Wireless Access

We are using our 2 Aruba controllers to redirect users to an external portal. We service up to 50,000 unique clients per day so I'm guessing some manual config of the internal web-server could be usefull.

Since we're using 2x 7220 controllers  the web-max-clients defaults to 75 concurrent clients.

When I pull up the statistics it seems I never have any available connection slots. Regardless of this, the portal  redirect doesn't appear slow or anything.

show web-server st

Web Server Statistics:
----------------------

Current Request Rate:            23 Req/Sec
Current Traffic Rate:            2 KB/Sec
Busy Connection Slots:           75
Available Connection Slots:      0
Total Requests Since Up Time:    43762818
Total Traffic Since Up Time:     15443592 KB
Avg. Request Rate Since Up Time: 3 Req/Sec
Avg. Traffic Rate Since Up Time: 1311 Bytes/Sec
Server Scoreboard RRRWRRWRCCRCRRRCRCRRRRRCRCCRRCCCCCCRRWRRRRCRRRRCCRCRCRRRRCRRCRRRCRWRRRCRRRC

Scoreboard Key: _ - Waiting for Connection, s - Starting up
                R - Reading Request, W - Sending Reply
                K - Keepalive, D - DNS Lookup
                C - Closing connection, L - Logging
                G - Gracefully finishing, I - Idle cleanup of worker
                . - Open slot with no current process

Since cpuload seems minimal I'm thinking about increasing the web-max-clients. Is there any reason not to do this as long as cpuload isn't an issue?

Second question, the external portal guys are complaining they get a ton of requests without the max address in the url. 

I vaguely remember that the controller treats certain traffic differently to lower cpu load. I cannot seem to find anything on this to corroborate this though. Can anybody confirm this/point me in the right direction on this?

hey Koen

No reason not to increase it, if your cpu is happy then have at it. I know of a couple of deployments running captive portal in airport terminals with the max sessions set to 320 (max) on 7240 hardware (be sure to be on recent s/w though in 6.5.1.x or higher, there are some httpd stability issues at very high loads in older 6.x versions). You might want to do it piecemeal, try 150 first and then work your way up. For the above mentioned network on 320 sessions, it's consuming 110 million plus CP requests in 30 days, and has a typical load of 60-70% CPU solely in httpd. 

For the different traffic treatment, maybe you're talking about this:

https://community.arubanetworks.com/t5/Wireless-Access/Enable-bypass-captive-portal-landing-page/td-p/358312

It shouldnt be causing the CP redirection URL to be malformed though, it should just prevent the initial redirection at the client side for non browser user agents

hth.

Thanks, already increased the number of connections to max 320.

Still uses up all connections at peak times but no noticable impact on cpu load whatsoever.

That post of you was indeed what I had in mind, thanks for referencing it again. Indeed not the behaviour we're seeing.

Portal people reported getting tons of requests with user-agent 'outlook'. These arrive at the external portal without any of the MAC-addres, ap-group and such info that other requests do carry with them.

Can't explain why but luckily it doesn't seem to matter as these aren't actual browsers apparently.