How do I do zero downtime restart go service?

The practice of graceful

Use the endless library to implement, such as Access gin:

r: = gin. Default () r.get ("/", index) endless. Listenandserve (":", R)

We write a handler function with a specified time-out:

Func index (c *gin. Context) {    Duration:= c.query ("duration")    Durationint, _:=  StrConv. Atoi (Duration)    * Time . Second)    c.string (http. Statusok, duration)}

Curl http://localhost:3000/?duration=10

When using CTRL + C at the time of the test, all requests will be processed to exit, if it is run in the background, when we get the PID of the process, if used kill -9 $PID , endless cannot catch the signal, need to usekill -s TERM $PID

Using Nginx upstream Load

Simple configuration of the Nignx:

Upstream graceful{Server127.0.0.1:8888; Server127.0.0.1:8889;} server {Listen8001;    server_name localhost; Location^~ /{proxy_set_header X-real-IP $remote _addr;        Proxy_set_header Host $host;        Proxy_set_header http_referer $http _referer; Proxy_set_header X-forwarded-For $proxy _add_x_forwarded_for; Proxy_pass http://graceful/;Proxy_redirectdefault; } error_page - 502 503 504/50x.html; Location= /50x.html {root html; } Location~ /\.. +{deny all; }}

To make a simpler load-balancing, update the program in order to update, when the fact that a service is no longer available, the load to another, after the sequential update, to meet the needs of service hot update, of course, it is best to use a dedicated deployment script to achieve.

How do I do zero downtime restart go service?