Issue is solved (I hope).
There was a bottleneck in one of my configurations - hAP AC lite (RB952Ui-5ac2nD) was connected via 5GHz WiFi to RB4011. Placement is shown at this picture. Between them there are 2 brick walls. hAP AC lite has RTSP 4k camera connected that constantly sends traffic.
During monitoring I've found that "Tx/Rx CCQ" is on average 60% with quite often drops to 20-30%. At place of hAP there is quite crowded place from neighbors. So my first test was to switch WiFi bridge to use 2GHz instead of 5GHz. Then I changed channel width to 20MHz and CCQ increased to 70% with drops to 40%. During those drops if there 1 or more calls at same time there were some short 'hungs' of traffic flow (monitored it on separate PC via LAN connection) for 0.1-2 seconds. I assume there is some 'bug' in processing of breaking radio connection and traffic on RB4011 that produces 'deadlock' in queue processing as at same time sometimes there were 'spike' of CPU usage from 3-4% average to 20%.
As I have old RB2011 laying around I added it as dedicated WiFi bridge. Now RB2011 connected to RB4011 via LAN cable on 1Gb ports. On RB2011 nv2 bridge configured. On hAP AC lite nv2 station WDS configured to connect to RB2011. Both use 20MHz channel width and less crowded 2GHz frequence 2432. On this setup I see Tx/Rx CCQ on average 85% with drops to 65%.
And no issues on RB4011 during high load speed tests and many parallel calls on several devices. Will still monitor it to make sure there are no issues. Also speedtest showing 350Mb download and 550Mb upload. Previously there were 100-200Mb/200-300Mb with some 'hangs' during test.
Thank you.
There was a bottleneck in one of my configurations - hAP AC lite (RB952Ui-5ac2nD) was connected via 5GHz WiFi to RB4011. Placement is shown at this picture. Between them there are 2 brick walls. hAP AC lite has RTSP 4k camera connected that constantly sends traffic.
During monitoring I've found that "Tx/Rx CCQ" is on average 60% with quite often drops to 20-30%. At place of hAP there is quite crowded place from neighbors. So my first test was to switch WiFi bridge to use 2GHz instead of 5GHz. Then I changed channel width to 20MHz and CCQ increased to 70% with drops to 40%. During those drops if there 1 or more calls at same time there were some short 'hungs' of traffic flow (monitored it on separate PC via LAN connection) for 0.1-2 seconds. I assume there is some 'bug' in processing of breaking radio connection and traffic on RB4011 that produces 'deadlock' in queue processing as at same time sometimes there were 'spike' of CPU usage from 3-4% average to 20%.
As I have old RB2011 laying around I added it as dedicated WiFi bridge. Now RB2011 connected to RB4011 via LAN cable on 1Gb ports. On RB2011 nv2 bridge configured. On hAP AC lite nv2 station WDS configured to connect to RB2011. Both use 20MHz channel width and less crowded 2GHz frequence 2432. On this setup I see Tx/Rx CCQ on average 85% with drops to 65%.
And no issues on RB4011 during high load speed tests and many parallel calls on several devices. Will still monitor it to make sure there are no issues. Also speedtest showing 350Mb download and 550Mb upload. Previously there were 100-200Mb/200-300Mb with some 'hangs' during test.
Thank you.
Statistics: Posted by KAjFASH — Mon Jan 15, 2024 10:39 pm