efa_installer.sh fail on Testing EFA device in the last few days

Question

Today I follow efa get started doc https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa-start.html#efa-start-enable to install EFA software on Amazon Linux 2,  it failed on the last step "Testing EFA device" with below error:

$ sudo ./efa_installer.sh -y

= Starting Amazon Elastic Fabric Adapter Installation Script =

= EFA Installer Version: 1.17.2 =
......

......

== Testing EFA device ==

Starting server...

Starting client...

Error: fi_pingpong test timed out.

==============================================================================
An EFA device has been detected but a ping test has failed. Please consult the

EFA documentation to verify your configuration.

===========================================================================

===================================================
EFA installation complete.

- Please logout/login to complete the installation.
- 
- Libfabric was installed in /opt/amazon/efa
- 
- Open MPI was installed in /opt/amazon/openmpi
- 
===================================================

Log from dmesg:

[   47.835056] efa 0000:00:06.0 rdmap0s6: Unregister ib device

[   47.966388] Elastic Fabric Adapter (EFA) v1.16.0g

[   48.080918] efa 0000:00:06.0: Setup irq:27 name:efa-mgmnt@pci:0000:00:06.0

[   48.085540] efa 0000:00:06.0 efa_0: IB device registered

[   50.633468] efa 0000:00:06.0 rdmap0s6: Failed to process command DEREG_MR (opcode 8) comp_status 7 err -22

[   50.637880] efa 0000:00:06.0 rdmap0s6: Failed to de-register mr(lkey-4) [-22]

[   50.647230] efa 0000:00:06.0 rdmap0s6: Failed to process command DEREG_MR (opcode 8) comp_status 7 err -22

[   50.651981] efa 0000:00:06.0 rdmap0s6: Failed to de-register mr(lkey-3) [-22]

**However, it  does work just a few days ago(last week) when I did the same testing.    Is there any  update for the EFA in the last few days?**

Answer

Hello, thank you for your post. Please refer to the documentation on [EFA limitations](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa.html#efa-limits). I suspect that there may be a problem with the Security Group configuration for the instance on which you ran the EFA installer.

>The EFA must be a member of a security group that allows all inbound and outbound traffic to and from the security group itself.

Please consider opening a support case so that a support engineer can assist you with reviewing the configuration of the EC2 instance and troubleshoot the EFA pingpong test.

efa_installer.sh fail on Testing EFA device in the last few days

Contenuto pertinente