Ceph » Linux kernel client

On a dual-stack cluster, the kernel client chokes on the list of monitor addresses:

libceph: mon2 (2)[xxxx:xxx:xx:xxxx::1]:3300 session established
libceph:  ceph_decode_entity_addrvec addr_cnt 2
libceph:  ceph_decode_entity_addrvec i 0 addr (2)[xxxx:xxx:xx:xxxx::1]:3300
libceph:  ceph_decode_entity_addrvec i 1 addr (2)xx.xxx.xxx.90:3300
libceph: another match of type 2 in addrvec
libceph: failed to decode server addrs: -22
libceph: mon2 (2)[xxxx:xxx:xx:xxxx::1]:3300 read processing error
libceph: mon2 (2)[xxxx:xxx:xx:xxxx::1]:3300 session lost, hunting for new mon

The above was captured with:

# grep libceph.*=p /proc/dynamic_debug/control 
net/ceph/decode.c:114 [libceph]ceph_decode_entity_addrvec =p " %s addr_cnt %d\n" 
net/ceph/decode.c:122 [libceph]ceph_decode_entity_addrvec =p " %s i %d addr %s\n" 

This repeats in a tight loop until aborted or some other timeout occurs