samba - Dec 2021 - smbstatus no longer working after upgrade to 4.13.14

On Mon, 2021-12-06 at 09:24 +0000, Julian Einwag via samba
wrote:
> Hi,
> 
> we?re running a Samba CTDB cluster on Ubuntu 20.04. The cluster is
> part of our windows domain.
> After a recent upgrade to 4.13.14 (from 4.11.6), the smbstatus
> command is no longer working. Clients can connect just fine. This is
> the output produced by smbstatus -a:
> 
> 
> Samba version 4.13.14-Ubuntu
> PID     Username     Group        Machine                            
>        Protocol Version  Encryption           Signing
> -------------------------------------------------------------------
> ---------------------------------------------------------------------
> db_open: Failed to initialize messaging
> Failed to initialize session_global: NT_STATUS_OBJECT_NAME_COLLISION
> 
> Service      pid     Machine       Connected
> at                     Encryption   Signing
> -------------------------------------------------------------------
> --------------------------
> db_open: Failed to initialize messaging
> Failed to initialize session_global: NT_STATUS_OBJECT_NAME_COLLISION
> Failed to traverse sessions: NT_STATUS_OBJECT_NAME_COLLISION
> 
> db_open: Failed to initialize messaging
> /run/samba/locking.tdb not initialised
> This is normal if an SMB client has never connected to your server.
> 
> 
> 
> 
> Our smb.conf:
> 
> [global]
> bind interfaces only = no
> interfaces = eno1
> security = ADS
> workgroup = OURWORKGROUP
> realm = ******
> netbios name = ******
> csc policy = disable
> winbind enum users = yes
> winbind enum groups = yes
> winbind use default domain = Yes
> log file = /var/log/samba/log.%m
> max log size = 1000
> syslog = 3
> log level = 1 auth_audit:3 auth_json_audit:3
> panic action = /usr/share/samba/panic-action %d
> clustering = yes
> include = registry
> allow insecure wide links = yes
> username map = /etc/samba/user.map
> veto files = /Thumbs.db/
> min domain uid = 0
> delete veto files = yes
> idmap config * : backend = tdb
> idmap config * : range = 3000-7999
> 
> idmap config OURWORKGROUP : backend = rid
> idmap config OURWORKGROUP : range = 10000-999999
> 
> # Template settings for login shell and home directory
> template shell = /bin/bash
> template homedir = /home/%U
> 
> 
> ceph: config_file = /etc/ceph/ceph.conf
> ceph: user_id = admin
> map acl inherit = yes
> store dos attributes = yes
Not an expert on clustering by any means, but is ctdb running ?
Also you do not seem to to have 'vfs objects = ceph' set, but you are
using Samba 'ceph' parameters, have you read 'man vfs_ceph' ?

Rowlad

> On 6. Dec 2021, at 11:41, Rowland Penny via samba <samba at
lists.samba.org> wrote:
>> : config_file = /etc/ceph/ceph.conf
>> ceph: user_id = admin
>> map acl inherit = yes
>> store dos attributes = yes
>
> Not an expert on clustering by any means, but is ctdb running ?
> Also you do not seem to to have 'vfs objects = ceph' set, but you
are
> using Samba 'ceph' parameters, have you read 'man vfs_ceph? ?
You are correct, our cluster is ceph backed, though we?re not using vfs_ceph,
but a kernel mount instead.
Removing the parameters doesn?t change anything on our test cluster, which is
running the same samba issue and experiences the same issue.

We did initially test vfs_ceph, but decided not to use it.

I posted a debug (smbstatus -a -d10) here:
https://gist.github.com/jeinwag/e5831526749226a4c710a0c60b572ebf

Looks like it's trying to communicate with ctdbd, but is failing?
Part of the log:

messaging_dgm_ref: messaging_dgm_get_unique returned Success
messaging_dgm_ref: unique = 12399100214177679836
messaging_ctdb_ref: messaging_ctdb_init returned Device or resource busy
messaging_ctdb_init failed: Device or resource busy
messaging_init_internal: messaging_ctdb_ref failed: Device or resource busy
msg_dgm_ref_destructor: refs=0x55f744cbc0e0
db_open: Failed to initialize messaging