Question 98075 in Nginx Question posted in
The official documentation can be found here.

why does the nginx passive health check configuration is not working?

I’m using Nginx OSS nginx/1.20.1 version as a reverse proxy and load balancer. Behind the load balancer, there are two servers available where the jfrog artifactory web app service has been running.

I would like to configure the Nginx in such a way that whenever the web app service gets down, the Nginx should stop sending the request to the particular server.

For this purpose, I have added the passive health check parameters(fail_timeout=900s) in the conf file

upstream artifactory {
    server 172.1.1.1:8082  fail_timeout=900s;
    server 172.1.1.2:8082  fail_timeout=900s;
}

upstream artifactory-direct {
    server 172.1.1.1:8081  fail_timeout=900s;
    server 172.1.1.2:8081  fail_timeout=900s;
}

server {
    listen 80  default_server;
    server_name xxx.xxx.xxxx.net xxx.xxx.xxx.xxx.net;
    return 301 https://xxx.xxx.xxx.xxx.net$request_uri;

}

Then I have stopped the web service in one of the upstream servers and checked the Nginx error log, it is redirecting the request to both the upstream servers. Ideally, it should mark one of the servers as "unavailable" but it is doing like that. Could someone help with this, please?

Attached the nginx config files

Tags:

2

Answers


  1. 0

    This is indeed an expected behavior with Nginx open-source version. You can find more details about this on this page

    As an alternate: you can automate the process of validating the backend (Artifactory) availability with a simple health check REST API and update the Nginx configuration if there is an issue. Keep running the script in a crontab

    Login or Signup to reply.
  2. 0

    Here’s how Nginx works with the directive fail_timeout, assume fail_timeout is set to Td :

    • if an upstream server (say U0) fails to process and respond with number of requests (specified by max_fails) in Td seconds, Nginx will mark U0 as unavailable at the time T0.
    • From here, Nginx temporarily passes new requests to other upstream servers except U0 for Td seconds, until the timeout happenes (at the time T0 + Td), then Nginx retries connecting and sending a request to U0:
      • if U0 is recovered and able to handle request, then mark it as active
      • otherwise Nginx repeats the same behavior, distributes requests to other upstream servers except U0 until next timeout event.

    To see whether Nginx marks any upstream server as unavailable, make sure the logging level is warn (or below that e.g. info, debug) so Nginx will dump appropriate detail to error log.

    Login or Signup to reply.
Please signup or login to give your own answer.

Back To Top
Search