As discussed in https://tracker.ceph.com/issues/69475 this
setting is effectively useless as it only controls the directory
inside the container where the ceph-exporter will write out
its asok file, and has no influence over where it is on the
host where the ceph-exporter daemon is deployed. Given any
custom values for the sock_dir setting would have always
been broken, we decided to skip out on writing a proper migration
step in cephadm to deal with this, and instead just force this
field to be unset
Signed-off-by: Adam King <adking@redhat.com>
extra_entrypoint_args=extra_entrypoint_args)
self.service_type = service_type
- self.sock_dir = sock_dir
+ self.sock_dir = None
self.addrs = addrs
self.port = port
self.prio_limit = prio_limit
def validate(self) -> None:
super(CephExporterSpec, self).validate()
+ if self.sock_dir and self.sock_dir != '/var/run/ceph/':
+ raise SpecValidationError(
+ 'sock_dir setting is deprecated and must be either unset or set to /var/run/ceph/'
+ )
+
if not isinstance(self.prio_limit, int):
raise SpecValidationError(
f'prio_limit must be an integer. Got {type(self.prio_limit)}')