Bug 8669 - upgradeprovision --full fails to find CN=NTDS Settings
Summary: upgradeprovision --full fails to find CN=NTDS Settings
Status: RESOLVED FIXED
Alias: None
Product: Samba 4.0
Classification: Unclassified
Component: Tools (show other bugs)
Version: unspecified
Hardware: x86 Linux
: P5 normal (vote)
Target Milestone: ---
Assignee: Matthieu Patou
QA Contact: samba4-qa@samba.org
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-12-18 16:45 UTC by Michael Wood
Modified: 2012-05-09 11:59 UTC (History)
1 user (show)

See Also:


Attachments
Add debug to the upgrade provision (755 bytes, patch)
2012-01-01 18:53 UTC, Matthieu Patou
no flags Details
upgradeprovision --full --debugall (77.25 KB, text/plain)
2012-01-05 11:00 UTC, Michael Wood
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Michael Wood 2011-12-18 16:45:02 UTC
> I'm trying to upgrade an old installation of Samba 4 from
> 4.0.0alpha12-GIT-77b9b97 to 4.0.0alpha18-GIT-35605fa.
>
> The build works fine.
>
> samba-tool dbcheck --fix finds and fixes a bunch of things.

The vast majority of the fixes are like this:

ERROR: missing GUID component for objectCategory in object
CN=System,DC=example,DC=com -
CN=Container,CN=Schema,CN=Configuration,DC=example,DC=com
Change DN to <GUID=47288117-6896-4151-b2af-4921180a3ccf>;CN=Container,CN=Schema,CN=Configuration,DC=example,DC=com?
[YES]
Fixed missing GUID on attribute objectCategory

and it ends like this:

ERROR: dsServiceName not in GUID form in @ROOTDSE
Change dsServiceName to GUID form? [y/N] y
Changed dsServiceName to GUID form
Checked 870 objects (226 errors)

If I run it again immediately I get this:

# samba-tool dbcheck --fix
Checking 870 objects
Fix isDeleted originating_change_time on 'CN=Deleted
Objects,DC=example,DC=com' [y/N/all/none] y
Checked 870 objects (1 errors)
# samba-tool dbcheck --fix
Checking 870 objects
Fix isDeleted originating_change_time on 'CN=Deleted
Objects,DC=example,DC=com' [y/N/all/none] y
Checked 870 objects (1 errors)

So although it asks if I want to fix that error, it can't seem to do it.

And with -d10:

# samba-tool dbcheck --fix --yes -d10
INFO: Current debug levels:
[...]
lpcfg_load: refreshing parameters from /usr/local/samba/etc/smb.conf
params.c:pm_process() - Processing configuration file
"/usr/local/samba/etc/smb.conf"
Processing section "[globals]"
Processing section "[netlogon]"
Processing section "[sysvol]"
pm_process() returned Yes
Security token SIDs (1):
 SID[  0]: S-1-5-18
 Privileges (0xFFFFFFFFFFFFFFFF):
[...]
 Rights (0x               0):
lpcfg_servicenumber: couldn't find ldb
schema_fsmo_init: we are master[yes] updates allowed[no]
Checking 870 objects
Fix isDeleted originating_change_time on 'CN=Deleted
Objects,DC=example,DC=com' [YES]
ldb:acl_modify: isDeleted
Sorting rpmd with attid exception 3 rDN=CN DN=CN=Deleted
Objects,DC=example,DC=com
Checked 870 objects (1 errors)

> upgradeprovision completes without complaint.

This is what it actually says:

# upgradeprovision
Creating a reference provision
No IPv6 address will be assigned
Copy privilege
Update base samdb by searching difference with reference one
You still have the old DNS object for managing dynamic DNS, but you
didn't supply --full so a correct update can't be done

so no errors, but I'm not sure if it did anything.
- Show quoted text -
With the following workaround upgradeprovision --full completes, but
obviously hasn't fixed everything it should have:

--- a/sbin/upgradeprovision
+++ b/sbin/upgradeprovision
@@ -886,7 +886,7 @@ def checkKeepAttributeWithMetadata(delta, att, message, refe
    dn = current[0].dn

    for att in list(delta):
-        if att in ["dn", "objectSid"]:
+        if att in ["dn", "objectSid", "msDS-hasMasterNCs"]:
            delta.remove(att)
            continue

# upgradeprovision --full
Creating a reference provision
No IPv6 address will be assigned
Copy privilege
Update base samdb by searching difference with reference one
Starting update of samdb
There are 76 missing objects
Reloading a merged schema, which might trigger reindexing so please be patient
Schema reloaded!
There are 7 changed objects
Update of samdb finished
Update of secrets.ldb
IMPORTANT!!! If you were using Dynamic DNS before you need to update
your configuration, so that the tkey-gssapi-credential has the
following value: DNS/samba.example.com
Update machine account
Some defaultSecurityDescriptors and/orsecurityDescriptor have changed,
recalculating SD
Unable to set ACLs on policies related objects: an integer is required
Upgrade finished!
Reopenning samdb to trigger reindexing if needed after modification
Reindexing finished

Also, the message about not being able to set ACLs on policies related
to objects looks like it might be a problem.

