D2 T1 S5 ACI Multisite Troubleshooting
D2 T1 S5 ACI Multisite Troubleshooting
D2 T1 S5 ACI Multisite Troubleshooting
Roland Ducomble
CX Technical Leader – ACI Solutions Team
27th March 2019
Introduction
MP-BGP - EVPN
ACI
Multi Site
Orchestrator
(MSO)
REST
GUI
API Availability Zone ‘B’
Availability Zone ‘A’
Region ‘C’
▪ Separate ACI Fabrics with independent APIC clusters ▪ MP-BGP EVPN control plane between sites
▪ ACI MSO pushes cross-fabric configuration to multiple ▪ Data Plane VXLAN encapsulation across sites
APIC clusters providing scoping of all configuration ▪ End-to-end policy definition and enforcement
changes
© 2019 Cisco and/or its affiliates. All rights reserved.
ACI Multi-Site Network information carried across Site
(VNID for VRF or BD)
Identity information carried across site
(Class-id aka pcTag
MP-BGP - EVPN
ACI
MSO
A same VRF, BD or EPG created on two different site will have different VNID and
Class-Id as those are allocated by the local APIC cluster (not by MSO)
Need for a translation !
In ACI multisite, data packet are across site are vxlan encapsulated with the
vnid/class of the source site
The Destination site spine will have the role of translating those value
© 2019 Cisco and/or its affiliates. All rights reserved.
ACI Multi-Site
Namespace translation
Translation of Class-ID, VNID
Translation of Source IP Network (scoping of name spaces)
VTEP address
MP-BGP - EVPN
…
ACI
MSO
5
ACI Multipod versus Multisite
Multipod Multisite
• Controller : Single APIC controller for all • Controller : separate APIC cluster per site,
pod MSO talks to each APIC cluster
• Namespace : No need of any translation • Namespace : need to translate vnid and
(all come from same APIC cluster) pcTag across site (done on target spine)
• Control Plane : BGP EVPN used to sync • Control Plane : BGP EVPN used to sync
COOP DB COOP DB
• Unicast Data : VXLAN encapsulated • Unicast Data : VXLAN encapsulated in
• Leaf to Leaf tunnel for established flow source site VNID
• Leaf to target site spine tunnel for establish flow
(Always need to hop by Spine for Xlate)
• BUM traffic in the IPN: multicast
encapsulated in BD GIPo • BUM traffic in the ISN : unicast copy to each
target site (HREP – Head end replication)
• Control plane : PIM BiDir
© 2019 Cisco and/or its affiliates. All rights reserved. • Control plane: no need of multicast control plane
Namespace Translation
troubleshooting
remote site
# fv.Ctx
Logical BD – site 1
Total Objects shown: 1
fvBD
Translation of BD vnid to spine site 1
From 15204288 to 150723234
fvSiteAssociated
bdsol-aci35-apic1# moquery -d uni/tn-RD-L2/BD-Web/stAsc/site-2
Total Objects shown: 1
# fv.RemoteId
siteId : 2
childAction :
fvRemoteId descr :
dn : uni/tn-RD-L2/BD-Web/stAsc/site-2
lcOwn : local
modTs : 2018-05-03T03:14:40.895+00:00
Site 2 BD monPolDn : uni/tn-common/monepg-default
name :
bdsol-aci36-apic1# moquery -c fvBD -f 'fv.BD.seg nameAlias :
=="15073234"' | egrep "dn|scope|seg" ownerKey :
ownerTag :
dn : uni/tn-RD-L2/BD-Web remoteCtxPcTag : any
scope : 2162688 remotePcTag : 15073234 Actually remote BD VNID
seg : 15073234 rn : site-2
status :
© 2019 Cisco and/or its affiliates. All rights reserved. uid : 15374
bdsol-aci35-apic1# moquery -d uni/tn-RD-L2/ap-App/epg-Web/stAsc/
Total Objects shown: 1
fvSiteAssociated
Site 2 EPG
lcOwn : local
modTs : 2018-05-03T03:14:40.895+00:00
monPolDn : uni/tn-common/monepg-default
name :
nameAlias :
bdsol-aci36-apic1# moquery -c fvAEPg -f 'fv.AEPg.pcTag ownerKey :
=="49155"' | egrep "dn|scope|pcTag" ownerTag :
dn : uni/tn-RD-L2/ap-App/epg-Web remoteCtxPcTag : any
pcTag : 49155 remotePcTag : 49155
rn : site-2
scope © 2019 Cisco and/or its affiliates.
