Defects Closed without Code Changes

Allow firmware download process to complete. Status of the same can be checked using command

efa inventory device firmware-download show --fabric {fabric name}

1) Configure non-clos fabric

2) Create tenant, vrf, epg

3) Admin down device

4) Create multiple epg's , delete an existing epg

5) Manually delete vrf from admin down device

6) Admin up device

7) After admin up, for epg which is in delete-pending theapp-state moved to cfg-ready

Please follow below steps,

1) Create fabric/tenant/PO/EPG

2) Take EFA backup

3) Delete EPG/PO/tenant/fabric

4) Restore EFA backup taken in step (2)

5) Delete tenant from which was created before backup

6) Create same tenant again

7) Create PO under same tenant

As after restore MCT peer details are Nil so we need to perform DRC after restore taken backup.

After step (4) above, we need to perform DRC using inventory CLI as

efa inventory drift-reconcile execute --ip <device ip 1> --reconcile

efa inventory drift-reconcile execute --ip <device ip 2> --reconcile

Restart kubernetes on both nodes of the cluster. As root user:

$ systemctl restart k3s

Error during epg update ctag-range-add operation if different anycastIP is provided for new ctag range with same l2-vni already configured in EPG.

Ex.

Error : EPG: e1(11:11.11.11.1/24) shares network name with EPG: e1(10:10.10.10.1/24) should have same Anycast IP

1. Create L3 EPG epg1 with ctag-1, l2vni-1, anycastIP1

2. Update EPG epg1 with ctag-2, l2vni-1, anycastIP2

For the same l2vni, anycastIP must be the same, the same condition will be verified as part of epg validation which provided above-mentioned error.

1. For same l2vni, anycastIP must be same.

2. Different l2-vni can be used for using different anycastIP.

1. "app-state" of some of the VRFs is shown as "cfg-refreshed" in "efa tenant vrf show" output.

2. The same VRFs whose "app-state" is shown as "cfg-refreshed" are shown as "unstable" VRFs in the "efa tenant epg show" output.

1. Create an 18 node CLOS fabric.

2. Create multiple tenants (e.g. 14) tenants under the fabric created in step 1.

3. Create multiple VRFs (e.g. 400+) under the tenant created in step 2.

4. Create multiple EPGs using the VRFs created in step 3.

5. Check the "efa tenant vrf show" output to know the "app-state" of the VRFs.

1. Check the configuration drift per SLX device using the CLI "efa inventory drift-reconcile execute --ip <slx-device-ip>" to identify if the VRFs (which are in the cfg-refreshed state) are shown in the drift output.

2. If the VRFs are shown in the drift output for a given SLX device, then delete the VRFs from that particular SLX device.

3. Perform DRC (Drift and Reconcile) for the SLX device from which the unstable VRFs were deleted.

4. Steps 3 and 4 need to be done for all the SLX devices.

Network and router creation during EFA HA fail-over, stale entries

Steps:

Started script/stack to create networks and routers

Did HA fail over when network/router creation was in progress

Few network and router creations failed.

EFA HA Failover during stack creation can result in failed network and router entries at OpenStack.

EFA services will be down for few minutes during HA failover, resulting in the failures.

In 2.4.x, adding the same ports/pos with different native vlan was allowed without any validation, this leads to two possible issues:

1. Adding the same ports/PO to multiple epgs with different native-vlan VLAN results in the initial input native vlan being overwritten with the latest native vlan value onto the SLX.

2. Adding the same port/po to multiple epgs, the first epg being created with native VLAN and the second EPG being created without native VLAN, will result in removal of native VLAN configured as part of the first epg after upgrading to 2.5.0 and reloading system with maintenance mode enabled.

Case1:

1. Create fabric, tenant

2. Create PO1 with Device1Port1

3. Create PO2 with Device1Port2

4. Create EPG1 with PO1, PO2, CTAG1, CTAG2 with "switchport mode trunk", and native vlan as CTAG1

5. Create EPG2 with PO2, CTAG3, CTAG4 with native vlan CTAG3

6. On the SLX, PO2 initial native vlan CTAG1 gets replaced with CTAG3, while PO1 will have the initial native vlan CTAG1

Case2:

1. Create fabric, tenant.

2. Create PO1 with Device1Port1

3. Create PO2 with Device1Port2

4. Create EPG1 with PO1, PO2, CTAG1, CTAG2 with "switchport mode trunk" and native vlan as CTAG1

5. Create EPG2 with PO2 CTAG3, CTAG4 without native vlan

6. Upgrade EFA to version 2.5.0

7a. Reload SLX with maintenance mode enable-on-reboot so that the auto DRC (Drift and ReConcile) gets triggered or

7b. Reload SLX and perform manual DRC

8. On the SLX, PO2 initial native vlan CTAG1 gets replaced with the default value, while PO1 will have the initial native vlan CTAG1

1. Identify EPGs with common PO/port with conflicting native vlan

2. Delete conflicting ctag (that are configured wrongly and the vlan that needs to be configured as native vlan but isnt configured as one) using EPG update "ctag-range-delete" operation

3. Add ctag back with correct native vlan (to be configured on SLX) using EPG update ctag-range-add" operation

Below are the steps to reproduce the issue:

1. Create Port-channel on the SLX using EFA

2. Delete the port-channel from the SLX manually

3. Re-create port-channel on the SLX manually with drifted speed configuration

4. Perform DRC without reconciliation