Feeds Protocol Developer's Guide

Google Search Appliance software version 5.0
Posted October 2006
Revised August 2008

This document is for developers who use the Google Search Appliance Feeds Protocol to develop custom feed clients that push content and metadata to the search appliance for processing, indexing, and serving as search results.

Feeds cannot be used on the Google Mini.

To push content to the search appliance, you require a feed and a feed client:

The feed is an XML document that tells the search appliance about the contents that you want to push.
The feed client is the application or web page that pushes the feed to a feeder process on the search appliance.

This document explains how feeds work and shows you how to write a basic feed client.

Overview
1. Why Use Feeds?
2. Writing Your Own Feed Client
Quickstart
Designing an XML Feed
Pushing a Feed to the Google Search Appliance
Turning Feed Contents into Search Results
Troubleshooting
Example Feeds
Google Search Appliance Feed DTD

Overview

You can use feeds to push data into the index on the search appliance. There are two types of feeds:

A web feed provides the search appliance with a list of URLs. A web feed:
- Must be named "web", or have its feed type set to "metadata-and-url".
- May include metadata, if the feed type is set to "metadata-and-url".
- Does not provide content. Instead, the crawler queues the URLs and fetches the contents from each document listed in the feed.
- Is incremental.
- Is recrawled periodically, based on the crawl settings for your search appliance.
A content feed provides the search appliance with both URLs and their content. A content feed:
- Can have any name except "web".
- Provides content for each URL.
- May include metadata.
- Can be either full or incremental.
- Is only crawled when it is pushed.

Web feeds and content feeds also behave differently when deleting content. See Removing Feed Content from the Index for a description of how content is deleted from each type of feed.

To see an example of a feed, follow the steps in the Quickstart section of this document.

Why Use Feeds?

You should design a feed to ensure that your search appliance crawls any documents that require special handling. Consider whether your site includes content that cannot be found through links on crawled web pages, or content that is most useful when it is crawled at a specific time.

Examples of documents that are best pushed using feeds include:

Documents that cannot be fetched using the crawler. For example, records in a database or files on a system that is not web-enabled.
Documents that can be crawled but are best recrawled at different times than those set by the automatic crawl scheduler that runs on the search appliance.
Documents that can be crawled but there are no links on your web site that allow the crawler to discover them during a new crawl.
Documents that can be crawled but are much more quickly uploaded using feeds, due to web server or network problems.

Writing Your Own Feed Client

You push the XML to the search appliance using a feed client. You can use one of the feed clients described in this document or write your own. To write your own feed client, you should be familiar with these technologies:

HTTP - Hypertext Transfer Protocol (specification)
XML - Extensible Markup Language (specification)
a scripting language, such as Python

Quickstart

Here are steps for pushing a content feed to the search appliance.

Download sample_feed.xml to your local computer. This is a content feed for a document titled "Fed Document".
In the Admin Console, go to Crawl and Index > Crawl URLs and add this pattern to "Follow and Crawl Only URLs with the Following Patterns":
```
http://www.localhost.test.com
```
This is the URL for the document defined in sample_feed.xml.
Download pushfeed_client.py to your local computer. This is a feed client script. You must install Python 2.2 or later to run this script.
Configure the search appliance to accept feeds from your computer. In the Admin Console, go to Crawl and Index > Feeds, and scroll down to List of Trusted IP Addresses. Verify that the IP address of your local computer is trusted.
Run the feed client script with the following arguments (you must change "APPLIANCE-HOSTNAME" to the hostname or IP address of your search appliance):
```
% pushfeed_client.py --datasource="sample" --feedtype="full"
  --url="http://<APPLIANCE-HOSTNAME>:19900/xmlfeed" --xmlfilename="sample_feed.xml"
```
In the Admin Console, go to Crawl and Index > Feeds. A data source named "sample" should appear within 5 minutes.
The URL http://www.localhost.test.com/ should appear under Crawl Diagnostics within about 15 minutes.
Enter the following as your search query to see the URL in the results:
```
info:http://www.localhost.test.com/
```
If your system is not busy, the URL should appear in your search results within 30 minutes.

Designing an XML Feed

The feed is an XML file that contains the URLs. It may also contain their contents, metadata, and additional information such as the last-modified date. The XML must conform to the schema defined by gsafeed.dtd. This file is also available on your search appliance at http://<APPLIANCE-HOSTNAME>:7800/gsafeed.dtd. Although the Document Type Definition (DTD) defines elements for the data source name and the feed type, these elements are populated when you push the feed to the search appliance. Any datasource or feedtype values that you specify within the XML document are ignored.

The maximum size for a feed XML file is 1 GB. Files larger than 1 GB cannot be pushed to the search appliance.

