Flexlink+ on C9400?

[RB6502]

So I recently learned that Flex Links have been deprecated starting with early IOS-XE 16.x, but that there's a new feature called Flexlink+ that's available as of IOS-XE 16.12.  When I found the config guide chapter for this feature, the config statements of interest are really REP statements.  It's really a pretty short chapter with not a lot of explanation behind the Flexlink+ feature.

So what I'm wondering is:

• how close Flexlink+ compares feature/performance-wise to the old Flex Link?  I'm not really familiar with REP and am reading up on it, but the reason I was looking at Flex Links was to avoid STP, but still allow for alternate path options with fast convergence. 
• Is Flexlink+ convergence on the same order as Flex Link? 
• With Flex link, it was possible to achieve concurrent use of the links by splitting up the VLANs to be carried by the trunk across the two Flex Links, with say half as primary for one port and the other half as primary for the other port of the Flex Link.  Is this possible with Flexlink+?
• Is there still no configuration necessary on the "other end" (the uplink switch's ports)? 
• Can STP be active on other ports/VLANs of the uplink and/or downlink switches without causing problems?

Only asking these questions here as I can find next to no other information on Flexlink+.

[pieterh]

I think the 9400 has other options for redundancy links

you can combine two uplink 9400's in a stackwise virtual configuration

for the connected switch both links can be member of a normal etherchannel

this way no need to for flex-link and/or spanning-tree

[RB6502]

Maybe, but I wasn't asking about alternatives, I was asking specifically about the Flexlink+ feature in order to understand it better.  No feature is perfect, there are always trade-offs :)

[pieterh]

ok no problem,

does this document help ?

Resilient Ethernet Protocol (REP) is a Cisco proprietary protocol that provides an alternative to the Spanning Tree Protocol (STP). REP provides fast reconvergence and traffic recovery when a link in a ring fails. Fast reconvergence is achieved by notifying all nodes in the ring with a control message that activates the redundant path to the core network.

and CiscoLive!

FLexlink+ may just be another name for REP?

Table 1. Feature Information for Flexlink+

Feature name	Releases	Feature Information
Flexlink+	Cisco IOS XE Gibraltar 16.12.1	The feature was introduced.

[RB6502]

Thanks, I have read that chapter, but that only raises more questions.  Since REP is a loop topology protocol, I don't understand the applicability to Flex Links/Flexlink+, each of which are star topology things.

[camihern · Accepted solution]

Hi!

REP is indeed a loop control protocol, primarily, but the configuration suggested uses only REP edge no-neighbor type of ports.

A REP edge no-neighbor port assumes whatever is connected to it is not a REP capable device. So, for practical purposes, it "disables" the loop control capabilities. REP also disables STP on the REP ports, and there is no configuration change required on the devices connected to it (they could be configured as STP portfast to enhance convergence times). This is usually done when you need to integrated a REP ring with an external network (STP).

Considering this, the Flexlink+ example uses a REP ring of a single switch, with this switch having both the Primary and Secondary edge ports, which in Flexlink terms will be the Active and backup ports, respectively.

The REP switch will have all VLANs allowed on Primary edge port, while the Secondary edge port will be in alternate/blocking state, for example:

AM270NP3102#show rep topology
REP Segment 1
BridgeName PortName Edge Role
-------------------------------- ---------- ---- ----
AM270NP3102 Gi1/13 Pri Open
AM270NP3102 Gi1/12 Sec Alt

If the Primary edge port fail, the Secondary edge port starts forwarding traffic. This achieves the same as Flexlink.

Convergence time for REP varies between platforms and depends on the size of the ring, but is said to be between 10 - 90 msecs. This being a REP ring of a single switch the convergence should be minimal as no exchange of REP LSL packets via the network happens.

VLAN load balancing is possible in REP too. Let's say you have VLANs 10 and 20. You want VLAN 10 to be forwarded out of the Primary edge port and VLAN 20 out of the Secondary edge port.

On the Primary edge-port, configure VLB to block VLAN 10 on the Secondary port with the following command:

Device(config)# interface x/x
Device(config-if)# switchport trunk allowed vlan 10,20
Device(config-if)# switchport mode trunk
Device(config-if)# rep segment 1023 edge no-neighbor primary
Device(config-if)# rep block port -1 vlan 10

This means, block VLAN 10 on the Secondary port (-1). Once this command is configured, the rest of VLANs not specified on the list will automatically be blocked on the Primary port (1). The result is only VLAN 10 is forwarding in Primary port, and only VLAN 20 is forwarding on Secondary port. A "show interface trunk" should show that.

VLB is not contemplated on the Flexlink+ document, and there is also no mention of the REP admin VLAN which is vital for VLB and preemption features. So will need to check if this is officially supported, but in good theory you should be able to load balance between REP ring ports even if it is just one switch.

Hope this helps.

[pieterh]

Flex Links
Flex Links are a pair of a Layer 2 interfaces (switch ports or port channels) where one interface is configured to act as a backup to the other. The feature provides an alternative solution to the Spanning Tree Protocol (STP). Users can disable STP and still retain basic link redundancy. Flex Links are typically configured in service provider or enterprise networks where customers do not want to run STP on the switch. If the switch is running STP, Flex Links is not necessary because STP already provides link-level redundancy or backup

=> from this text, flexlinks is not really a star-topology its just an interface pair