CERN Accelerating science

Is there an ipv4 dependency in xrdfs?

Been testing EOS 4.5.2 with XROOTD 4.9.1 with ipv6.

We have a node that is not dual-stacked, but rather is ipv6 only so we can ensure we are testing over the ipv6 stack.

In doing some xrdfs commands, I’ve noticed that xrdfs seems to have a dependency on ipv4.

No xrdfs ls operation will work on a node with only an ipv6 stack configured:

[dszkola@cmswn1922 ~]$ xrdfs root://cmseos-itb.fnal.gov ls -l /store/user
[FATAL] Socket error: Network is unreachable

[dszkola@cmswn1922 ~]$ xrdfs root://eoscms.cern.ch ls -l /store/user/
[FATAL] Socket error: Network is unreachable

Doing debug operations, it is clear it is attempting to use the ipv4 address for the name given in the command.

[2019-09-20 12:12:22.175492 -0500][Debug  ][ExDbgMsg          ] [137.138.19.8:1094] MsgHandler created: 0x877140 (message: kXR_dirlist (path: /store/user/) ).
[2019-09-20 12:12:22.175520 -0500][Debug  ][PostMaster        ] [137.138.19.8:1094] Found 1 address(es): [::ffff:137.138.19.8]:1094
[2019-09-20 12:12:22.175537 -0500][Debug  ][AsyncSock         ] [137.138.19.8:1094 #0.0] Attempting connection to [::ffff:137.138.19.8]:1094
[2019-09-20 12:12:22.175557 -0500][Error  ][AsyncSock         ] [137.138.19.8:1094 #0.0] Unable to initiate the connection: [ERROR] Socket error: Network is unreachable
[2019-09-20 12:12:22.175568 -0500][Error  ][XRootD            ] [137.138.19.8:1094] Unable to send the message kXR_dirlist (path: /store/user/): [FATAL] Socket error: Network is unreachable

Is this something anyone is concerned about or do you feel it is inconsequential? Are there plans to eliminate this dependency in xrdfs?


Dan Szkola
FNAL