Running dbcheck again after the upgradeprovision finds a bunch more GUID errors:

# samba-tool dbcheck --fix --yes
Checking 944 objects
ERROR: incorrect GUID component for objectCategory in object
DC=l.root-servers.net,DC=RootDNSServers,CN=MicrosoftDNS,DC=DomainDnsZones,DC=example,DC=com
- <GUID=aae56a27-1de9-4188-afff-3a05b57aef73>;CN=Dns-Node,CN=Schema,CN=Configuration,DC=example,DC=com
Change DN to <GUID=cad206ef-f29f-4d88-8822-4b30cf4aef1b>;CN=Dns-Node,CN=Schema,CN=Configuration,DC=example,DC=com?
[YES]
Fixed incorrect GUID on attribute objectCategory
[...]
Checked 944 objects (83 errors)

And re-running dbcheck immediately finds another 2 errors (in addition
to the one from before) that it doesn't fix:

# samba-tool dbcheck --fix --yes
Checking 944 objects
ERROR: missing GUID component for wellKnownObjects in object
DC=DomainDnsZones,DC=example,DC=com -
B:32:18E2EA80684F11D2B9AA00C04F79F805:CN=Deleted
Objects,DC=DomainDnsZones,DC=example,DC=com
unable to find object for DN CN=Deleted
Objects,DC=DomainDnsZones,DC=example,DC=com - (No such Base DN:
CN=Deleted Objects,DC=DomainDnsZones,DC=example,DC=com)
Not removing dangling forward link
ERROR: missing GUID component for wellKnownObjects in object
DC=ForestDnsZones,DC=example,DC=com -
B:32:18E2EA80684F11D2B9AA00C04F79F805:CN=Deleted
Objects,DC=ForestDnsZones,DC=example,DC=com
unable to find object for DN CN=Deleted
Objects,DC=ForestDnsZones,DC=example,DC=com - (No such Base DN:
CN=Deleted Objects,DC=ForestDnsZones,DC=example,DC=com)
Not removing dangling forward link
Fix isDeleted originating_change_time on 'CN=Deleted
Objects,DC=example,DC=com' [YES]
Checked 944 objects (3 errors)
Comment 1 Matthias Dieter Wallnöfer 2011-12-18 19:04:28 UTC
Reassigning as desired on the technical mailing list.
Comment 2 Michael Wood 2012-01-01 13:37:59 UTC
Where I pasted "- Show quoted text -" in the description of the problem, it should have had the following:

> But upgradeprovision --full fails as follows (with some debug print
> statements added just before the failure):
>
> Creating a reference provision
> No IPv6 address will be assigned
> Copy privilege
> Update base samdb by searching difference with reference one
> Starting update of samdb
> There are 76 missing objects
> Reloading a merged schema, which might trigger reindexing so please be patient
> Schema reloaded!
> dn= CN=NTDS Settings,CN=SAMBA,CN=Servers,CN=Default-First-Site-Name,CN=Sites,CN=Configuration,DC=example,DC=com
> basedn= DC=example,DC=com
> attrs= msDS-hasMasterNCs
> res.count == 0
> Exception during upgrade of samdb:
> Traceback (most recent call last):
>  File "/usr/local/samba/sbin/upgradeprovision", line 1205, in update_partition
>    provisionUSNs)
>  File "/usr/local/samba/sbin/upgradeprovision", line 1080, in update_present
>    basedn, usns, samdb)
>  File "/usr/local/samba/sbin/upgradeprovision", line 904, in
> checkKeepAttributeWithMetadata
>    curval, refval, delta)
>  File "/usr/local/samba/sbin/upgradeprovision", line 788, in handle_links
>    for e in res[0][att]:
> IndexError: list index out of range
> Update failed
> Rolling back all changes. Check the cause of the problem
> Your system is as it was before the upgrade
>
> So it seems that this search for CN=NTDS Settings returns nothing:
>
>    res = samdb.search(expression="dn=%s" % dn, base=basedn,
>                        controls=["search_options:1:2", "reveal:1"],
>                        attrs=[att])
>
> The strange thing is that if I do the same search from the command
> line it works fine:
>
> # ldbsearch -H private/sam.ldb --controls="search_options:1:2
> reveal:1" "dn=CN=NTDS
> Settings,CN=SAMBA,CN=Servers,CN=Default-First-Site-Name,CN=Sites,CN=Configuration,DC=example,DC=com"
> msDS-hasMasterNCs
> # record 1
> dn: CN=NTDS Settings,CN=SAMBA,CN=Servers,CN=Default-First-Site-Name,CN=Sites,CN=Configuration,DC=example,DC=com
> msDS-hasMasterNCs: CN=Configuration,DC=example,DC=com
> msDS-hasMasterNCs: CN=Schema,CN=Configuration,DC=example,DC=com
> msDS-hasMasterNCs: DC=example,DC=com
Comment 3 Matthieu Patou 2012-01-01 18:52:50 UTC
Hello Michael,

Can you try to apply the following patch, it should help me to understand what are the difference.

