Bug #15262
open
Captive Portal Has High CPU Interrupts With Large Number of Users
Added by Kris Phillips about 1 year ago.
Updated 4 months ago.
Affected Plus Version:
23.09.1
Description
When 700+ Captive Portal users are in use, CPU interrupts will cause high load averages to occur. This can lead to connectivity problems, such as packet loss on WAN uplinks, webConfigurator responsiveness issues, etc.
Tested with a customer who had load averages of 14-16 with Captive Portal on with 1400+ users. Once Captive Portal was turned off, load averages dropped to 0.5.
Load seems higher for Captive Portal when there is significant numbers of users since the transition to pf from ipfw.
Customer in ticket 2947838007 is reportedly running into this issue as well.
- Status changed from New to Confirmed
Marking this as confirmed, as we have multiple instances of people reporting issues with high CPU usage with a high number of captive portal users.
Kris Phillips wrote in #note-4: [[
Marking this as confirmed, as we have multiple instances of people reporting issues with high CPU usage with a high number of captive portal users.
I reported this issue in the forums in February 2024. We have large deployments with 8,000 to 10,000 captive portal users. Version 2.6 continues to function perfectly with the captive portal; however, attempts to switch to 2.7.2 have twice resulted in the same issue: excessively high CPU load, with CPU core 0 consistently reaching 100% utilization while other cores remain available. This appears to be related to the switch from IPFW to PF.
https://forum.netgate.com/topic/184647/increase-php-fpm-listening-queue-not-working-in-2-7-0/5?_=1735573058510
Also available in: Atom
PDF