We just updated one of our ceph clusters from 14.2.15 to 14.2.19, and
see some unexpected behavior by radosgw - it seems to ignore parameters
set by the ceph config database. Specifically this is making it start up
listening only on port 7480, and not the configured 80 and 443 (ssl) ports.
Downgrading ceph on the rgw nodes back to 14.2.15 restores the expected
behavior (I haven't yet tried any intermediate versions). The host OS is
CentOS 7, if that matters...
Here's a ceph config dump for one of the affected nodes, along with the
radosgw startup log:
# ceph config dump|grep tier2-gw02
client.rgw.tier2-gw02 basic log_file /var/log/ceph/radosgw.log *
client.rgw.tier2-gw02 advanced rgw_dns_name s3.msi.umn.edu *
client.rgw.tier2-gw02 advanced rgw_enable_usage_log true
client.rgw.tier2-gw02 basic rgw_frontends beast port=80 ssl_port=443 ssl_certificate=/etc/ceph/civetweb.pem *
client.rgw.tier2-gw02 basic rgw_thread_pool_size 512
# tail /var/log/ceph/radosgw.log
2021-04-08 11:51:07.956 7f420b78f700 -1 received signal: Terminated from /usr/lib/systemd/systemd --switched-root --system --deserialize 22 (PID: 1) UID: 0
2021-04-08 11:51:07.956 7f420b78f700 1 handle_sigterm
2021-04-08 11:51:07.956 7f4220bc5900 -1 shutting down
2021-04-08 11:51:07.956 7f420b78f700 1 handle_sigterm set alarm for 120
2021-04-08 11:51:08.010 7f4220bc5900 1 final shutdown
2021-04-08 11:51:08.159 7f2ac6105900 0 deferred set uid:gid to 167:167 (ceph:ceph)
2021-04-08 11:51:08.159 7f2ac6105900 0 ceph version 14.2.19 (bb796b9b5bab9463106022eef406373182465d11) nautilus (stable), process radosgw, pid 88256
2021-04-08 11:51:08.300 7f2ac6105900 0 starting handler: beast
2021-04-08 11:51:08.302 7f2ac6105900 0 set uid:gid to 167:167 (ceph:ceph)
2021-04-08 11:51:08.317 7f2ac6105900 1 mgrc service_daemon_register rgw.tier2-gw02 metadata {arch=x86_64,ceph_release=nautilus,ceph_version=ceph version 14.2.19 (bb796b9b5bab9463106022eef406373182465d11) nautilus (stable),ceph_version_short=14.2.19,cpu=AMD EPYC 7302P 16-Core Processor,distro=centos,distro_description=CentOS Linux 7 (Core),distro_version=7,frontend_config#0=beast port=7480,frontend_type#0=beast,hostname=tier2-gw02.msi.umn.edu,kernel_description=#1 SMP Tue Mar 16 18:28:22 UTC 2021,kernel_version=3.10.0-1160.21.1.el7.x86_64,mem_swap_kb=4194300,mem_total_kb=131754828,num_handles=1,os=Linux,pid=88256,zone_id=default,zone_name=default,zonegroup_id=default,zonegroup_name=default}
BTW I can also change "rgw_frontends" to specify a civetweb frontend
instead and it will still start the default beast...
I haven't seen anyone else report such a problem so I wonder if this is
something local to us - like perhaps I'm using "ceph config" incorrectly
in a way which happened to be accepted before? Has anyone else seen this
behavior?
Graham
--
Graham Allan - gta@xxxxxxx
Associate Director of Operations - Minnesota Supercomputing Institute
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
