Hi.
Can anyone tell me what it is? Intel server, xeon processor, integrated 2 port network card (ethernet), and 3 intel x520 DA2 cards. 4 BGP sessions running and that’s it.
Almost all the time, even with low traffic ~ 100Mbps, one cpu goes to 100%, and this machine has practically nothing to do. There are a lot of unknown devices and I don’t know where from.
Please help.
The naming of devices has nothing to do with their proper functioning in linux in general and in ROS in particular.
For naming linux uses PCI IDs database, available from https://pci-ids.ucw.cz … you can browse the database by ID and you’ll see that device from your screenshot is “Xeon E7 v3/Xeon E5 v3/Core i7 PCI Express Root Port 3” which usually doesn’t need any driver to work properly.
For proper functioning a proper driver is usually needed (and driver identifies devices it can drive by numerical values so it doesn’t need PCI IDs database). Many times ROS doesn’t include driver for a device which is properly identified (and named). It is possible (not under ROS) that there is driver available for a device missing from locally available PCI IDs database.
As to the low performance of your router: run CPU profile and check what kind of process(es) are CPU-intensive and see if that can be optimized. Generally single connections (e.g. single thread speedtest or single thread iperf test) are bound to single CPU core and it’s generally not possible to achieve high throughputs. Should be able to get more than 100Mbps though.
CPU Profile looks like this:
Any sugestions ?
Anyone ???
How many firewall rules do you have?
Is there something particularly complex about your setup?
Have a look in the log files, anything of interest?
10 rules in queue, connection tracking disabled, maybe 15 rules in firewall , mostly drop from outside like dns , winbox.
2 full feed bgp sessions and 3 to ix.
uptime: 2d2h24m3s
version: 6.48.3 (stable)
build-time: May/25/2021 06:09:45
free-memory: 14.6GiB
total-memory: 15.6GiB
cpu: Intel(R)
cpu-count: 12
cpu-frequency: 1895MHz
cpu-load: 10%
free-hdd-space: 110.8GiB
total-hdd-space: 110.9GiB
write-sect-since-reboot: 807256
write-sect-total: 807256
architecture-name: x86_64
board-name: x86
platform: MikroTik
Traffic summary: 850Mbps / 70Mbps DL/UL
In logs nothing special, some drops logged.
This is Server, 2U INTEL, Xeon CPU, 16GB RAM. 3 card Intel x520-DA2 and integrated onboard 2x 1GE.
That CPU profile is that taken when your speed testing through that device out to the internet?
Or is that idle?
No.
The CPU usage level is still the same, it increases a little as traffic goes higher on the network.
hi,
please post your config here and obfuscate / hide sensitive information so that others can see it and might help you in the process
Ok , main config is in attachments, bgp, filters, ip address.
res_pci.txt (14.9 KB)
system.txt (619 Bytes)
bgp.rsc (2.49 KB)
filtry.rsc (2.32 KB)
Maybe I’m blind because I’ve been sick for a week, but I can’t see the main config.
This is config , 4 files, most important things.
Whilst I’m not sure the CPU load comes from some configuration issue, in general, there is no such thing as “important part of configuration” when it comes to analysis of an unexpected behaviour. Typically, the root cause of that behaviour is in the part of the configuration you do not expect to be relevant. So post the complete one, obfuscating only public IPs, AS numbers, passwords, and user names/e-mail addresses/domain names that might tell too much about you, but don’t let the obfuscation break the structure of the configuration (e.g. prefixes of all IP addresses in the same subnet must be changed to the same value).
Can you say something based on this setup i uploaded? It’s a waste of time to fix the config file. This situation is happening gradually, in the beginning it was ok, and the configuration has not been changed since the beginning of the server operation.