Christian Kuntz
2020-Sep-29 05:44 UTC
[Samba] [CTDB] "use mmap = no" Causes wibind to fail
Hello all, I'm currently running samba on debian buster, version 2:4.9.5+dsfg-5 with lustre 2.13.0. The problem I'm experiencing is pretty straightforward, if I set "use mmap = no" in my global config (as suggested by the wiki), winbind fails to start and the cluster grinds to a halt. Here is the error messages from systemd: Sep 28 22:14:17 tenzin systemd[1]: Starting Samba Winbind Daemon... Sep 28 22:14:18 tenzin winbindd[21609]: [2020/09/28 22:14:18.037190, 0] ../source3/lib/dbwrap/dbwrap_ctdb.c:1924(db_open_ctdb) Sep 28 22:14:18 tenzin winbindd[21609]: Could not open tdb /var/lib/ctdb/volatile/g_lock.tdb.0: Invalid argument Sep 28 22:14:18 tenzin winbindd[21609]: [2020/09/28 22:14:18.037574, 0] ../source3/lib/dbwrap/dbwrap_open.c:174(db_open) Sep 28 22:14:18 tenzin winbindd[21609]: failed to attach to ctdb g_lock.tdb Sep 28 22:14:18 tenzin winbindd[21609]: [2020/09/28 22:14:18.037626, 0] ../source3/lib/dbwrap/dbwrap_ctdb.c:1946(db_open_ctdb) Sep 28 22:14:18 tenzin winbindd[21609]: g_lock_ctx_init failed Sep 28 22:14:18 tenzin winbindd[21609]: [2020/09/28 22:14:18.037677, 0] ../source3/lib/dbwrap/dbwrap_open.c:174(db_open) Sep 28 22:14:18 tenzin winbindd[21609]: failed to attach to ctdb secrets.tdb Sep 28 22:14:18 tenzin winbindd[21609]: [2020/09/28 22:14:18.037720, 0] ../source3/passdb/secrets.c:66(secrets_init_path) Sep 28 22:14:18 tenzin winbindd[21609]: Failed to open /var/lib/samba/private/secrets.tdb Sep 28 22:14:18 tenzin winbindd[21609]: [2020/09/28 22:14:18.037761, 0] ../source3/winbindd/winbindd.c:1800(main) Sep 28 22:14:18 tenzin winbindd[21609]: Could not initialize domain trust account secrets. Giving up Sep 28 22:14:18 tenzin systemd[1]: winbind.service: Failed with result 'protocol'. And my testparm output: [global] clustering = Yes disable spoolss = Yes dns proxy = No load printers = No local master = No log file = /var/log/samba/clients/%m.log logging = syslog at 0 file map to guest = Bad User max log size = 100 max stat cache size = 65536 max xmit = 1048676 name resolve order = host wins bcast ntlm auth = ntlmv1-permitted printcap name = /dev/null socket options = IPTOS_LOWDELAY SO_RCVBUF=131072 TCP_NODELAY use mmap = No username map cache time = 60 usershare path idmap config * : range = 1000000-1999999 idmap config * : backend = autorid include = /etc/samba/smb-shares.conf printing = bsd [defaultsettings] access based share enum = Yes available = No browseable = No case sensitive = Yes create mask = 0666 directory mask = 0777 include = /etc/samba/smb-shares.conf.local read only = No strict locking = No use sendfile = Yes vfs objects = catia fruit acl_xattr aio_pthread readdir_attr:aapl_rsize = no fruit:veto_appledouble = no fruit:resource = file fruit:metadata = stream fruit:locking = none fruit:encoding I've tested removing the "use mmap" line and the error disappears. The documentation on the samba wiki ( https://wiki.samba.org/index.php/Configuring_clustered_Samba#Lustre_filesystem) states that the setting is necessary to prevent possible corruption, but states the last tested version was on the 1.4.x and 1.5.x versions of lustre. Is the wiki information still accurate? If so, what needs to change about my configuration to fix this issue? If the information is out of date and the modern utility of "use mmap = no" is unknown, how can I test it? As always, thank you for your time and attention, Christian Kuntz -- <http://opendrives.com>
Am 9/29/20 um 7:44 AM schrieb Christian Kuntz via samba:> Is the wiki information still accurate? If so, what needs to change about > my configuration to fix this issue? > If the information is out of date and the modern utility of "use mmap = no" > is unknown, how can I test it?hm, I guess the mmap no setting should have been removed alongside this change: <https://wiki.samba.org/index.php?title=Configuring_clustered_Samba&diff=11665&oldid=11663#Lustre_filesystem> The option only affects the use of mmap or not in TDB and on a Samba ctdb cluster the tdb databases must be located on cluster node local filesystems, not the cluster filesystem. I'm going to remove the setting from the wiki. -slow -- Ralph Boehme, Samba Team https://samba.org/ Samba Developer, SerNet GmbH https://sernet.de/en/samba/ GPG-Fingerprint FAE2C6088A24252051C559E4AA1E9B7126399E46 -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 833 bytes Desc: OpenPGP digital signature URL: <http://lists.samba.org/pipermail/samba/attachments/20200929/45ce3aab/signature.sig>
On 29/09/2020 08:35, Ralph Boehme via samba wrote:> Am 9/29/20 um 7:44 AM schrieb Christian Kuntz via samba: >> Is the wiki information still accurate? If so, what needs to change about >> my configuration to fix this issue? >> If the information is out of date and the modern utility of "use mmap = no" >> is unknown, how can I test it? > hm, I guess the mmap no setting should have been removed alongside this > change: > > <https://wiki.samba.org/index.php?title=Configuring_clustered_Samba&diff=11665&oldid=11663#Lustre_filesystem> > > The option only affects the use of mmap or not in TDB and on a Samba > ctdb cluster the tdb databases must be located on cluster node local > filesystems, not the cluster filesystem. > > I'm going to remove the setting from the wiki. > > -slowRalph, the wikipage, that Christian referred to, says to use: netbios name = something clustering = yes idmap config *?: backend = autorid idmap config *?: range = 1000000-1999999 My understanding is that autorid is meant to be used with multiple domains, yet Christian's smb.conf is definitely for a standalone server, so does he need the idmap config lines ? Rowland