With 10.0.1 code, a health check that checks for http GET requests and receive “200 OK” passed with no problems. When the code was upgraded from 10.0.1 to 10.2.1, all servers that had this health check were marked down causing an unexpected outage.
I run the same health check on the F5 CLI by doing telnet to port 80 on the server and sending the same GET request and I get a “200 OK” response.
I am not sure what the problem is. I increased the timeout but the problem is not due to the timeout.
In 10.2.0, bigd behavior was modified to not append any \r\n’s to the end of the monitor send string for HTTP, HTTPS and TCP. If you add two \r\n’s to the end of the send string does the pool get marked up?
See SOL10655 for details:
sol10655: Change in Behavior: CR/LF characters appended to the HTTP monitor Send string
I played around with it adding and removing extra CR and LF but cannot bring up the pool. The only work around I have to bring up the pool and the web page is to remove the check for receive string. Tcpdump shows 200 OK string on the response and I have no idea why it fails. I modified the receive string to exactly match the case etc.
The same monitor works fine for the same servers from Big-IP running 9.4.7 code. When I run the same GET request on the Big-IP CLI (Linux OS), it works fine. It is just the TMOS version 10.2.1 that is having issues.
GET /login.asp HTTP/1.1\r\nHost: 10.22.0.196\r\nConnection: Close\r\n\r\n
And then start with a receive string of HTTP. If that works, then try 200 OK. If one or both don’t work, you can try the steps outlined here to troubleshoot the issue:
That wouldn’t be a valid HTTP request as the double \r\n would terminate the request and then you have another HTTP header appended. Can you try it like this:
GET /login.asp HTTP/1.1\r\nHost: 10.22.0.196\r\n\r\n
Did you try move the space after colon instead of before for Connection eg GET /exploit-mag-csg/ HTTP/1.1\r\nHost:bfc.saint-gobain.com\r\nConnection: Close\r\n\r\n