动态配置的upstream有一定的几率会使用系统的resolve

millken · 2016-08-29T21:02:36+00:00

Hello! 2016-08-29 1:59 GMT-07:00 Jinhua Luo: > 找到原因了，我在balancer_by_lua里头写&#x...

动态配置的upstream有一定的几率会使用系统的resolve

millken

   server {
        listen 84 reuseport;
        set $ups "default";
        set $ups_scheme "http";
        set $access_log_path "/tmp/access.log";
        access_log $access_log_path compression;

        location / {
            proxy_http_version 1.1;
            proxy_set_header Connection "";
            proxy_pass $ups_scheme://$ups;
 ....

ng启动从数据库加载vhost配置，同时dyups.update(ups_key, ups_value)。
这里的ups_key形如ups_1.com
这个过程没错，测试也都正常，但是中间偶尔会有几次，upstream不是走的动态配置，而是走了resolve。
日志如下：

2016/08/19 17:46:11 [error] 29738#29738: recv() failed (111: Connection refused) while resolving, resolver: 127.0.0.1:53
2016/08/19 17:46:16 [error] 29738#29738: recv() failed (111: Connection refused) while resolving, resolver: 127.0.0.1:53
2016/08/19 17:46:21 [error] 29738#29738: recv() failed (111: Connection refused) while resolving, resolver: 127.0.0.1:53
2016/08/19 17:46:26 [error] 29738#29738: recv() failed (111: Connection refused) while resolving, resolver: 127.0.0.1:53
2016/08/19 17:46:31 [error] 29738#29738: recv() failed (111: Connection refused) while resolving, resolver: 127.0.0.1:53
2016/08/19 17:46:36 [error] 29738#29738: recv() failed (111: Connection refused) while resolving, resolver: 127.0.0.1:53
2016/08/19 17:46:41 [error] 29738#29738: *147 ups_1.com could not be resolved (110: Operation timed out), client: 127.0.0.1, server: , request: "GET /a HTTP/1.1", host: "1.com:84"

造成的现象：返回503，或卡在那里。

提了个issue在：https://github.com/yzprofile/ngx_http_dyups_module/issues/64

但不确定是nginx本身的问题，还是dyups模块问题。