Feature #14348
closed
systemd: support multiple clusters
Added by Artemy Kapitula over 8 years ago.
Updated about 3 years ago.
Description
systemd unit files has a hardcoded "ceph" cluster name.
It makes new releases completely unusable on modern Linux systems who are systemd-based (Fedora, CentOS, RedHat).
The suggested solution is to create a wrapper to parse a one-word daemon identifier and pass it to daemons after splitting to cluster/id parts:
ceph-custom-daemon osd clustername.1 => ceph-osd --cluster clustername --id 1
ceph-custom-daemon mon clustername.monid => ceph-mon --cluster clustername --id monid
Files
- Assignee set to Loïc Dachary
Would you be so kind as to create a pull request with these proposed changes ?
The "ceph" cluster name isn't hardcoded in the systemd unit files. "Environment=CLUSTER=ceph" is just a default, which can be overridden by setting CLUSTER=whatever in /etc/sysconfig/ceph. Does that not provide enough flexibility for using alternate cluster names? What am I missing here?
Tim Serong wrote:
The "ceph" cluster name isn't hardcoded in the systemd unit files. "Environment=CLUSTER=ceph" is just a default, which can be overridden by setting CLUSTER=whatever in /etc/sysconfig/ceph. Does that not provide enough flexibility for using alternate cluster names? What am I missing here?
For the most users it's enough, but not for a heavy usage cases.
The first, that's could be OK if you get only the one cluster. If you get many, it's not an option.
The second, configuration files is hard to automate. With @-style it's easy to configure node form command line.
- Assignee deleted (
Loïc Dachary)
- Tracker changed from Bug to Feature
- Subject changed from systemd unit files broken for all non-default cluster name to systemd: support multiple clusters
- Priority changed from High to Normal
Sounds like it works fine with non-default names, but doesn't work if you have multiple clusters.
- Status changed from New to Resolved
Also available in: Atom
PDF