[cifs-protocol] [REG:110051857479854] RE: Replicating deleted object procedure clarifications
hongweis at microsoft.com
Tue May 25 12:29:01 MDT 2010
You are welcome. Glad to help.
Based on the documentation for RemoveObj, it always does everything all together ,including mangling object DN, updating attribute and moving to Deleted Container etc, to make the deleted object conform to the invariants specified in 18.104.22.168.5 MS-ADTS. If the attributes or values already conform to the invariants , they will not be changed. This maybe the case when it is used in UpdateObj() , where some of the attributes such as DN have been changed in PerformModifyOperation and will not been changed again.
Please let me know if you have further question on this topic.
From: kamenim at gmail.com [mailto:kamenim at gmail.com] On Behalf Of Kamen Mazdrashki
Sent: Sunday, May 23, 2010 5:28 AM
To: Hongwei Sun
Cc: cifs-protocol at samba.org; pfif at tridgell.net; MSSolve Case Email
Subject: Re: [REG:110051857479854] RE: [cifs-protocol] Replicating deleted object procedure clarifications
You are totally right and you understood my question, thanks!
The thing is that although the comment before calling RemoveObj() describes what should happen next,
RemoveObj() descriptions itself doesn't mention about this situation at all.
I think the documentation is not complete enough and RemoveObj() prototype and description should
make a distinction between 'mangling object DN, state and update attributes' and just 'update attributes'.
Now, if someone reads the description it looks like RemoveObj() will act blindly on DSNAME given,
which should lead to deleting the object, modified already by PerformModifyOperation() :)
On Sat, May 22, 2010 at 02:40, Hongwei Sun <hongweis at microsoft.com<mailto:hongweis at microsoft.com>> wrote:
For your question regarding the algorithm used in UpdateObject() , as documented in 22.214.171.124.9 of MS-DRSR, I think that the RemoveObj() was performed after PerformModifyOperation just to ensure the attributes on the deleted object conform to the invariants of a tombstone or deleted-object(see 3.1..1.5.5 of MS-ADTS). This is mentioned in the comments of the algorithm before RemoveObj is called. Furthermore, as per 5.154 of MS-DRSR, RemoveObj procedure performs a delete operation on an object so the targeted object will be transformed to a deleted-object or a tombstone ,depending on the state of the Recycle Bin option feature. This includes modifying the attributes and moving into the Deleted Container of its NC. But it will not remove the objects from the directory directly. If the object after PerformModifyOperation already conforms to the invariant of deleted-object or tombstone, the RemoveObj may do nothing to the object.
Please let me know if I understand you question properly and if you have further questions.
From: cifs-protocol-bounces at cifs.org<mailto:cifs-protocol-bounces at cifs.org> [mailto:cifs-protocol-bounces at cifs.org<mailto:cifs-protocol-bounces at cifs.org>] On Behalf Of Kamen Mazdrashki
Sent: Tuesday, May 18, 2010 7:33 AM
To: Interoperability Documentation Help; cifs-protocol at samba.org<mailto:cifs-protocol at samba.org>; pfif at tridgell.net<mailto:pfif at tridgell.net>
Subject: [cifs-protocol] Replicating deleted object procedure clarifications
I am currently trying to refactor Delete object implementation in Samba and I need some help
with algorithm used for deleting objects and how the deletion is replicated to other DCs.
ProcessGetNCChangesReply [MS-DRSR] - http://msdn.microsoft.com/en-us/library/dd207758(v=PROT.13).aspx
UpdateObject [MS-DRSR] - http://msdn.microsoft.com/en-us/library/dd207780(v=PROT.13).aspx
Delete Operation [MS-ADTS] - http://msdn.microsoft.com/en-us/library/cc223480(v=PROT.13).aspx
Consider following sutiation:
0. We have two DCs configuration.
We have an OU object with following props:
1. We delete this OU on DC1. The state of this object on each dc should be as follows:
dn: OU=TEST_DELETE_0417\0ADEL:b7f1b90d-d247-45b7-87fb-f6bc916bd729,CN=Deleted Objects,DC=samba,DC=devel
2. Replication is triggered from DC1 to DC2.
Now, according to UpdateObject() procedure, we will identify that Object's DN has changed from
to "dn: OU=TEST_DELETE_0417\0ADEL:b7f1b90d-d247-45b7-87fb-f6bc916bd729,CN=Deleted Objects,DC=samba,DC=devel".
Hence we will modify object's DN (calling PerformModifyDNOperation() operation).
Which will make this object a Deleted-object right?
While progressing further in UpdateObject() procedure, we will check and see, that 'isDeleted'
attribute value is TRUE, so we shall call RemoveObj() procedure. At this point I am a little bit puzzled as
there are two possible outcomes from this procedure:
1. Object's RDN should be transformed to a delete-mangled RDN. So we should end with an RDN like:
2. Or, as the object is already under "Deleted Object" container (moved there by previous call to PerformModifyOperation()),
RemoveObj() procedure should delete it further - i.e. if the object is a Tombstone, it will be completely removed.
Sorry that my description gets so messy.
Basically what UpdateObject() states is that first we should execute PerformModifyOperation() and then RemoveObj().
Which is a little bit confusing, as PerformModifyOperation() will turn the object into a Deleted-object.
Calling RemoveObj() later will actually act on already modified object, so I wonder - how does RemoveObj()
knows that we just converted the object and this object should not be completely removed?
Another possibility is for PeformModifyOperation() to determine that target DN will move the object under
"Deleted Objects" container, and in this case to modify only Attribute values on the object, but not to call
kamen.mazdrashki at postpath.com<mailto:kamen.mazdrashki at postpath.com>
CISCO SYSTEMS BULGARIA EOOD
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cifs-protocol