Home / user-546300

Mystic's questions

Martin Hope
Mystic
Asked: 2021-10-23 04:28:18 +0800 CST
  • 0

I have a VPN server with Intel X710 and Xeon E5-2630 v4. Debian 10. Based on accel-ppp, pptp protocol. Single SFP interface, two vlans. Problem - interrupts are not evenly distributed across queues. In the first queue, more than 10 times more interrupts come than the rest. Because of this, after running set_irq_affinity -x all ens15f0, my entire load falls on the first core.

  • I have tried updating the i40e drivers.
  • Tried using another script instead of set_irq_affinity - , with it the same result, just not the 1st core, but 2 and 20.
  • I tried to turn off offload: ethtool --offload ens15f0 rx off tx off sg off tso off gso off gro off lro off rxvlan off txvlan off rxhash off
  • Tried: ethtool -C ens15f0 adaptive-rx on adaptive-tx on

I reread half of the Internet, but the overall picture has not yet emerged. If anyone has a lot of experience with this, please point in the right direction. enter image description here

enter image description here

nic intel irq interrupts
Martin Hope
Mystic
Asked: 2021-10-06 08:54:52 +0800 CST
  • 1

A machine with debian 10. It has 3 interfaces with different ip. One of them is needed in rare cases, it has a static route to the management network (the rest of the routing works via ospf through another interface).
The problem is that if you pull the cable out of the network card, the interface is still up and the route remains. Can't this be automated? And I could not find anything on this question, maybe I am doing something wrong?
In addition, if I use auto or allow-hotplug, the interface is enabled after reboot even without a link, so I have to manually raise it if necessary.

networking debian interface
Martin Hope
Mystic
Asked: 2021-02-13 04:27:14 +0800 CST
  • 1

Some servers in the cluster have a lot of irrelevant processes in the process list. There are hundreds of them and some of them are about 2 years old.

root      9903  0.0  0.0  10752   180 ?        Ss    2019   0:00 bash -c /usr/local/bin/accel-cmd show stat | grep -C3 sessions | grep activ
root      9904  0.0  0.0   6172   148 ?        S     2019   0:00 /usr/local/bin/accel-cmd show stat
root      9905  0.0  0.0   7832   148 ?        S     2019   0:00 grep -C3 sessions
root      9906  0.0  0.0   7832   140 ?        S     2019   0:00 grep active
root      9907  0.0  0.0   5616   104 ?        S     2019   0:00 cut -d  -f4
root     10047  0.0  0.0   4084   112 ?        S     2020   0:00 run-parts /etc/ppp/ip-up.d --arg=ppp955 --arg=none --arg=0 --arg=10.x.x.x -
root     10051  0.0  0.0  17676   212 ?        S     2020   0:00 /bin/bash /etc/ppp/ip-up.d/dump ppp955 none 0 ...
root     10061  0.0  0.0  17676    56 ?        S     2020   0:00 /bin/bash /etc/ppp/ip-up.d/dump ppp955 none 0 ...
root     10063  0.0  0.0   6172   168 ?        S     2020   0:00 accel-cmd show sessions
root     10064  0.0  0.0   6304    32 ?        S     2020   0:00 grep 172.x.x.x
root     10065  0.0  0.0   4116     0 ?        S     2020   0:00 cut -f 2 -d | -s
root     10066  0.0  0.0   4120     0 ?        S     2020   0:00 tr -d
root     10082  0.0  0.0   4084    92 ?        S     2020   0:00 run-parts /etc/ppp/ip-up.d --arg=ppp1062 --arg=none --arg=0 --arg=10.x.x.x

I could not find information on similar cases or an explanation for this. This does not seem to affect the performance of the servers, but would still like to fix the problem.

ps