EFA: Error on registration for a 'relp' handler on endpoint

[Alberto_Oter]

Hi experts,

I have 2 scenarios of 2 different pair of SLXs, where 1 syslog-relp has been configured successfully with an FQDN at EFA, for the second EFA scenario is not allowing to configure to set up this syslog destination. Should I have to check something special?; both are EFA 2.4.5

Here some highlights of the scenario:

-I'm trying to add a new entry to efa notification subscribers list, no other subscriber exists currently.

-In my command instead of the address I'm trying to configure FQDN, and I'm getting an error:

Command: (efa:extreme)hostname#efa notification subscribers add-syslog-relp --address hostname01-.eng.phone.net --cacert hostname01-ca-combined.pem

Output: Error : Error on registration for a 'relp' handler on endpoint 'hostname01-.eng.phone.net:514'. ERROR: The host from the endpoint URL is not reachable.

-I currently have IP reachability from EFA external Network to the remote DNS server. (ping/tracepath, both works properly). ping is working even using the FQDN.

-NSLookup is answering properly with the IP resolved

-Port is open:

(efa:extreme)extreme@TPVM-L1B:/$ nc -vz hostname01-.eng.phone.net 514
Connection to hostname01-.eng.phone.net 514 port [tcp/shell] succeeded!

-I did some tcpdumps at EFA but those are not showing noting interesting, just when I try to add the endpoint it requests an ICMP to the destination, this ping is answered properly.

-The only way to add this entry is with the pure IP Address. <-- Interesting that in the first EFA Scenario I'm not using the IP Address but the FQDN.

-Should it be a problem on the endpoint? what I can ask for?

Thanks in advance experts!!

I'm still not able to register the relp destintation 

[Michael_Morey · Accepted solution]

Alberto,

I have not seen this issue before, but I looked through our documentation and defect database.  The use of FQDN should absolutely be supported in your version, which you saw worked in your first example.  Also as using the IP address successfully registered the client, so we know the device is reachable.

I found an interesting defect that was found by our internal testers that describes the issue you are seeing.  Ultimately, the issue was found to be delay in the processing DNS requests as the k3s container was not efficiently utilizing the host system DNS.  This issue was reported as a "timing issue", meaning that after several attempts the request would eventually come back within the timeout period and successfully register.

I would recommend trying to repeatedly register the client as a workaround.  If this does not work, I would recommend using the latest version of EFA.  I have been testing the notification service in EFA 3.0 lately and have seen no issues with registering clients using FQDN/IP.

Thanks,