the operating system) management system can batch ghost, in addition, the.net framework itself is integrated with the click once technology, MEF, and so on, can write their own ten lines of code to invoke framework, like the. Net framework of distributed Actor new version supports hot update mechanism, presumably,

CodePudding user response:

It is as you say, "cluster environment update file?"

Such as a DLL to 100 machines do you want?

I use goodsync... A do upload FTP client. Then another 99 open FTP service on the first installation software and configuration once more change directory and update mechanism.

So once you have uploaded to the first of this software will automatically change the file to distribute to the other 99 stage...

CodePudding user response:

Hot update needs to be considered, that is, the runtime can replace process have been loaded and cached in memory of all sorts of DLL,

CodePudding user response:

Once can't hot update, you may only "stable update way" is on the outside of the operating system, using a lower level of network publishing system to stop the whole system, then the local to the whole system to replace,

CodePudding user response:

Use automation tools, I use a Jenkins

CodePudding user response:

A few years ago I wrote a simple example to demonstrate the function of MEF at that time: https://bbs.csdn.net/topics/392444516
Actually we also use a very simple technique, because we are on the server side has a "waiting" service is responsible for according to the configuration file regularly pull up various console application, so we wait program also go to the website regularly compare our console program a monitoring files in the directory list, compare each need to monitor the files md5, once found, will kill the console process, then automatically to the site is not the same as the up and down with the md5 files, downloaded automatically to pull up the process of the console, it can run very stable, of course, the premise is to make the necessary testing, do not worry about the task was forced to interrupt,

CodePudding user response:

Finally, the emphasis is not actually a began to pursue the bells and whistles, but the success rate of 100% in order to get stable, will not allow the occasional update failure phenomenon,

CodePudding user response:

Oh, I write the demo in post https://bbs.csdn.net/topics/391970445
I still want to emphasize that the so-called CICD is very costly, best most valuable solution is very simple and direct,

CodePudding user response:

refer to the eighth floor interacting in a professional developer response:

oh, I wrote the demo in post https://bbs.csdn.net/topics/391970445
I still want to emphasize that the so-called CICD is very costly, the best the most valuable solutions are very simple and direct,

ok, thank you for your reply, I study here, thank you very much!

CodePudding user response:

reference 5 floor is nu month god reply:

using automation tools, I use a Jenkins

ok, thank you, I study

CodePudding user response:

refer to the second floor by_ love reply:

it is as you say, "cluster environment update file?"

Such as a DLL to 100 machines do you want?

I use goodsync... A do upload FTP client. Then another 99 open FTP service on the first installation software and configuration once more change directory and update mechanism.

So once you have uploaded to the first of this software will automatically change the file to distribute to the other 99 stage...

thank you, I study the

CodePudding user response:

It is as you say, automated deployment

The normal situation is the first update to git (public or private, there is no)

Then each side and git synchronization

CodePudding user response:

How much human development, no matter how much time, but it is in order to meet the requirements of the test, not for the sake of application, and the deployment of real application oriented, more server deployment automatically, pay attention to don't use any extra things don't always go wrong, this is very important,

CodePudding user response:

upstairs said goodsync rather easy to use