Deployment Strategy

1. In-Place Update

In-place means that it does not require extra servers during the update.

Pros:

you don't need extra serve

Cons:

the traffic capacity will be reduced during the update


1. All-At-Once

All At Once strategy is the basic one. You stop all the servers, and update all of your servers, and restart your service.

Fatal disadvantage is, the service should be completely shut down during the update.

2. Rolling Update

Rolling Update means you update your server one-by-one. 

Let's suppose you have 4 servers running software ver 1.0. When you try to update those to 2.0,

1.0 1.0 1.0 1.0 -> 1.0 1.0 1.0 2.0 -> 1.0 1.0 2.0 2.0 -> 1.0 2.0 2.0 2.0 -> 2.0 2.0 2.0 2.0


2. Blue/Green Update

Blue/Green Update maintains server running the sw of current version(Blue) and starts the new server that runs the sw of next version(Green). When the Green servers are able to handle all the traffics properly, and the Blue servers are killed.

1. All At Once

Blue/Green Update also has an All-At-Once strategy. Unlike the In-Place All-At-Once, it does not have any downtime because blue servers are not down for the update; New sw are installed on the Green servers.


2. Canary(Two-Step Canary)

In this strategy, a small percentage of the traffic(which is called 'Canary group', usually 5 to 10%) is handled by the Green servers, while the rest traffics are still handled by the Blue servers: Step 1. After a period of time, the rest traffics are also handled by the Green servers and the Blue servers are killed: Step 2.

The key is that you can test and monitor during the Step 1 whether the Green server is running properly. If not, you can take actions like stopping the update, or rollback the changes done by Green servers.


3. Linear(Linear Canary)

This is similar to the Canary(Two-Step Canary), but the difference is that while the Two-Step Canary shifts a small Canary Group at first and shifts all of the rest later(e.g. 10% -> 100%), Linear strategy shifts the traffic gradually by fixed time interval(e.g. 10% -> 20% -> .... 100%)



Q1. What are the advantages of Blue/Green Update over Rolling Update?

One difference is that B/G Update can hold the full traffic while update. However, another key difference is that there is no environment isolation in Rolling Update because new sw are installed in the existing server. This also means that Rolling Update carries increased risks and complicated process of rollback if a deployment fails.

댓글

댓글 쓰기

이 블로그의 인기 게시물

Interface of Java

 You can define(declare + initialize) variable like interface_name variable_name = new class_name() In such cases, the variable can only use method overriding the interfaces methods. Also, the language does not recognized it as the class; the language recognizes it as interface. Still, it does have member variables of the class and there are some libraries that enables the programmer to tell what class it is constructed(i.e. Reflection) Interface can be used as parameter of functions or return types of functions. When interface is a parameter, the class that implements the interface should be given to the function. When interface is a return type, it can be used for polymorphism. Also, variable of java(in case its not primitive type like int, float), it is rather closer to pointer of C++; it refers to the value, not contains value itself. The important point is that, when the class implements the interface, all of its public methods should be the interface's methods. If not, client...
자세한 내용 보기

Data Analytics Overview(OLTP vs OLAP)

You’ve likely heard about the lucrative careers of high-level Data Analysts, but have you ever wondered how they actually handle "Big Data"? Of course, they aren’t just using Microsoft Excel. To understand the technology behind it, we must first distinguish between the two primary ways we process information:  OLTP  and  OLAP . 1. OLTP OLTP (Online Transactional Processing)  is the DB used for company’s daily operations. Whenever you post on Reddit or place an order on Amazon, you are interacting with an OLTP database. It is triggered by the business's customer and handled by the backend server so that it can provide the business. Fast:  It handles millions of requests with millisecond response times. Small&Frequent:  It focuses on small, frequent transactions—usually just a few rows at a time. Write-intensive:  Its primary job is to record or update new data as it happens. 2. OLAP OLAP (Online Analytical Processing)  is the very technology we...
자세한 내용 보기

Leetcode

이미지
1. Two Sum: find a pair with specific sum   class Solution : def twoSum ( self , nums : List [ int ] , target : int ) - > List [ int ] : numMap = { } n = len ( nums ) # Build the hash table for i in range ( n ) : numMap [ nums [ i ] ] = i # Find the complement for i in range ( n ) : complement = target - nums [ i ] if complement in numMap and numMap [ complement ] != i : return [ i , numMap [ complement ] ] return [ ] # No solution found You might wonder how does the code deals with duplicates. As you can see, in the iteration of building the hash table, always the latter(with bigger index) will be saved in the hash table. In the latter iteration finding the complement, the former one(with smaller index) always goes through the iteration earlier, so the [i, numMap[complement]] can return two different indexes, carrying duplicate...
자세한 내용 보기