Forum Discussion
ringoseagull_77
Nimbostratus
NimbostratusHealth monitoring when using port 0 for persistence
I'm using session persistence across virtual servers with backend HTTPS re-encryption using port 0 on pool members.
Is there way of configuring a health monitor on those pool members?
I tried using icmp gateway, but then of course found that when the web server was stopped the icmp check still succeeded so the member was not dropped from the pool and users were redirected to the fallback "Site not found" error page.
6 Replies
hoolioCirrostratus
Yes, you can hard code the destination port to be checked to 443 using the alias port option on an HTTPS monitor.
Aaron
ringoseagull_77Thanks Aaron, I've hit some issues trying to get this to work.
When I apply the new https monitor with 443 as the alias the F5 reports the node as down.
This may be because we haven't got a proper SSL cert yet, so when you browse to the URL you have to add an exception to get round the invalid certificate. Will this also work if I applied the same change to an http monitor and set the alias to port 80?
I also wonder if I have a problem with the monitor send string.
The old monitor worked with:
GET / HTTP/1.1\r\nHost: \r\nConnection: Close\r\n
I tried this without success, and then a string adapted from SOL2167:
GET / HTTP/1.1\r\nHost: moodle-test.city.ac.uk\r\nConnection: Close
This also showed the node to be down.- ringoseagull_77I've tested setting the port alias to 80 and this also reports the node as down. Not having much luck with this one.
- ringoseagull_77Double post.
- hoolioWhich LTM version are you running? bigd, the monitoring daemon, appends one or two \r\n's to the end of the monitor send string.
This looks like a good start for 9.4 - 10.1, assuming you've set the alias port to 443:
GET / HTTP/1.1\r\nHost: moodle-test.city.ac.uk\r\nConnection: Close
I added some testing suggestions for LTM monitors to this wiki page:
http://devcentral.f5.com/Wiki/default.aspx/AdvDesignConfig/TroubleshootingLtmMonitors.html
Aaron - ringoseagull_77Thanks for your help, the config was fine (I'm on 9.4.7), the monitoring was confused by my colleague restarting Apache without telling me :-)
