Eosxd hangs on the client


I am trying eosxd for fuse mounting a test EOS cluster that uses kerberos authentication. It was working well apparently for some time but know it hangs after a few minutes and does not recover.

On the workstation, as root:

 [root@nanpc118 ~]# eosxd -ofsname=nansubaeos.in2p3.fr:/eos /eos
# fsname='nansubaeos.in2p3.fr:/eos'
# -o allow_other enabled on shared mount
# -o big_writes enabled
# no config file - running on default values
# extracted remote mount dir from fsname is '/eos'
# extracted connection host from fsname is 'nansubaeos.in2p3.fr'
# File descriptor limit: 524288 soft, 524288 hard
# allowing max read-ahead buffers of 1073741824 bytes
# allowing max write-back buffers of 1073741824 bytes
# Disabling nagle algorithm (XRD_NODELAY=1)
# Setting MALLOC_CONF=dirty_decay_ms:0

On the client as a user I can see the content of EOS but after a while, the ls command hangs and never returns.

The eosxd process is still here:

[root@nanpc118 ~]# ps -ef | grep eos
root     12718     1  0 13:47 ?        00:00:23 eosxd -ofsname=nansubaeos.in2p3.fr:/eos /eos

It still add entries in /var/log/eos/fusex/fuse.nansubaeos.in2p3.fr-eos.stats.

nansubaeos.in2p3.fr is a round-robin DNS alias that lists the 2 EOS managers (alternatively)

How can I debug this ?

Thank you


I have a mount that did not hung for about 24h. Instead of using the DNS alias nansubaeos.in2p3.fr I used the name of one of the managers (nanxr15.in2p3.fr) that is part of this alias.
=> Could the use of a DNS alias be the cause of the problems ?

Also, I cannot use the command eos fusex ls:

[root@nanxrd15(EOSMASTER) ~]#eos -b fusex ls
error: you have to be root to list VSTs (errc=1) (Operation not permitted)