Anyone use a 10 GIG Network for 800xa ? Is it supported ?
We have a customer when using 10 GIG network with 800xa RNRP always gives a Suspected Network Loop when they Copy Files. 800xa 6.1.
When they go to a 1 GIG network no issues at all. There are no Physical Loops everything has been checked including RSS. It seems this fast Network
it's like too fast for RNRP and it gets Multiple Messages and thinks its a Network Loop. Yet 1 Gig is fine. Fault Tracer has no Errors. Yet during copy of
files Suspect Network Loop in RNRP Logs and Both Path 0 and 1 Go Down.
All the Documentation discusses up to 1 GIG.
Is 10 GIG Supported for 800xa 6.1 ? I cannot find this anywhere.
Has anyone used 10 GIG without any RNRP issues ?
When they go to a 1 GIG network no issues at all. There are no Physical Loops everything has been checked including RSS. It seems this fast Network
it's like too fast for RNRP and it gets Multiple Messages and thinks its a Network Loop. Yet 1 Gig is fine. Fault Tracer has no Errors. Yet during copy of
files Suspect Network Loop in RNRP Logs and Both Path 0 and 1 Go Down.
All the Documentation discusses up to 1 GIG.
Is 10 GIG Supported for 800xa 6.1 ? I cannot find this anywhere.
Has anyone used 10 GIG without any RNRP issues ?
Voted best answer
RSS = rappid spanning tree? That can cause loop detected problems.
Answers
The only reason for RNRP to detect loop is if it receives the exact same routing message more than (I believe) 13 times in a row.
Each RNRP local area routing message ends with a sequence counter with a value ramping between 0 and 255. You can actually monitor this counter value if you use Wireshark to capture UDP port 2423 traffic. The last byte of the hello message should never be the same between two routing messages. If you see such, the network or driver is likely at fault.
Maybe the 10G NIC driver is leaking duplicated multicast messages to the RNRP service?
Or maybe RSS is mandatory ON (with no means to turn it off) to improve the handling of the immense amounts of traffic a 10G interface may create?
RSS enables optimizing network traffic on multiprocessor archictecture. It is an accelerator which assist a computer to handle network traffic.
Even though the network manual states the opposite, see if you can enable Flow Control on the switch port and computer NIC and what effect it may have on the loop problem (Ethernet flow control allow throttling of traffic to avoid congestion).
Windows File Copy will claim and use all available NICs between the hosts taking part in the copying. With redundant network, "double speed" will be seen when copying. It would be interesting to know if you get dual loops reported also if you disconnect the remote peer's secondary interface while copying (forcing the copy to use single LAN only).
And finally, no hardware on the ABB Arrow Portal (login account required) have 10G NICs (ABB stopped maintaining a "System 800xA Certified HW" list some time ago).
Each RNRP local area routing message ends with a sequence counter with a value ramping between 0 and 255. You can actually monitor this counter value if you use Wireshark to capture UDP port 2423 traffic. The last byte of the hello message should never be the same between two routing messages. If you see such, the network or driver is likely at fault.
Maybe the 10G NIC driver is leaking duplicated multicast messages to the RNRP service?
Or maybe RSS is mandatory ON (with no means to turn it off) to improve the handling of the immense amounts of traffic a 10G interface may create?
RSS enables optimizing network traffic on multiprocessor archictecture. It is an accelerator which assist a computer to handle network traffic.
Even though the network manual states the opposite, see if you can enable Flow Control on the switch port and computer NIC and what effect it may have on the loop problem (Ethernet flow control allow throttling of traffic to avoid congestion).
Windows File Copy will claim and use all available NICs between the hosts taking part in the copying. With redundant network, "double speed" will be seen when copying. It would be interesting to know if you get dual loops reported also if you disconnect the remote peer's secondary interface while copying (forcing the copy to use single LAN only).
And finally, no hardware on the ABB Arrow Portal (login account required) have 10G NICs (ABB stopped maintaining a "System 800xA Certified HW" list some time ago).
Add new comment