: 2162688
All rights reserved. status :
uid : 15374
Process involved for vnidmap/sclass/site-etep
on Spine only
This runs on
This run on sup-lc. Call
sup. Listen sdk hal api to
for events Sdk Programs the
Dcimgr Dcimc program site-
from PE and etep, Hal hardware
send it Dcimc vnidmap
on Lc sclassmap
• HAL CLI :
module-2# show platform internal hal objects dci ?
all Dump All HAL objects
remotesite Remotesite or wan instance
remotesiteetep Unicast etep that belongs to this remotesite
remotevrfvnid Vrf for remotesite object
sclassmap Sclass mapping for remotesite vrf
vnidmap Vnid mapping for remotesite object
© 2019 Cisco and/or its affiliates. All rights reserved.
DCI mgr – xlate
pod36-spine1# show dcimgr repo vnid-maps detail pod36-spine1# show dcimgr repo sclass-maps detail
-------------------------------------------------------------- ----------------------------------------------------------
Remote | Local Remote | Local
site Vrf Bd | Vrf Bd Rel-state site Vrf PcTag | Vrf PcTag Rel-state
-------------------------------------------------------------- ----------------------------------------------------------
1 2981888 | 2293760 [formed] 1 2981888 49153 | 2293760 49153 [formed]
0x2d8000 | 0x230000 0x2d8000 0xc001 | 0x230000 0xc001
-------------------------------------------------------------- -------------------------------------------------------------------
1 2981888 16678778 | 2293760 16154554 [formed] 1 2981888 49154 | 2293760 49155 [formed]
0x2d8000 0xfe7f7a | 0x230000 0xf67fba 0x2d8000 0xc002 | 0x230000 0xc003
-------------------------------------------------------------- -------------------------------------------------------------------
1 3014656 | 2457600 [formed] 1 2981888 16387 | 2293760 16386 [formed]
0x2e0000 | 0x258000 0x2d8000 0x4003 | 0x230000 0x4002
-------------------------------------------------------------------
1 3014656 49153 | 2457600 49153 [formed]
0x2e0000 0xc001 | 0x258000 0xc001
-------------------------------------------------------------------
1 3014656 16387 | 2457600 32772 [formed]
0x2e0000 0x4003 | 0x258000 0x8004
• For VPNv4 : ..
VRF RD : 10.0.88.95:6
Export RT list:
• RD is typically the PTEP of the origination:vrf_id 132:2654211
Import RT list:
• RT is typically the BGP_ASN:VRF_VNID 132:2654211
• In multisite, RT will be different across site for the same vrf, so we must configure cross
route-target import/export for correct BGP path exchange to happen
• This is done using bgp EVI (EVPN instance) (show bgp internal evi XXXX) . Similar to pcTag
and VNID translation this is also pushed using object model
© 2019 Cisco and/or its affiliates. All rights reserved.
• Note that here we need route-target per BD (not per VRF)
BGP VNI
• Route Exchange issues can be seen either in the source or on the
remote site.
• Check if the BGP MOs are created for VNIs/RTs and RDs correctly. These
MOs are created only on spines in every site. These MOs are created
when the VRF/BD/EPGs are stretched or the contracts are created at EPG
level
• Following shows mapping of BGP VNIDs and what routes are
requested from COOP and why they are used:
1. Proxy – Spine COOP lookup 3. In all case Rx on spine does vnid/sclass translation
SIP and DIP outer rewritten IP Network
1 3
MP-BGP - EVPN
2
2. Known EP – EPM lookup on leaf
© 2019 Cisco and/or its affiliates. All rights reserved.
