I am seeing weird issue with one of my RB911G-5HPnD.
When I run Bandwith test from the device to another device connected via LAN, immediately after Btest starts the Rx CCQ drops from 100% to 5%. As soon as the Btest is stopped, CCQ returns to 100%. The Tx bandwith is very low compared to another 911G connected to same switch. Same target and the difference is huge (10-15Mbps Tx on affected device, 90Mbps from another device to same target).
I guess its the only 911G which is using 40MHz channel single chain with nstreme. I am seeing this CCQ drop also when there is high traffic and CPU goes above 80%…
I know it is, normally not doing this, but I am trying to diagnose a problem I am seeing now - low tcp performance between devices connected to same Cisco switch… getting around 30Mbps with single TCP connection even I am sure there is no problem to max out the 100Mbps with devices using the 400MHz CPU.
And when I tried on the 911 I realized the CCQ went suddenly down to 3%. I have replaced this device with a new spare NetMetal 5, configured from scratch, upgraded to 6.22 but the problem is still there
I am pretty sure even I know that the Btest is CPU intensive, I have never seen such behaviour which would lead to CCQ drop caused by either Btest or high CPU. Sometimes mostly on v6 devices I see SNMP causing high CPU usage but it never affected CCQ of wireless card.
for the 911 I have used the standard 24V 0,8A (Cable is outdoor shielded carrier class Ubi Level 2 - 12m), when swapped the NetMetal used the original 24V 1,2A and installed new cable, this time its Ubi outdoor Level 1. I dont think it could be a power issue…
Tested more and when I switch to NV2 from nstreme its similar, the CCQ drops to 40% but doesnt go back up until the link goes down/up… Very strange..
Upgrade to 6.23rc7 as advised by support made thing even worse. Flash access is taking whole CPU time, link disconnecting etc.
Downgrade to 6.19 - FIXED! Btest pushing 200Mbps to a neighbor connected to same switch without affecting the CCQ… My GOD…
UPDATE: Just one thing not right with 6.19 - 20-30% CPU Usage by unclassified services which apparently has been solved in 6.22… But I cannot use that version as it has the problem described above