I issued a query, and it didn’t match a document that I know exists. Why?
Sometimes the Files to be filtered counter is greater than the Total # documents counter. How is that possible?
How do I handle a large number of documents in my corpus if I have a powerful computer?
There are a lot of event log messages. Can I turn them off?
Why do I get the event Very low disk space… even after freeing disk space?
How much disk space should I allocate for Index Server data?
My computer is idle but the Files to be filtered counter is not zero. What is the reason?
Why are all my floating point numbers formatted with the same number of decimal places?
Why do unreadable files in Microsoft® FrontPage® _vti directories show up in my query results?
How can I exclude certain directories from search results?
How can I index remote UNC shares?
Why does Index Server not index remote virtual roots?
When a query is sent and the message "Query too expensive" is returned, what can be done to solve the problem?
What can be done if I have just installed an application and suddenly users start seeing extra characters (and possibly HTML code, depending on the application) in the summaries on the results page of a query?
- Different languages were used to filter the page and issue the query. The split of a page into words (the fundamental unit stored in the index) is a language-dependent process. When a query is issued from one language and the document was filtered in another, query results can be unpredictable. The locale used to filter a document is dependent on the filter. Some file formats, such as Microsoft Word, mark documents with a language; this mark is used during filtering. Other formats, such as plain text, contain no language specifier. Most filters default to the system locale for these files. The locale for the query is specified by use of the CiLocale variable. If CiLocale is not specified, the locale of the browser is used (if available) or the default locale of the server.
- The page has not yet been filtered. There is often a delay between the time a page is modified and the time it is indexed. You can use the Performance Monitor or an administrative query to determine the number of pending pages. The CiOutOfDate variable is set to TRUE if one or more files remain to be filtered.
- An error occurred during filtering of this document. There are a number of reasons why a document may not have been filtered. See the filtering page for a detailed discussion of filtering. You can check for errors either in the event log or by issuing a query for the unfiltered property by using the administrative query for unfiltered files.
- The query was too complex. Certain classes of resource-intensive queries are turned off when CiForceUseCi is set to TRUE. When results are missing for this reason, the CiQueryIncomplete variable is set to TRUE.
- The query took too much CPU time. An administrator can limit the amount of CPU time a single query can consume. When this limit is reached, the CiQueryTimedOut variable is set to TRUE.
- You do not have permission to read the page. A query result will not be returned for a page that you do not have permission to read. See Security for a detailed discussion of security as it relates to Index Server.
- Some files have been modified more than once before being filtered. If a file is modified multiple times, there may be multiple entries for the same file. The Files to be filtered counter only specifies how big the change queue is.
- If there is sufficient RAM (more than 48 MB) and the MaxFreshCount registry parameter is 5000 or less, try increasing the MaxFreshCount to 10,000 or 20,000. If your hard disk is overworking, try adding more RAM to your computer.
- If this event is logged during a MasterMerge, freeing up additional space will help. During a master merge, additional free disk space (up to 15 percent of the size of all the documents that are being indexed) may be required. After the master merge completes, disk space is usually freed up.
- The minimum disk space allocated should be at least 30 percent of the size of your corpus. If you increase the registry
value of MaxCharacterization , then more space should be allocated. Temporarily during a master merge, Index Server may need up to 45 percent of the corpus size.
- Some files failed to filter. If a file is in use by another user process when the CiDaemon process tries to filter its contents, the file is moved to a lower-priority queue and will be retried for filtering at a later time. The interval between the retries is controlled by the registry parameter FilterRetryInterval. This typically happens with applications such as Microsoft Word that keep the file open for exclusive access and fast saves.
- Index Server uses the settings specified in the Regional Settings application of Control Panel. Users from different locales will see a view of the data specific to their locale. U.S. users may see a period used as a decimal point, and French users would see a comma, for example. Likewise date, time, and currency values are formatted according to the parameters in Regional Settings. If you are using .asp files and VBScript for queries, you can control decimal precision with the FormatNumber function.
- This is a limitation of scoping in Index Server. You can disable indexing on a virtual directory through the IIS Internet Service Manager by clearing the Index this directory check box in the Directories property sheet. Alternatively, you can modify the user’s restriction. For example, to prevent FrontPage directories from being indexed, change the CiRestriction in the .idq file to:
CiRestriction=(%UserRestriction%) &! #vpath *\_vti_*.
To display just the user's query in the result page, add the following restriction to the .idq file:
UserRestriction=%UserRestriction%
In the .htx file, add the following line:
<%UserRestriction%>
- Indexing and searching in Index Server is based on virtual roots, and you can exclude virtual directories from indexing through the IIS Internet Service Manager by clearing the Index this directory check box on the Directory property sheet. Alternatively, in your .idq files you can specify files to be prevented from appearing in search results. For example, if the virtual root /Docs points to C:\Docs but you do not want results from C:\Docs\Private to appear in the result set, you can modify the CiRestriction parameter in your .idq file as follows:
CiRestriction=%CiRestriction% AND NOT #path c:\docs\private*
You can append more AND NOT #path path* clauses to the CiRestriction to exclude more directories.
- If you add a virtual root pointing to the remote UNC share, it will automatically be indexed by Index Server. Make sure that you specify the user ID (domain\user) and password correctly. Note the backslash (\) in the domain\user format of the user ID.
When specifying the logon ID for a remote virtual root, type both the domain name and the user name, separated by a backslash (\):
domain\username
If you do not give the domain name, Index Server will not index the remote virtual roots. Note that the domain name may actually be the name of the computer, if the account is local to that computer.
- When you install certain applications, such as Microsoft® Office 97, they will overwrite some of the registry settings for Index Server. To restore the Index Server registry settings, you should reinstall Index Server.
© 1997 by Microsoft Corporation. All rights reserved.