Defects Closed with Code Changes

Add the devices to fabric in the following sequence

1)First add devices that have preconfigured configs

2)Add remaining devices that don't have any configs stored

Removing the devices and adding the devices again to fabric in following sequence

1)First add devices that have preconfigured configs

2)Add remaining unconfigured devices.

1. Check if we have enough disk space as mentioned in system requirements,

2. On the install dir , change to docker_images and import the images using following command

k3s ctr image import docker_k3s_images.tar

3. execute efactl start

There are two workarounds:

1. Remove the health.gob following the below steps

sudo systemctl stop monitor

sudo systemctl status monitor

sudo rm /apps/efadata/misc/health.gob

sudo systemctl start monitor

2. Add the device back

Set the higher value for the user token expiry using "efa auth settings token update" CLI. The default access token expiry value is 1 hour.

Example: efa auth settings token update --type=ACCESS --hours=2 --minutes=30

On BR-SLX9540, BR-SLX9640, SLX9740-40C, SLX9740-80C

1. Configure a fabric and create a BD based tenant, vrf and Port channel under the tenant

2. Configure L3 EPG under the BD based tenant using the vrf and Port channel created in step 1.

3. Step 2 will result in 'Tenant-profile' pw-profile configuration under the Bridge domain corresponding to the Tenant ctag.

4. Delete the port channel which has the associated LIF to the BD (discussed in step 3) on the switch.

OR

Delete the LIF present in the Port channel or Ethernet Port that is associated with the BD (discussed in step 3) on the switch.

5. Execute the DRC for the switch using CLI 'efa inventory drift-reconcile execute --ip <ip-address> --reconcile'

6. Step 5 will result in 'default' pw-profile configuration instead of 'Tenant-profile' under the Bridge domain corresponding to the Tenant ctag and EPG ctags will remain in 'cfg-refreshed' state.

Note: The issue is reproducible on BR-SLX9540, BR-SLX9640, SLX9740-40C, SLX9740-80C platforms only.

Restart iptables service on the node for closing 8080 port to be accessed from outside.

sudo systemctl restart efa-iptables.service

1. Configure a single rack(MCT Pair) Non-CLOS fabric with the SLX devices

2. All links between the MCT pair are brought down

3. "efa fabric show" output indicates the devices with the app-sate set as "cfg-refresh-err"

4. Reload the SLX devices

5. There are connectivity issues towards the SLX from EFA after reload

1. Fix the network connectivity issue between EFA and the SLX devices

2. Execute "lldp disable" followed by "lldp enable" under the physical interfaces interconnecting the MCT pair

3. Execute "efa inventory device update --ip <device-ip>" on the MCT pair

Port-group add operation on a Layer-3 EPG of a bridge-domain enabled tenant that shares ctag with other EPGs may fail. on certain conditions with the error:

Device: <device-1-IP> Ctag: <ctag> Anycast <IP-1> subnet is conflicting with already configured Ve 4097 : Anycast <IP-1> on the device <device-1>

1. Configure two layer-3 EPGs with shared ctags and with ports from different SLX devices that are connected as MCT pair

2. Do an EPG port-group-delete update operation on one EPG to remove all its ports

3. Re-add the same ports back to the EPG

The step 3 will fail with the symptom mentioned above.

rabbitmq pod logs which are present after failover are not purged.

The fix cleans up logs that are not from the current running pod and older than 5 days. The cleanup is triggered once on start of efa-monitor service and then repeated every 5 days.

Delete the rabbitmq logs manually at /apps/efa_logs/rabbitmq/ on TPVM

and /var/log/efa/rabbitmq/ on server.

An incorrect error message is shown to the EFA user when the user tries to add an SLX device (with an expired password) to the fabric, using the below command:

"efa fabric device add --name <fabric name> --rack <rackname> --ip <ip>--username <value> --password <value>"

1.Make slx password expire by configuring the below on the SLX

# show running-config password-attributes

password-attributes force-default-password-change

password-attributes max-password-age 30.

2. Modify the date on SLX past 30 days so that it expires.

3. Try to use device registration in EFA using the command "efa fabric device add". Observe the error message displayed is not complete or meaningful.

Check for password attribute configuration

# show running-config password-attributes

password-attributes force-default-password-change

password-attributes max-password-age 30.

Remove the above configuration so the password does not expire.

There are 2 recoveries:

1. Login to switch through management IP and change password and use that changed password to add device to the fabric using the below CLI

efa fabric device add --name fab2 --rack room1-rack1 --ip 10.20.246.5 --username admin --password <changed password>

2. Reset the device to default configuration so the default credentials can be used in the below CLI

efa fabric device add --name fab2 --rack room1-rack1 --ip 10.20.246.5 --username admin --password <default password>

On fresh installation of EFA/XCO on Ubuntu 20.04 LTS, deployment fails while installing glusterFS with the following error.

Dependency libnl-3-200 is not installed.

Errors were encountered while processing:

libnl-route-3-200:amd64

libibverbs1:amd64

librdmacm1:amd64

ibverbs-providers:amd64

dpkg: dependency problems prevent configuration of libnl-route-3-200:amd64:

libnl-route-3-200:amd64 depends on libnl-3-200 (= 3.4.0-1); however:

Package libnl-3-200 is not installed.

Snmp community creation without view will fail after device registration is done but before the device persist to add the default view efav3View to the DB.

The fix is to move the creation of default view to the beginning of device persist to reduce the time delay.