DIP outer set on lef
SIP outer Rewritten when passing by ingress spine
LAB Stretched VRF ACI Multi-Site
Pod35 Pod36
Tenant IPA
BD GW
Aci-35-interconnect 172.16.2.254/24 VRF DC:DC1
BD GW
1/51-52 172.16.4.254/24 BD2 172.16.2.54/24
172.16.1.254/24 1/49-50 Route
BD1 172.16.1.54/24
Web-EPG2
172.16.3.254/24 172.16.[1-4].0/24
Web-EPG1
Route 2/5-6 2/5-6
172.16.[1-4].0/24
aci35-spine1
aci36-spine1
C1 C1 C2 C2
1/49-50 1/49-50
aci35-leaf1 aci36-leaf1
Test :
172.16.3.2 to 172.16.2.2
IP address : 172.16.2.2
Vrf : 2457600
pod36-leaf1# show system internal epm endpoint ip 172.16.2.2 Flags : 0
EP bd vnid : 16220082
MAC : 0050.56b1.4403 ::: Num IPs : 1 EP mac : 00:50:56:B1:44:03
IP# 0 : 172.16.2.2 ::: IP# 0 flags : Publisher Id : 10.1.48.64
Vlan id : 21 ::: Vlan vnid : 8194 ::: VRF name : DC:DC1 Record timestamp : 05 02 2018 02:29:12 339899902
BD vnid : 16220082 ::: VRF vnid : 2457600 Publish timestamp : 05 02 2018 02:29:12 340145880
Phy If : 0x1a001000 ::: Tunnel If : 0 Seq No: 0
Interface : Ethernet1/2 Remote publish timestamp: 01 01 1970 00:00:00 0
Flags : 0x80004c04 ::: sclass : 32771 ::: Ref count : 5 URIB Tunnel Info
EP Create Timestamp : 04/19/2018 07:03:23.999543 Num tunnels : 1
EP Update Timestamp : 05/02/2018 02:33:29.507208 Tunnel address : 10.1.48.64
EP Flags : local|IP|MAC|sclass|timer| Tunnel ref count : 1
:::: Remote COOP entry site 1
Extract BGP table site 2 Publisher id is the spine DP TEP in site 2
pod36-spine1# show bgp l2vpn evpn vrf overlay-1 | egrep "Route Dis|172.16.2.2\]” pod35-spine1# show coop internal info ip-db | egrep -A
Route Distinguisher: 1:49774514 (L2VNI 16220082) 15 -B 1 "172.16.2.2$"
*>l[2]:[0]:[16220082]:[48]:[0050.56b1.4403]:[32]:[172.16.2.2]/272 ------------------------------
Route Distinguisher: 10.10.35.102:136 (L2VNI 1) IP address : 172.16.2.2
*>l[2]:[0]:[16220082]:[48]:[0050.56b1.4403]:[32]:[172.16.2.2]/272 Vrf : 3014656
Flags : 0x4
EP bd vnid : 15925206
Extract BGP table site 1 EP mac : 00:50:56:B1:44:03
Publisher Id : 10.10.35.102
pod35-spine1# show bgp l2vpn evpn vrf overlay-1 | egrep "Route Dis|172.16.2.2\]" Record timestamp : 01 01 1970 00:00:00 0
Publish timestamp : 01 01 1970 00:00:00 0
Route Distinguisher: 1:49774514 Seq No: 0
*>e[2]:[0]:[0]:[48]:[0050.56b1.4403]:[32]:[172.16.2.2]/272 Remote publish timestamp: 04 24 2018 05:05:34 611613733
Route Distinguisher: 1:32702422 (L2VNI 15925206) URIB Tunnel Info
*>e[2]:[0]:[15925206]:[48]:[0050.56b1.4403]:[32]:[172.16.2.2]/272 Num tunnels : 1
© 2019 Cisco and/or its affiliates. All rights reserved.