Can you also do a ldbsearch with --show-deleted --reveal on the CN=NTDS ... object for msDS-hasMasterNCs.

Can you also open another bug report for the dbcheck not being able to fix problems in the DB ?

Thanks.
Comment 4 Matthieu Patou 2012-01-01 18:53:15 UTC
Created attachment 7218 [details]
Add debug to the upgrade provision
Comment 5 Michael Wood 2012-01-01 22:31:05 UTC
(In reply to comment #3)
> Hello Michael,
> 
> Can you try to apply the following patch, it should help me to understand what
> are the difference.

This is what it prints:

[...]
Reloading a merged schema, which might trigger reindexing so please be patient
Schema reloaded!
Current provision: CN=Configuration,DC=example,DC=com
Current provision: CN=Schema,CN=Configuration,DC=example,DC=com
Current provision: DC=example,DC=com
Reference provision: CN=Configuration,DC=example,DC=com
Reference provision: CN=Schema,CN=Configuration,DC=example,DC=com
Reference provision: DC=example,DC=com
Reference provision: DC=ForestDnsZones,DC=example,DC=com
Reference provision: DC=DomainDnsZones,DC=example,DC=com
Exception during upgrade of samdb:
[...]

> Can you also do a ldbsearch with --show-deleted --reveal on the CN=NTDS ...
> object for msDS-hasMasterNCs.

Perhaps what you're looking for here is in comment 3?

This returns nothing:

# ldbsearch -H private/sam.ldb --show-deleted --reveal "CN=NTDS Settings" msDS-hasMasterNCs
# Referral
ref: ldap://example.com/CN=Configuration,DC=example,DC=com

# returned 1 records
# 0 entries
# 1 referrals

but using --controls="search_options:1:2" shows it as in comment 3:

dn: CN=NTDS Settings,CN=SAMBA,CN=Servers,CN=Default-First-Site-Name,CN=Sites,CN=Configuration,DC=example,DC=com
msDS-hasMasterNCs: CN=Configuration,DC=example,DC=com
msDS-hasMasterNCs: CN=Schema,CN=Configuration,DC=example,DC=com
msDS-hasMasterNCs: DC=example,DC=com

> Can you also open another bug report for the dbcheck not being able to fix
> problems in the DB ?

OK thanks.
Comment 6 Michael Wood 2012-01-03 21:36:14 UTC
A few days ago I updated to e39df67669f61056692736db9c8dc16fbf2c3624 and the problem was still there (of course).  I have now added the debugging patch (attachment 7218 [details]) and also these three commits:

f66ef5cfbc932dc03a5bea61e9cb10dd8d948128
f05edc0ecb9da2cb00a83b38d0be5812cc4ccf77
3213d1e0b770690b1a964f38fb57ebbcd8ce0746

The upgradeprovision no longer gets stuck on the NTDS Settings.  It also did not print the debugging info like it did in comment 5.  Does this mean that this issue is likely resolved now?

I notice that it still complains about not being able to set ACLs on policy-related objects.  Should I open another bug report about that?

Creating a reference provision
No IPv6 address will be assigned
Copy privilege
Update base samdb by searching difference with reference one
Starting update of samdb
There are 76 missing objects
Reloading a merged schema, which might trigger reindexing so please be patient
Schema reloaded!
There are 3 changed objects
Update of samdb finished
Update of secrets.ldb
IMPORTANT!!! If you were using Dynamic DNS before you need to update your configuration, so that the tkey-gssapi-credential has the following value: DNS/samba.example.com
Update machine account
Some defaultSecurityDescriptors and/orsecurityDescriptor have changed, recalculating SD 
Unable to set ACLs on policies related objects: an integer is required
Upgrade finished!
Reopenning samdb to trigger reindexing if needed after modification
Reindexing finished

