Summary of FAST Search Server 2010 for SharePoint SP1 improvements and fixes

Summary of FAST Search Server 2010 for SharePoint SP1 improvements and fixes

Table of Contents



This is a summary of the improvements and fixes in FAST Search for SharePoint Service Pack 1. The list contains some additional info compared to the excel sheet published with the SP1 update.

SP1 includes new fixes and improvements, but also includes the previous cumulative updates. For details on important fixes in the CUs, see FAST Search for SharePoint Cumulative Updates.

Note that some fixes (such as the scope fix introduced in the August 2011 CU) are introduced in the SharePoint Server SP1 rather than FAST Search Server 2010 for SharePoint SP1. You should always install both service packs.

For details on how to install service packs and cumulative updates, see Install a software update (FAST Search Server 2010 for SharePoint).

SP1 improvements

  • More flexible property extraction. For details, see Creating a custom property extractor.
  • Titles and dates for Microsoft Office documents in search results do not correctly reflect the contents of the documents. For details on new SP1 configuration option, see Configuring Optional Item Processing.
  • Service Pack 1 includes new deployment configuration file templates that better reflects recommended deployment scenarios.
  • Added support for Greek stemming (lemmatization) and spell checking. For an updated list of supported languages, see Linguistic features per language.

SP1 fixes

  • Multiple fixes related to feeding and query performance. This is particularly important for large deployments with many index columns. SP1 is therefore required if you want to deploy an XL size farm as described in Test results: Extra-large scenario. The test results in that article is based on a SP1 deployment.
    • Prior to SP1, the maximum query throughput (QPS) is roughly 50 divided by the number of columns (for example, a 10 column system would have a throughput of 5).
    • Fast Search feeding and indexing hangs when feeding recovery is triggered and there are many content feeding sessions running at the same time. This may happen frequently if you have many index columns.
  • SP1 removes lowercase naming restrictions on query interface and web parts:
    • Refiner names must be specified in lowercase in the refinement panel configuration, query web service or query OM. This fix ensures that you can specify the correct index schema name, even if it includes a leading uppercase, e.g. 'Write'.
    • Consider a scope that has a FAST specific scope filter for which the filter string contains one or more managed property restrictions, and the managed property names contain uppercase letters. When this scope is used in a query, the error message, "Property doesn't exist or is used in a manner inconsistent with schema settings" is returned.
  • The FAST Search for SharePoint October CU introduced an important fix for full backup and restore. This fix introduces a new -multithreaded option to the backup and restore scripts that you should use. For details, see Full backup and restore.
  • The default schema settings produced irrelevant results because unnecessary content is indexed in the full-text index. The default schema also produces inconsistent SiteName values between indexed content from the SharePoint and FAST Lotus Notes connectors. Before installing SP1 you may see that queries like 'http', 'sharepoint', 'html' may match all items in the index due to too much metadata mapped to the full-text index. See also this blog post.
  • When a user tries to install Fast Search Server 2010 for SharePoint using a SQL server, which has been configured to be accessed using SQL authentication, Post setup fails in database creation step. This fix addresses the above issue and the Service Pack must be applied before the post setup. For scripted install, two new parameters are added to the psconfig command-line tool: sqlUserName and sqlPassword. For details, see Run scripted installation and psconfig command-line reference.
  • Keyword Syntax queries with a non-default full-text index returns zero results.
  • If you try to create a new rank profile that has an illegal name (such as a name that contains angle brackets), the schema update will fail on the FAST Search servers, but you will not receive an error message in the PowerShell console. For details, see Problems deploying an index schema update (FAST Search Server SP1 fix).
  • PowerPoint SlideShow (.ppsx) files are not searchable when you use the latest Microsoft Filter Pack. For details, see IFilter and file type reference.
  • Feeding may fail after customization of the inclusion dictionary for a property extractor.
  • Indexer ignores errors when it copies the index to the backup indexer server or additional search rows.
  • Indexer crashes and corrupts the document index.
  • You cannot remove crawled property mappings for a managed property in SharePoint Central Admin.
  • The indexer goes into an invalid state and does not shut down.
  • The Configuration process (PSConfig) throws an error and terminates at resourcesstoreinstall.exe execution.
  • The "total size of documents (bytes)" value that is reported by the Indexer process does not decrease correctly when documents are removed from the search index.
  • Document preview does not work in search results when https is used.
  • When an I/O exception occurs during index activation, the system enters a state in which no more data is indexed.
  • The terms in simplified Chinese language multi-term queries are automatically combined into a phrase.
  • When a user is authenticated by a third-party system through form-based authentication, search results do not list documents to which anonymous users have read access.
  • FAST scope filters are dropped if they are longer than the maximum size, but no warning message is returned. In this situation after the fix, the filter is reverted to the previously configured scope filter (if exists), and a warning is displayed in the user interface.
  • An uncaught exception makes the click-through log analysis stop processing new data.
  • The primary indexer becomes unresponsive during long index copies.
  • If the click-through analysis engine fails to write its internal state to disk, the file that contains the state is truncated, and the application cannot restart.
  • The indexer can run out of disk space when old fixml data is not cleaned up.
  • Backup indexer or query matching may fail when disk space is low.
  • Search fails until samworker or qrproxy services are restarted.
  • In cases where indexing fails due to corrupt data, then new content will not become searchable.
  • After startup, the backup indexer gets out of sync with the primary indexer with respect to index copying. As a result, the backup indexer does not have the necessary indexes available in case of failover, and everything will have to be re-indexed. Error logs may include this message: "Communication error when registering file receiver from ...: File receiver unavailable."
  • The click-through analysis engine (SPRel) breaks when error handling is triggered during feeding.
  • The link and click-through analysis components will not work if you set up a multi-server installation for which the hostnames in the deployment file use different casing than the hostnames that are given to the Post Setup Configuration script. This restriction was previously mentioned in Configure a multiple server deployment.
  • Corruption of the search index stops deletion of expired normalized index dictionaries. Those files might never be cleaned up, wasting disk space.
  • In a fault tolerant setup under certain circumstances when the amount of available memory is low on the primary indexer node, the primary indexer shuts down. This can cause the backup indexer to get our of sync.
  • Installation on the Italian language version of Windows 2008 Server fails in the post-setup configuration stage.
  • The infotool command line tool does not pick up the pipeline extensibility and schema files from the administration server. This information is very useful for troubleshooting schema and item processing scenarios.
  • Information on the backup and crawler configuration are missing from the infotool report when FAST Search Server is installed in a folder whose name includes a white space character.
  • When you run the SecureFASTSearchConnector.ps1 script in silent mode, a message is printed to the console that states that the script was not successful even if it was successful. This fix corrects that problem and resolves a problem of several log messages being printed to the console while the script is run in silent mode.
  • When the user stops a process by using the command line tool nctrl, the message "Failed to stop the following process" may be displayed, However, the process might just be taking a long time to stop.
  • The dump file (.dmp) that is created when Contentdistributor, Indexingdispatcher, or Nameserver crashes is unusable.
  • Other performance related fixes:
    • The fsearch process grows in memory size until all system memory is used under a sustained heavy query load.
    • Feeding performance for large batches, with many links per document or many documents per batch, is unexpectedly poor.
    • Administrators experience performance drops in feeding due to the overhead of error-handling and resubmitting of batches.
    • Temporary index directories that remain when the search controller is stopped are not cleaned up, which wastes disk space.
    • When an item processor cannot send link data to the link analysis engine, the item processor stores the data until it can be sent. In some cases, the data is never sent, performance suffers, and errors are logged.
  • Fixes related to the FAST specific connetors:
    • FAST Search Web Crawler: In some scenarios, the crawler crawls documents in the wrong order and may not return the expected results.
    • FAST Search Web Crawler: If the user updates the crawler configuration on a multi-node crawler setup that specifies a proxy for crawling, the crawling comes to a halt and all requests fail. This leads to no documents being updated or added to the index.
    • FAST Search Web Crawler: If you have a crawl that uses sitemaps as start URIs, you may observe that certain documents are removed from the index by the crawler after a number of refresh cycles, even though these documents still exist on the web server and are also still linked by the sitemap.
    • FAST Search Web Crawler: An error occurs when you use the command line tool "crawleradmin" together with the "--nodestatus" option to query the status of multiple web crawlers.
    • The FAST Search Web Crawler becomes unresponsive when it is looking up DNS host entries for a large set of websites. The fix introduces a new option 'dns_resolver_threads' in CrawlerGlobalDefaults.xml. For details, see crawlerglobaldefaults.xml reference.
    • During crawling by the FAST Search Web Crawler when JavaScript support is enabled, the Browser Engine suddenly stops processing certain documents. Therefore, these documents are not crawled and indexed.
    • Adding items using the FAST database connector and multi-row normalizer functionality may cause the string "NULL" to be indexed instead of the value NULL.
    • When indexing documents using the FAST specific JDBC connector, and warning or errors occurs, indexing speed drops significantly.
    • When the same managed property is added and deleted many times (more than 20), memory consumption for several processes continues to increase, even if the actual schema remains the same.
Leave a Comment
  • Please add 3 and 4 and type the answer here:
  • Post
Wiki - Revision Comment List(Revision Comment)
Sort by: Published Date | Most Recent | Most Useful
Comments
  • Fernando Lugão Veltem edited Revision 5. Comment: remove en-us from title

