When I try to create the .smb pool, I get an error message: # ceph osd pool create .smb pool names beginning with . are not allowed I assume I can just change to using a pool without the leading period. When I do the shares, how do I format the share path? Does the ceph file system get mounted in a specific location? -Rob -----Original Message----- From: John Mulligan <phlogistonjohn@asynchrono.us> Sent: Tuesday, September 3, 2024 4:08 PM To: ceph-users@ceph.io Cc: Robert W. Eckert <rob@rob.eckert.name> Subject: Re: [ceph-users] SMB Service in Squid On Tuesday, September 3, 2024 3:42:29 PM EDT Robert W. Eckert wrote:
I have upgraded my home cluster to 19.1.0 and wanted to try out the SMB orchestration features to improve my hacked SMB shared using CTDB and SMB services on each host.
Hi there, thanks for trying out the new SMB stuff!
My smb.yaml file looks like
service_type: smb service_id: home placement: hosts: - HOST1 - HOST2 - HOST3 - HOST4 spec: cluster_id: home features: - domain #clustered: true config_uri: rados://.smb/home/scc.toml custom_dns: - "<DNS SERVERS>" join_sources: - "rados:mon-config-key:smb/config/home/join1.json" # cluster_meta_uri: rados://.smb/home/meta #cluster_lock_uri: rados://.smb/home/lock include_ceph_users: - client.smb.fs.cluster.home
#cluster_public_addrs: #address: "192.168.2.175" #destination: "192.168.2.0/24"
When I first ran ceph orch apply -i smb.yaml, it didn't like the sections I commented out related to clusters- this may be that I formatted them wrong?
This is probably because you are using squid. Those fields are only in main and we do not plan on backporting them (yet, see below).
I would get errors like:
Error EINVAL: ServiceSpec: __init__() got an unexpected keyword argument 'cluster_meta_uri'
After commenting out the clustering (for now), I successfully applied this YAML, however the .smb pool was never created so I cannot go on to the next task of fiddling around with the config files and config json.
Is there a way to create the .smb pool manually?
Yes, you need to create the pool manually with `ceph osd pool create` [1]. Also assign it an application (for example `smb`) [2]: ceph osd pool application enable .smb smb [1] https://docs.ceph.com/en/latest/rados/operations/pools/#creating-a-pool [2] https://docs.ceph.com/en/latest/rados/operations/pools/#associating-a-pool-w... Once you have a JSON configuration file you can upload it using the `rados` cli tool. Something like `rados --pool=.smb --namespace=<ns> put <objname> <filename>`
Also is there any good basic examples of a config json? I am not connecting to active directory (On Windows 365 accounts so no local AD).
I> will eventually write a script to pull the user details and map I> them to
The configuration json is defined by the sambacc project [3]. It's a JSON wrapper around samba's configuration plus some container setup magic. [3] https://github.com/samba-in-kubernetes/sambacc/blob/master/docs/ configuration.md the local hosts, but want to get basic services up first.
That sounds pretty cool. I'd love to see what you come up with. One word of caution about the smb service, especially as it appears in Squid. While it should be entirely usable on it's own it will be much more work to configure it manually when you use it directly. We've developed a new smb manager module that allows you to manage clusters and shares without having to know as many of the lower level details needed to use the service spec directly. That said, I will support the service if you find bugs, etc. It's just not the intended interface for most users of smb on Ceph. This is similar to Ceph's NFS module and NFS service (IMO). It happens that the smb service spec got added to ceph main before squid got branched and so it is available on squid but almost everything else SMB related is only on the Ceph main branch today. The smb module and related features should be usable by the wider community with the Ceph Tentacle release. Adam King and I have discussed the possibility of feature backports to Squid but I wanted to get the overall suite of smb things more mature on ceph main first.