Incidentally, if I run dbcheck after this it again fixes a large number of issues, but now has 3 issues it is apparently unable to fix instead of just 1.
Comment 7 Matthieu Patou 2012-01-05 06:48:14 UTC
(In reply to comment #6)
> A few days ago I updated to e39df67669f61056692736db9c8dc16fbf2c3624 and the
> problem was still there (of course).  I have now added the debugging patch
> (attachment 7218 [details]) and also these three commits:
> 
> f66ef5cfbc932dc03a5bea61e9cb10dd8d948128
> f05edc0ecb9da2cb00a83b38d0be5812cc4ccf77
> 3213d1e0b770690b1a964f38fb57ebbcd8ce0746
> 
> The upgradeprovision no longer gets stuck on the NTDS Settings.  It also did
> not print the debugging info like it did in comment 5.  Does this mean that
> this issue is likely resolved now?
> 
Yes and no, There was a 2 bugs, one easy and one not so easy.
The thing is that there is an issue with DNS container not being created.

So I would not recommend using it for the moment.

> I notice that it still complains about not being able to set ACLs on
> policy-related objects.  Should I open another bug report about that?
> 
Well I'll try to have a quick look it seems that you provided me some info so I could slip the change pretty quickly in master tree.

> Incidentally, if I run dbcheck after this it again fixes a large number of
> issues, but now has 3 issues it is apparently unable to fix instead of just 1.
What are those issues did you open a bug for it ?
Comment 8 Michael Wood 2012-01-05 09:55:09 UTC
(In reply to comment #7)
> > The upgradeprovision no longer gets stuck on the NTDS Settings.  It also did
> > not print the debugging info like it did in comment 5.  Does this mean that
> > this issue is likely resolved now?
> > 
> Yes and no, There was a 2 bugs, one easy and one not so easy.
> The thing is that there is an issue with DNS container not being created.
> 
> So I would not recommend using it for the moment.

OK.

> > I notice that it still complains about not being able to set ACLs on
> > policy-related objects.  Should I open another bug report about that?
> > 
> Well I'll try to have a quick look it seems that you provided me some info so I
> could slip the change pretty quickly in master tree.

Thanks.

> > Incidentally, if I run dbcheck after this it again fixes a large number of
> > issues, but now has 3 issues it is apparently unable to fix instead of just 1.
> What are those issues did you open a bug for it ?

I added them to the existing bug report for dbcheck not being able to fix the other issue, although in this case it doesn't seem to try to fix it.  (bug 8683)

Here's what I get:

# samba-tool dbcheck --fix --yes
Checking 944 objects
ERROR: missing GUID component for wellKnownObjects in object
DC=DomainDnsZones,DC=example,DC=com -
B:32:18E2EA80684F11D2B9AA00C04F79F805:CN=Deleted
Objects,DC=DomainDnsZones,DC=example,DC=com
unable to find object for DN CN=Deleted
Objects,DC=DomainDnsZones,DC=example,DC=com - (No such Base DN: CN=Deleted
Objects,DC=DomainDnsZones,DC=example,DC=com)
Not removing dangling forward link
ERROR: missing GUID component for wellKnownObjects in object
DC=ForestDnsZones,DC=example,DC=com -
B:32:18E2EA80684F11D2B9AA00C04F79F805:CN=Deleted
Objects,DC=ForestDnsZones,DC=example,DC=com
unable to find object for DN CN=Deleted
Objects,DC=ForestDnsZones,DC=example,DC=com - (No such Base DN: CN=Deleted
Objects,DC=ForestDnsZones,DC=example,DC=com)
Not removing dangling forward link
Fix isDeleted originating_change_time on 'CN=Deleted Objects,DC=example,DC=com'
[YES]
Checked 944 objects (3 errors)

so perhaps this is related to not creating the DNS containers you mentioned above.
Comment 9 Michael Wood 2012-01-05 11:00:38 UTC
Created attachment 7230 [details]
upgradeprovision --full --debugall
Comment 10 Michael Wood 2012-01-07 23:54:47 UTC
The original provision is of course very old and uses bind9 flat files instead of DLZ or the built-in name server.

I suppose upgradeprovision should just upgrade and keep the flat file DNS scheme.

I would like to migrate to DLZ, but would understand if that was a separate process or if you needed to use something like "upgradeprovision --migrate-dns-to-dlz".

Would it be simplest now to get upgradeprovision to check if the old provision is using flat files/DLZ and upgrade to the same method?
Comment 11 Matthias Dieter Wallnöfer 2012-05-04 08:35:12 UTC
Still an issue?
Comment 12 Michael Wood 2012-05-04 08:50:50 UTC
(In reply to comment #11)
> Still an issue?

Well, I haven't used upgradeprovision again because as far as I know it's still broken.  But it seems that the issue with NTDS Settings was resolved as per comment 6, so maybe this bug should be closed.
Comment 13 Matthias Dieter Wallnöfer 2012-05-09 11:59:11 UTC
I mark this as fixed, feel free to reopen if needed.