How to migrate BAGG from 4x10G to 2x40G and remain connectivity

[Dennis_HP-ATP]

Hello guys,

We are about to assign new up and downlinks (BAGG's), between our Datacenter core and DC access switches, but the challenging part is how to remain connectivity, while moving from 4x10G links (current situation) to 2x40G links (new situation). 

LACP cannot handle different speeds, so it looks like I will be having downtime, or is there a possibility to configure a LACP BAGG that can handle both 10 - and 40Gbps speeds? This would make my life much easier :-)

Specs are:

Datacenter core switch:

[UCDC1]dis dev manu
 Slot 1 CPU 0:
DEVICE_NAME          : FlexFabric 5930-4Slot JH179A

with software running

]dis boot-loader
Software images on slot 1:
Current software images:
  flash:/5930-cmw710-boot-r2418p06.bin
  flash:/5930-cmw710-system-r2418p06.bin
Main startup software images:
  flash:/5930-cmw710-boot-r2418p06.bin
  flash:/5930-cmw710-system-r2418p06.bin

Datacenter access layer switch

dis dev manu
 Slot 1 CPU 0:
DEVICE_NAME          : FlexFabric 5700-48G-4XG-2QSFP+ JG894A

running software:

dis boot-loader
Software images on slot 1:
Current software images:
  flash:/5700-cmw710-boot-r2418p06.bin
  flash:/5700-cmw710-system-r2418p06.bin
Main startup software images:
  flash:/5700-cmw710-boot-r2418p06.bin
  flash:/5700-cmw710-system-r2418p06.bin

Hopefully you can help me out with no or subsecond downtime for this one.

KR
Dennis

[HP-Browniee]

I don't think you can accomplish this without a little downtime. Even with spanning-tree you will have a recalculation.

[Vince-Whirlwind]

If you make sure your STP port priority is configured to prioritise the 10Gb links, you can get all your connectivity up and running, then switch STP priorities to flick over.
If you have STP configured correctly, you would miss at most one ping, so TCP traffic would be unaffected, and UDP would lose about 3 or 4 seconds.

[sdide · Accepted solution]

Hi,

I've done some testing. I used 2 x HPE-5130 24G SFP 4SFP+ EI (R3116P05)

I made two BAGGs  on each switch ( TE1/0/26 and TE1/0/27 as BAGG1,  and Gi1/0/17 and Gi1/0/18 as BAGG2)

Both BAGGs have the same configuration on both switches (just trunks with some vlans on it) like below:

[HPE-CMW7-1]display cu inter bridge 2
#
interface Bridge-Aggregation2
 port link-type trunk
 undo port trunk permit vlan 1
 port trunk permit vlan 2 503 599 1234
 link-aggregation mode dynamic

RSTP is enabled, I've not opened BAGG1 - it is shutdown, and the cost of BAGG1 which is naturally lower then BAGG2 (10G default cost 1000 vs. 1G defailt cost 10000) has been raised to 20000.

[HPE-CMW7-2]display cu inter Bridge-Aggregation 1
#
interface Bridge-Aggregation1
 port link-type trunk
 undo port trunk permit vlan 1
 port trunk permit vlan 2 503 599 1234
 shutdown
 link-aggregation mode dynamic
 stp instance 0 cost 20000

I ping a the management address on both switches every 200ms throught the whole process as I do each step to see if I loose packets (i should have pinged a device on an access port - not on the management vlan - for proper testing, but i didn't bother since the management vlan in this setup has the same topology as the others (just running RSTP). 

STP: (Switch 2 is root)

[HPE-CMW7-2]display stp
-------[CIST Global Info][Mode RSTP]-------
Bridge ID : 28672.d894-037c-f8bb

[HPE-CMW7-2]display stp root
 MST ID   Root Bridge ID        ExtPathCost IntPathCost Root Port
 0        28672.d894-037c-f8bb  0           0

[HPE-CMW7-1]display stp
-------[CIST Global Info][Mode RSTP]-------
 Bridge ID           : 32768.d894-037c-1abf

[HPE-CMW7-1]display stp root
 MST ID   Root Bridge ID        ExtPathCost IntPathCost Root Port
 0        28672.d894-037c-f8bb  10000       0           BAGG2

And switch 1 has rootport BAGG2, cost 10000 - all good.

Bridgeaggregations are as follows

[HPE-CMW7-2]display link-aggregation verbose
...
Aggregate Interface: Bridge-Aggregation1
Aggregation Mode: Dynamic
Loadsharing Type: Shar
Management VLAN : None
System ID: 0x8000, d894-037c-f8bb
Local:
  Port             Status  Priority Oper-Key  Flag
--------------------------------------------------------------------------------
  XGE1/0/27        U       32768    1         {AC}
  XGE1/0/28        U       32768    1         {AC}
Remote:
  Actor            Partner Priority Oper-Key  SystemID               Flag
--------------------------------------------------------------------------------
  XGE1/0/27        27      32768    1         0x8000, d894-037c-1abf {ACEF}
  XGE1/0/28        28      32768    1         0x8000, d894-037c-1abf {ACEF}

Aggregate Interface: Bridge-Aggregation2
Aggregation Mode: Dynamic
Loadsharing Type: Shar
Management VLAN : None
System ID: 0x8000, d894-037c-f8bb
Local:
  Port             Status  Priority Oper-Key  Flag
--------------------------------------------------------------------------------
  GE1/0/17         S       32768    2         {ACDEF}
  GE1/0/18         S       32768    2         {ACDEF}
Remote:
  Actor            Partner Priority Oper-Key  SystemID               Flag
--------------------------------------------------------------------------------
  GE1/0/17         17      32768    2         0x8000, d894-037c-1abf {ACDEF}
  GE1/0/18         18      32768    2         0x8000, d894-037c-1abf {ACDEF}

Now I can enable BAGG1 on both switches:

interface Bridge-Aggregation 1
 undo shutdown

If I look at STP, I can observe that the BAGG1 is now up and on switch 1 has role ALTERNATE, as expected.

[HPE-CMW7-1]display stp brief
 MST ID   Port                                Role  STP State   Active Protect
 0        Bridge-Aggregation1                 ALTE  DISCARDING  NONE
 0        Bridge-Aggregation2                 ROOT  FORWARDING  NONE

Now i can lower the cost on BAGG1 on both switches to 1000.

interface Bridge-Aggregation 1
 stp cost 1000

and again look at stp

[HPE-CMW7-1]display stp brief
 MST ID   Port                                Role  STP State   Active Protect
 0        Bridge-Aggregation1                 ROOT  FORWARDING  NONE
 0        Bridge-Aggregation2                 ALTE  DISCARDING  NONE

and observe that BAGG1 is now the the root port on switch 1.

Last thing to do is to shutdown BAGG2, which is now harmless since traffic is on BAGG1

No pings lost using $ fping -l -p200 <ip-of-switch-1> <ip-of-switch-2>

ip-of-switch-1 : xmt/rcv/%loss = 732/732/0%, min/avg/max = 0.88/3.58/82.4
ip-of-switch-2 : xmt/rcv/%loss = 732/732/0%, min/avg/max = 0.91/2.34/36.6

no loss and worst responce time was 82ms. (switch was slow to answer during STP recalculation, which is why it would have been better to ping somthing else)

PS: This is with Rapid STP, if you're running plain STP. The concept of alternate ports is not available, and you need to consider what will happen.

Regards

[Vince-Whirlwind]

RSTP is nice and rapid then.