We are attempting to troubleshoot a Search Results issue. We have a couple of test cases that we are working with from the client. The problem is not limited to the test cases, though.

The essential problem is that documents that should be returned for Search Criteria are not returned. The documents in question are in a folder that has permissions limiting access to Administrators and one additional Role. The client-user has this role (and I believe also had the role when I REINDEXED the entire repository a few days ago.)

To attempt to narrow down the source of the problem I have determined the following:

* ADMINISTRATOR user *can* find the documents in question via test case search criteria -- DEMONSTRATES that folder is in Search Indexes

* CLIENT-user *can* find the documents in question via ALTERNATE search criteria, -- DEMONSTRATES that client-user has SOME SORT of access to the folder in terms of the Search Indexes

As best I can tell at this point it seems like CLIENT-user can only get search hits in this folder based on the TITLE field, not the CONTENTS.

If this level of granularity is configurable I do not know where to find it to check it or change it.

If this is a defect or similar issue, I wonder if you (Peter) have had any customers experience something similar? I'm not quite finding anything in the forum but that does not mean that it is not out there...

Thanks in advance...

Hi David,

Can you send me an email with exact steps to reproduce? What I'll need is: what is the search term in the document and how exactly the forms for search were filled.

"As best I can tell at this point it seems like CLIENT-user can only get search hits in this folder based on the TITLE field, not the CONTENTS." - retry the test with something that is only in the contents and not in the title. Is the content search flawed or the title search?

Peter

Peter -- I am not certain now that Permissions actually are a part of the problem.

I'm having a lot of difficultly getting Search Results to match my expectations under a number of different scenarios and it makes the symptom gathering a little bit tricky. I'm going to lay the details for one scenario that does *not* involve any permission issues and then see if we can work from there.

I have a document that we will call TARGET. TARGET contains the two-word phrase:

"low guidance"

I also have document A and document B. Both document A and document B contain both of the words:

"low"
"guidance"

but DO NOT contain the phrase:

"low guidance"

I also have document C, which contains only the word:

"guidance"


Search | Search | Title, Keyword, Contents | guidance (no quotes)
returns A, B, C (also expect TARGET)

Search | Search | Title, Keyword, Contents | "guidance" (with quotes)
returns A, B (also expect C and TARGET)

Search | Search | Title, Keyword, Contents | low guidance (no quotes)
returns A, B (also expect TARGET)

Search | Search | Title, Keyword, Contents | "low guidance" (with quotes)
returns TARGET (*AS EXPECTED*)


Search | Advanced | Contents | Exact Phrase | low guidance (no quotes)
returns A, B (expect TARGET only, because I expect the QUOTES are implied by EXACT PHRASE)

Search | Advanced | Contents | Exact Phrase | "low guidance" (with quotes)
returns TARGET (AS EXPECTED **EXCEPT THAT**, I expected the QUOTES to implied by EXACT PHRASE and not need to be entered by me)

Is there something I am misunderstanding about the way these results should be returned?

I wonder if there is anything I can do to better see/understand the raw Lucene queries that DMX is issuing? I have Luke installed on the server.

I have verified that using Luke to query the index in the manner that I expect from the search described above. Using Luke I do get the expected results, for example I can return the document TARGET using only the search term 'guidance' -- with or without quotes.

How do I reconcile the difference between the result I see in Luke and the results I see in DMX / DNN?

Hi David,

Thanks for the elaborate example. I'll replicate this here and try to determine where this diverges from what you see in Luke.

Peter

Hi David,

I recreated the exact situation (note I'm using DMX 6.1). To test this I made 4 text documents as described each with a lorum ipsum in them and then the words low guidance as described in your post for each. The results:

Search | Search | Title, Keyword, Contents | guidance (no quotes)
returns A, B, C *and* TARGET

Search | Search | Title, Keyword, Contents | "guidance" (with quotes)
returns A, B as well as C and TARGET

Search | Search | Title, Keyword, Contents | low guidance (no quotes)
returns A, B as well as C and TARGET

Search | Search | Title, Keyword, Contents | "low guidance" (with quotes)
returns TARGET (*AS EXPECTED*)

Search | Advanced | Contents | Exact Phrase | low guidance (no quotes)
returns A, B, C and TARGET (I agree this is *not* expected and I'll see abuot fixing this)

Search | Advanced | Contents | Exact Phrase | "low guidance" (with quotes)
returns TARGET (AS EXPECTED **EXCEPT THAT**, I expected the QUOTES to implied by EXACT PHRASE and not need to be entered by me) - agreed

So from my perspective there is 1 error to resolve here, which is the "exact phrase without quotes" scenario.

Best,
Peter

Peter -- I delete all index files and reindexed.

We do not get correct results after the reindex. But the results do not look identical either.

Search | Search | Title, Keyword, Contents | guidance (no quotes)
returns NO RESULTS

Search | Search | Title, Keyword, Contents | "guidance" (with quotes)
returns NO RESULTS

Search | Search | Title, Keyword, Contents | low guidance (no quotes)
returns NO RESULTS
Search | Search | Title, Keyword, Contents | “low guidance” (with quotes)
returns TARGET only

Search | Advanced | Contents | All Words | “low guidance” (with quotes)
returns NO RESULTS

NEW SEARCH:
Search | Advanced | Contents | All Words | guidance FedEx (no quotes)
!!! returns TARGET and 13 other matches

Search | Advanced | Contents | All Words | guidance (no quotes)
!!! returns NO RESULTS

Why can I not find results on just ‘guidance’?!?!?!?!

I can get results from LUKE using just the word 'guidance' but it seems I have to include an adjacent word to get results from DMX.

How do I reconcile the difference?

You are on DMX 6.1.1, now? And (just making sure) this user has access to documents a, b, and c?

Peter

Hi Peter, I am working with David on this same problem.

The particular user referred to above has 2 security roles which allows her access to two different folders. Combined, the two different folders probably contain about 100 documents, about 75 of which contain the word "guidance." When this user logs in and goes to the document library, she sees only the folders she has access to. When she drills down into the folders, all of the documents display correctly.

However, when she uses the search function and searches for the word "guidance" in the Document Repository, only three documents display, as opposed to the expected 75 or so. And of these three documents, ONE is from a folder that this person does NOT have access to. All three documents definately contain the word "guidance." However, so do hundreds (if not thousands) of other documents in the repository as a whole, and certainly the folders she has access to contain dozens of documents that contain this word.

We have built indexes up the wazoo. We really have no idea what is going on, but this is an extremeley serious problem. Is there some way we can escalate it?

Sorry, we are on 6.1.0.

OK, once you're on 6.1.2 (should be available in a few days) let me know if this happens again.

Peter

[quote]
Posted By Kate on 10 Jan 2013 17:15:30

And of these three documents, ONE is from a folder that this person does NOT have access to.
[/quote]

This is impossible. This user must have access to the documents displayed. You can trace this by carefully examining the properties of the document. Before display the search results are combined with the permissions, the user's roles and the user's ID. We've not seen a security leak in a very long time (years IMO). So there must be some permission that links this person to the document. I'd be happy to analyze if you can grant me host access to the system and tell me the user's ID and entry ID of the document (you can see this if you "navigate" to the document in question in the querystring).

Peter