@Voxel Thanks.
I've used telnet for the last weeks cause I just didn't have enough time to configure ssh access. Now I've set up ssh and logged in and it's a different story. I've always used ssh tunneling to access routers and internal LAN network devices. Is the ssh configuration going to survive a firmware upgrade. View attachment 31464
My experience has been that a firmware upgrade will wipe out the ssh configuration. It takes a few seconds to manually recreate.
Another good alternative to PuTTY is Git Bash, included in the Git software downloadable at https://git-scm.com/download/win
Git Bash provides UNIX command-line tools, including SSH
@Voxel Thanks, I have a backup of the public and private keys.
Can you or anyone else explain the %CPU0 and %CPU1 values before and after the slash symbol. Are these readings for the dual-core Qualcomm® Krait™ CPU (1.7 GHz) load only or they show the load of the dual-core 800 MHz Network Subsystem (NSS) Accelerator either.
I still try to understand why sometimes I have relatively high CPUs load (Mostly the value/s after the slash are high. Maybe these are for the NSS?). That high load is observed mostly when I have a torrent client program running (idling) on one Windows 10 PC. The top-procps-ng doesn't show the process responsible for the high load. Maybe the top-procps-ng doesn't show the NSS load.
Can you or anyone else explain the %CPU0 and %CPU1 values before and after the slash symbol. Are these readings for the dual-core Qualcomm® Krait™ CPU (1.7 GHz) load only or they show the load of the dual-core 800 MHz Network Subsystem (NSS) Accelerator either.
I do not quite understand you guys... Nothing new. This top is usual for me so it is why I added it to firmware. I am using Debian to compile firmware and its top is the same:
The same is with 'ps-procps-ng' vs 'ps' from busybox. I type the command ('ps') automatically, so just automatically I type:
Code:
ps aux
instead of just 'ps'. And I see more advanced:
So it is why my /root/.profile is:
Code:
#!/bin/sh
alias ps='/usr/bin/ps-procps-ng'
alias top='/usr/bin/top-procps-ng'
export PATH=/opt/bin:/opt/sbin:/bin:/sbin:/usr/bin:/usr/sbin
'alias' is very useful. When now I type 'top' I have: '/usr/bin/top-procps-ng'. When 'ps aux' I have '/usr/bin/top-procps-ng aux'
P.S. This 'top-procps-ng' is very useful. For example when I press 'Shift-M' I see now sort by memory load. Hit 'h' for a help when running to see the list of commands...
And with the Kamoj 5.4beta add-on you get "ps2" and "top2" as well.
Let's see if you can find out what they do ;-)
PS
To override alias, prefix the command with a backslash, e.g.: \ps or \top
No complaints here.
I'm using Debian too for compilation of OpenWRT firmwares. I was just trying to understand why there are no relevant values for CPU load on the rows below the %CPU0/CPU1 when the CPU0/1 values on top were around 40-50%. I thought if it is possible to monitor NSS load too. Obviously it isn't. @kamoj ps2 and top2 doesn't work from telnet session in case anyone tries to run them. Via SSH both run as intended. I have configured SSH session that runs top2 immediately after logging with private key from Pageant.
To all interested users and developers willing to help resolve this. @Voxel@kamoj@HELLO_wORLD
I continue this thread with my investigations of the high CPU load of R7800.
After spending over a week trying to figure out the reason (really tricky) for the high CPU usage I think I finally made it.
The situation was described in my previous posts here, here, here and several other posts here.
Disregard all my previous doubts and probable reasons (torrents, DNSCrypt, etc.) I've suspected before.
I have the following network setup. First LAN port of the router is bridged (for IPTV using VLAN and LED blinks in orange). A LAN cable is going to a switch where I have two IPTV set-top boxes connected. I have a PC connected to LAN port 4.
I usually have this normal CPU load when both IPTV set-top boxes are turned off.
The picture changes to this when one or both IPTV set-top boxes are turned on. Below is the minimum load that I was able to capture.
Next depending on the combination of the TV channels being watched simultaneously (HDTV, FullHD or Standard resolution TV) there are different CPUs loads.
The worst case supposedly is when 2 FullHD TV channels work simultaneously. From the second picture with 21%/21% load and above I'm unable (always) to open Top command from @kamoj add-on.
With load close to 70-90% it becomes difficult/impossible to open the kamoj Information page too and all negative impacts on the work of the router reported earlier.
What I have tried so far.
First of all I've thought the bridging is somehow (due to a bug) not hardware offloaded by NSS. In the document here it's written that
NSS acceleration model
• Features
• Designed for Home Gateways (CPE)
• Flow detection based “All-or-nothing”
offload
• Acceleration supports:
• IPv4, IPv6, NAT, PPPoE, L2TP, VLAN, Qdisc
• Functional behavior
• 0% cpu load seen in Linux
• Keep Linux counters up to date
• Does not require functional changes at an
upper level (user space)
That's why I decided to connect my PC to the first LAN port (disconnecting the cable to the switch) and see what would happen. My Internet provider gives five public IPs that I can use. The PC got a public IP from the ISP via DHCP on the bridged port (now the LED blinks in white). Then I've run several speed tests while I was monitoring the load on my Phone with ConnectBot (my connection is 512Mbps down/150Mbps up) but the CPU load remained low as on the first picture.
The bitrate for one IPTV set-top box is fairly low - maximum of 10Mbps per HDTV channel and only 4Mbps for standard resolution channel (ISP uses HEVC H.265 for data compression).
Second, I've tried to disable DNSCrypt and later Aegis @HELLO_wORLD (thinking it may interfere somehow). With these disabled I have the same high CPU usage when IPTV set-top box/es work.
I still didn't have enough time to try the next things that are in my mind.
I hope someone has good suggestions or maybe similar setup with IPTV (VLAN) and can make same experiments.
Sorry for this long post but using the router this way is not very healthy.
First, nice job narrowing it down to the IPTV boxes. That is a progress, and was tricky.
How is your WAN setup? Do you have a modem or are you directly connected to the ISP with the R7800?
The system load shows that for some reason the data bridged to your IPTV boxes are not fully accelerated. Something is hitting the kernel hard. Maybe installing netdata via Entware could help a little to narrow this further down.
I hope someone has good suggestions or maybe similar setup with IPTV (VLAN) and can make same experiments.
Sorry for this long post but using the router this way is not very healthy.
Guys really fast and short update. I've enabled the IGMP Proxy and now no more high CPU usage even with working IPTV. I will give more details later.
Thanks a lot for your help.
SNBForums is a community for anyone who wants to learn about or discuss the latest in wireless routers, network storage and the ins and outs of building and maintaining a small network.
If you'd like to post a question, simply register and have at it!
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
