DNS SRV default service name not used anymore
I am in the process of upgrading from Kraken to Luminous.
I am using DNS SRV records to lookup MON servers.
I've started by upgrading my ceph admin node (ceph-deploy) and I get this error after upgrading:
# ceph status 2017-08-31 15:57:24.811639 7f67164ab700 -1 WARNING: the following dangerous and experimental features are enabled: btrfs 2017-08-31 15:57:24.889154 7f67164ab700 -1 WARNING: the following dangerous and experimental features are enabled: btrfs unable to get monitor info from DNS SRV with service name: ceph-mon no monitors specified to connect to. 2017-08-31 15:57:24.892900 7f67164ab700 -1 failed for service _._tcp [errno 2] error connecting to the cluster
tcpdump output is when executing 'ceph status' command:
15:57:24.889523 IP 192.168.16.194.38469 > 192.168.16.2.domain: 28989+ SRV? _._tcp. (24) 15:57:24.892233 IP 192.168.16.2.domain > 192.168.16.194.38469: 28989 NXDomain 0/1/0 (99) 15:57:24.892364 IP 192.168.16.194.42709 > 192.168.16.2.domain: 35339+ SRV? _._tcp.<DOMAIN>. (34) 15:57:24.892816 IP 192.168.16.2.domain > 192.168.16.194.42709: 35339 NXDomain* 0/1/0 (90)
As you can see, ceph is looking for _._tcp.<DOMAIN>. SRV record, instead of _ceph-mon._tcp.<DOMAIN>.
The issue was fixed by forcing service name in ceph.conf to its default value:
mon_dns_srv_name = ceph-mon
But it was working fine in kraken.