MarkLuun Posted January 15, 2021 Posted January 15, 2021 (edited) Hi Armbians I am running mulitiple TOR network relays on Cubietruck boards (Cubieboard 3) in different Datacenters with synchronous 100 MBit/s internet connection each board (by ethernet plug) for 1- 1,5 years now. At the beginning each board could handle 25 MBit/s incomming and outgoing traffic the same time easy and 24/7. The boards are running the latest Armbian build with kernel headers installed and full firmware package. But now, as you can see here, my Cubietrucks are getting slower after time, with the network traffic they can handle via eth0: https://metrics.torproject.org/rs.html#details/EC9C4CDB403204731122C9D3FA44DFAE667DC82D They have been kept up to date and are rebootet and the system cleaned once a week. So whats the problem? SD card getting old and dieing? Is there a way to speed them up again and (even better) tune the throughput without risking stability? Please help Edited January 15, 2021 by zahiir Typing error
_r9 Posted January 15, 2021 Posted January 15, 2021 Providing logs with armbianmonitor -u helps with troubleshooting and significantly raises chances that issue gets addressed.
MarkLuun Posted January 15, 2021 Author Posted January 15, 2021 (edited) Done! http://ix.io/2M8m one more: http://ix.io/2M8p Edited January 15, 2021 by zahiir additional info
Igor Posted January 16, 2021 Posted January 16, 2021 17 hours ago, zahiir said: the latest Armbian build with kernel headers installed and full firmware package. According to your logs, not really: It's not very well tested upgrade path - if you are coming from some old build, a clean install is recommended. But that solo doesn't explain this odd behaviour. I am having one Cubietruck in the quality assurance lab and it performs within expectations, #28 https://users.armbian.com/igorp/2021-01-14_23.53.13.html Perhaps firewall rules are responsible for speed downgrade? They eat CPU and if there is a lot on CPU, network speed goes down. Allwinner A20 is not very powerful device.
MarkLuun Posted January 16, 2021 Author Posted January 16, 2021 Hmm, I would wonder if the firewall is the problem. I only set 6 single IPv4 port rules via UFW. I have CT boards running with 15-20 IPv4 port rules plus the same in IPv6 rules that do not have this issue. The issue also does not affect all boards, just about 1/3 of them. The others are doing their job as allways! But the CPU could be a point to look at. I also addet a 2048MB SWAP partition with dphys-swapfile and set these parameters in systctl.conf for it: vm.swappiness=10 vm.vfs_cache_pressure=50 Maybe this could be a possible point to find the bug. I will try your fix above today and upgrade the system and kernel. Thanks for that! A fresh OS install would be the best choise, of course. But this may be a time consuming process to completely setup more then 20 relays from zero again. I will try the above fix first...
Igor Posted January 16, 2021 Posted January 16, 2021 26 minutes ago, zahiir said: A fresh OS install would be the best choise, of course. Absolutely. And no messing up with SWAP https://forum.armbian.com/search/?q=zram&fromCSE=1 28 minutes ago, zahiir said: But this may be a time consuming process to completely setup more then 20 relays from zero again. https://www.ansible.com/
Solution MarkLuun Posted January 20, 2021 Author Solution Posted January 20, 2021 After setting the parameter „vm.swappiness=0“ and deleting any not absolutely necessary UFW parameter (also the IPv6 ones) I set the network connection parameters again in „armbian-config“. Now the traffic, passing my relays, seems to go back to a normal value again. So I would say this problem is solved! Thanks a lot!
MarkLuun Posted January 22, 2021 Author Posted January 22, 2021 But how do I enable the last resorts of the CT and tune the network throughput to the max.? Is there a way???
Recommended Posts