Hello,
I have a machine which I want to boot via pxe/tftp. The machine is behind a CRS106.
First of all, it needs more time to get the DHCP/PXE done (15 sec. vs. 2 sec if I use a netgear switch), after this getting the kernel and initrd via tftp takes also longer as if I use a stupid netgear switch (15 sec. vs. 25 sec.).
The CSR106 is running with the default config, so all ports are bridged.
Any ideas whats wrong here? Did I miss something?
Regards
Daniel
Could it be that CRS needs longer time to autonegotiate ethernet configuration (speed, duplex)? It would explain both initial delay as well as the later one (when kernel driver re-initializes network interface). Stupid switches are usually much faster at it.
You can verify that by inspecting log on CRS … interface downs and ups are logged…
I’d expect it is the xSTP learning phase on the port after it comes physically up what prevents it from forwarding traffic, unless the port is manually configured as edge one (which is not the default).
Yes, I see some ups and downs.
Ok, that would explain the delay when getting the IP via DHCP. But what about the download of Kernel and initrd via tftp?
Regards
Daniel
When OS (whatever it is) starts NIC driver, driver often restarts the interface. Then it’ll be down/up and DHCP procedure again.
And whenever down/up happens, also what @sindy mentioned happens …
Set the port as edge one (/interface bridge port set [find interface=etherX] edge=yes) and try the boot again to see the total boot time.