[Pkg-samba-maint] Bug#603729: Repeated winbind_cache.tdb corruption in Samba 3.5.x

[Dale Schroeder]

Package: winbind
Version: 2:3.5.6~dfsg-1

Samba 3.5.6~dfsg-1 on Debian Squeeze.  Updates are current.

Ever since upgrading from 3.4.x to 3.5.x, winbind_cache.tdb is 
repeatedly corrupted, rendering authentication impossible.

I have a script that checks every 15 minutes for connectivity to the AD 
server and restarts samba and winbind if the connection is corrupted. 
This happens on every winbind system, so it is not system specific.

ls of this file in /var/cache/samba is as follows:
-rw-------  1 root root 139264 Nov 16 11:21 winbindd_cache.tdb
-rw-------  1 root root 946176 Nov 16 11:20 winbindd_cache.tdb.bak
-rw-------  1 root root 946176 Nov 16 09:16 winbindd_cache.tdb.bak.old

The corrupted tdb's are almost 7 times as large as the working version. 
  In 3.3.5, a winbind_cache.corrupt was generated at each restart after 
a corruption.  In 3.5.6, 2 backups are maintained instead.  This is the 
only difference I have noted between the two versions.

I notice this is repeated in the logs at the time of the corruption:

[2010/11/16 11:19:40.933366, 10] 
winbindd/winbindd_cache.c:4674(wcache_fetch_ndr)
    Entry has wrong sequence number: 1573542
[2010/11/16 11:19:40.933684,  1] winbindd/winbindd_util.c:289(trustdom_recv)
    Could not receive trustdoms

A level 10 log during a corruption is attached.

Dependencies:
adduser                        3.112
libc6                          2.11.2-7
libcomerr2                     1.41.12-2
libkrb53                       1.6.dfsg.4~beta1-13
libldap-2.4-2                  2.4.23-6
libpam0g                       1.1.1-6.1
libpopt0                       1.16-1
libtalloc1                     1.2.0~git20080616-1
libwbclient0                   2:3.5.6~dfsg-1
lsb-base                       3.2-23.1
samba-common                   2:3.5.6~dfsg-1

Also:
linux-image-2.6.32-5-686       2.6.32-27

smb.conf:

[global]
	workgroup = DOMAIN
	realm = DOMAIN.COM
	server string = %h server (Samba %v)
	security = ADS
	allow trusted domains = No
	map to guest = Bad User
	obey pam restrictions = Yes
	password server = ad_dc
	passdb backend = tdbsam
	username map = /etc/samba/users.map
	log level = 1 winbind:10 idmap:4
	log file =/var/log/samba/%m
	max log size = 1000
	name resolve order = wins host bcast
	deadtime = 15
	load printers = No
	printcap name = cups
	wins proxy = Yes
	wins server = 192.168.x.xxx
	ldap ssl = no
	panic action = /usr/share/samba/panic-action %d
	#idmap backend = rid:DOMAIN=1000-20000000
	idmap backend = tdb
	idmap uid = 1000-20000000
	idmap gid = 1000-20000000
	idmap config DOMAIN : backend	= rid
	idmap config DOMAIN : range		= 1000 - 20000000
	template homedir =/home/domain/%U
	template shell = /bin/bash
	winbind cache time = 10
	winbind enum users = Yes
	winbind enum groups = Yes
	winbind use default domain = Yes
	winbind offline logon = Yes
	admin users = root, DOMAIN\user1, "@DOMAIN\group1"
	ea support = Yes
	map archive = No
	map readonly = no
	store dos attributes = Yes

I found that the corruptions occurred less frequently if I converted to 
the newer idmap_rid syntax; however, they still occur at the rate of 5 
to 6 daily.

Thanks,
Dale Schroeder

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: debug1.txt
URL: <http://lists.alioth.debian.org/pipermail/pkg-samba-maint/attachments/20101116/93dac315/attachment-0001.txt>