  • Craig Lussier edited Revision 4. Comment: added en-US to tags and title

Page 1 of 1 (2 items)
Wikis - Comment List
Sort by: Published Date | Most Recent | Most Useful
Posting comments is temporarily disabled until 10:00am PST on Saturday, December 14th. Thank you for your patience.
Comments
  • Regarding: "Multiple fixes related to feeding performance. This is particularly important for large deployments with many index columns. Prior to SP1, the maximum query throughput (QPS) is roughly 50 divided by the number of columns (for example, a 10 column system would have a throughput of 5)"

    The first sentence is towards feeding performance, but the rest of the paragraph talks about QPS and search. Could you clarify this a bit? Is it related to query performance while feeding?

  • Regarding: "Multiple fixes related to feeding performance" - see updated article. This was actually related to feeding and query performance for large deployments. The other issue listed there effectively prevented you from being able to feed an XL-style deployment.

  • Can FAST SP1 by applied by itself or must we apply Sharepoint SP1 at the same time???

  • Can FAST SP1 by applied by itself or must we apply Sharepoint SP1 at the same time???

  • Craig Lussier edited Revision 4. Comment: added en-US to tags and title

  • It is noted in this post that  

    some fixes (such as the scope fix introduced in the August 2011 CU) are introduced in the SharePoint Server SP1 rather than FAST Search Server 2010 for SharePoint SP1. You should always install both service packs.

    Hope that answers your question.

    Srinivas Dutta | Sr.EE | CTS

  • Please note that as noted in this article  that some fixes (such as the scope fix introduced in the August 2011 CU) are introduced in the SharePoint Server SP1 rather than FAST Search Server 2010 for SharePoint SP1. You should always install both service packs.

    Srinivas Dutta | Sr.EE | CTS

  • Fernando Lugão Veltem edited Revision 5. Comment: remove en-us from title

  • Carsten Siemens edited Revision 6. Comment: Added tag: has comment, has TOC

Page 1 of 1 (9 items)