Route Distinguisher: 10.10.35.101:135 (L2VNI 1) Tunnel address : 10.10.35.102
*>e[2]:[0]:[15925206]:[48]:[0050.56b1.4403]:[32]:[172.16.2.2]/272 Tunnel ref count : 1
DCI Mgr on spine pod 36 (site 2)
Remote Site pod36-spine1# moquery -c dciAnycastExtn
DP-ETEP and Total Objects shown: 2
Aci-35-interconnect
DCI mgr vnid map 1/49-50
1/51-52
aci35-leaf1 aci36-leaf1
Ex :
hom_elam_in_l4v_tn.tn_nonce_info: 0x188002
2/1-2 2/1-2
Outer : pod35-leaf1 PTEP to 10.10.35.102 (site2 dp-etep) Outer : 10.10.35.101 (site1 DP-ETEP) to pod36-leaf1 PTEP
Outer L4 : site1 vrf VNID – Sclas App-EPG1 Outer L4 : site2 vrf VNID – Sclass App-EPG1 Translated
Inner : 172.16.3.2 to 172.16.2.2 Inner : 172.16.3.2 to 172.16.2.2
2/1-2 2/1-2
Outer : pod35-leaf1 PTEP to 10.0.88.66 (site1 Outer : 10.10.35.101 (site1 DP-ETEP) to pod36-leaf1 PTEP
anycast proxy) Outer L4 : site2 vrf VNID – Sclass App-EPG1 Translated
Outer L4 : site1 vrf VNID – Sclas App-EPG1 Inner : 172.16.3.2 to 172.16.2.2
Inner : 172.16.3.2 to 172.16.2.2
No EPM entry relying1/49-50
on 1/49-50
BD subnet route to EPM learning for 172.16.3.2 to tunnel to site1 DP-ETEP
SPine aci35-leaf1 aci36-leaf1
Pod35 Pod36
Aci-35-interconnect
DCI mgr translation DCI mgr translation
Vrf vnid 2457600 -> 3014656 1/49-50
1/51-52
Vrf vnid 3014656 -> 2457600
Sclass 32771 -> 32772 Sclass 32770 -> 16390
2/5-6 2/5-6
aci35-spine1
aci36-spine1
Policy Enforcement
- Ingress leaf derives sclass and vnid based on local EPM
2/1-2 2/1-2
- If Remote EPM is populated – Enforce Policy (as usual)
ACI Multi-Site
L2 flooding
L2 flooding
BD GW 1/49-50
1/51-52 BD GW Tenant RD-L2
10.1.1.254/24 10.1.1.254/24
10.2.2.254/24 2/5-6 2/5-6 10.2.2.254/24 VRF L2
aci35-spine1 BD1/10.1.1.254/24
aci36-spine1
Web-EPG
BD2/10.2.2.254/24
App-EPG
1/49-50 1/49-50
aci35-leaf1 aci36-leaf1
Vm 10.1.1.35 Vm 10.1.1.36
# fv.BD
name : Web
OptimizeWanBandwidth : yes
arpFlood : yes
bcastP : 225.0.216.80
• BD must be set with intersite childAction
configIssues
:
:
L2 flooding
BD GW 1/49-50
1/51-52 BD GW Tenant RD-L2
10.1.1.254/24 10.1.1.254/24
10.2.2.254/24 2/5-6 2/5-6 10.2.2.254/24 VRF L2
aci35-spine1 BD1/10.1.1.254/24
aci36-spine1
Web-EPG
BD2/10.2.2.254/24
App-EPG
1/49-50 1/49-50
aci35-leaf1 aci36-leaf1
BD GW 1/49-50
1/51-52 BD GW
10.1.1.254/24 10.1.1.254/24
10.2.2.254/24 2/5-6 2/5-6 10.2.2.254/24
aci35-spine1
aci36-spine1
aci35-leaf1 aci36-leaf1
Pod35 Pod36
Aci-35-interconnect
BD GW BD GW
10.1.1.254/24 1/51-52 10.1.1.254/24
1/49-50
10.2.2.254/24 10.2.2.254/24
2/5-6 2/5-6
aci35-spine1
aci36-spine1
2/1-2 2/1-2
1/49-50 1/49-50
aci35-leaf1 aci36-leaf1
OSPF OSPF
Vm 10.1.1.35 Vm 10.1.1.36
Lo1 - 10.30.1.1/24 Lo1 10.30.2.1/24
Vm 10.1.1.35
Lo1 10.30.2.1/24
Vm 10.1.1.36 In summary: we do not support traffic
Lo1 - 10.30.1.1/24
From EPG in siteX to L3 out in siteY
© 2019 Cisco and/or its affiliates. All rights reserved. (planned for 4.2)
Why EP to remote L3 out do not work
• No VPNv4 route exchange across multisite BGP session
• No l2vpn evpn type 5 neither
• Site 2 never got route from Site 1 – L3 out