Choosing a Name for the Feed Data Source

When you push a feed to the search appliance, the system associates the fed URLs with a data source name, specified by the datasource element in the feed DTD.

If the data source name is "web", the system treats the feed as a web feed. A search appliance can only have one data source called "web".
If the data source name is anything else, and the feed type is metadata-and-url, the system treats the feed as a web feed.
If the data source name is anything else, and the feed type is not metadata-and-url, the system treats the feed as a content feed.

To view all of the feeds for your search appliance, log into the Admin Console and choose Crawl and Index > Feeds. The list shows the date of the most recent push for each data source name, along with whether the feed was successful and how many documents were pushed.

Note: Although you can specify the feed type and data source in the XML file, the values specified in the XML file are currently unused. Instead, the search appliance uses the data source and feed type that are specified during the feed upload step. However, we recommend that you include the data source name and feed type in the XML file for compatibility with future versions.

Choosing the Feed Type

The feed type determines how the search appliance handles URLs when a new content feed is pushed with an existing data source name.

Content feeds can be full or incremental; a web feed is always incremental. To support feeds that provide only URLs and metadata, in versions 4.6.2.S.12 and later you can also set the feed type to metadata-and-url. This is a special feed type that is treated as a web feed.

When the feedtype element is set to full for a content feed, the system deletes all the prior URLs that were associated with the data source. The new feed contents completely replace the prior feed contents. If the feed contains metadata, you must also provide content for each record; a full feed cannot push metadata alone. You can delete all documents in a data source by pushing an empty full feed.
When the feedtype element is set to incremental, the system modifies the URLs that exist in the new feed as specified by the action attribute for the record. URLs from previous feeds remain associated with the content data source. If the record contains metadata, you can incrementally update either the content or the metadata.
For example, if your incremental feed specifies metadata for the record without any content information, the metadata will update but the content will remain unchanged; if your incremental feed only specifies content for the record, the content will update but the metadata will remain unchanged. You can only update metadata without content for records that had content in an earlier push.
When the feedtype element is set to metadata-and-url, the system modifies the URLs and metadata that exist in the new feed as specified by the action attribute for the record. URLs and metadata from previous feeds remain associated with the content data source. You can use this feed type even if you do not define any metadata in the feed. The system treats any data source with this feed type as a special kind of web feed and updates the feed incrementally. See the External Metadata Indexing Guide for more information about external metadata.

Full Feeds and Incremental Feeds

Incremental feeds generally require fewer system resources than full feeds. A large feed can often be crawled more efficiently if it is divided into smaller incremental feeds.

Here is an example that illustrates the effect of a full feed:

Day 1 - You create a new data source by pushing a feed that contains documents D0, D1 and D2. The system serves D0, D1, and D2.
Day 2 - Using the same data source name, you push a full feed that contains documents D0, an updated D1, and a new D3. When the feed processing is complete, the system serves D0, the updated D1, and the new D3. Because document D2 was not defined in the full feed, it is removed from the index.

Here is an example that mixes full and incremental feeds:

Day 1 - You create a new data source by pushing a feed that contains documents D0, D1 and D2. The system serves D0, D1 and D2.
Day 2 - You push an incremental feed that defines the following actions: "add" for D3, "add" for an updated D1, and "delete" for D2. The system serves D0, updated D1, and D3. D0 was pushed by the first feed; because it is not referenced in the incremental feed, the D0's contents remain in the search results.
Day 7 - You push a full feed that contains documents D0, D7, and D10. The system serves D0, D7, and D10 when the full feed processing is complete. D1 and D3 are not referenced in the full feed, so the system removes them from the index and does not add them back.

Defining the XML Record for a Document

You include documents in your feed by defining them inside a record element. All records must specify a URL which is used as the unique identifier for the document. If the original document doesn't have a URL, but has some other unique identifier, you must map the document to a unique URL in order to identify it in the feed.

Each record element can specify following attributes:

url (required) - The URL is the unique identifier for the document. This is the URL used by the search appliance when crawling and indexing the document. All URLs must contain a FQDN (fully qualified domain name) in the host part of the URL. Because the URL is provided as part of an XML document, you must escape any special characters that are reserved in XML. For example, the URL http://www.mydomain.com/bar?a=1&b2 contains an ampersand character and should be rewritten to http://www.mydomain.com/bar?a=1&b2.
displayurl - The URL that should be provided in search results for a document. The search appliance enables use of the display URL only for content feeds and not for web feeds. This attribute is useful for web-enabled content systems where a user expects to obtain a URL with full navigation context and other application-specific data, but where a page does not give the search appliance easy access to the indexable content.
action - Set action to "add" when you want the feed to overwrite and update the contents of a URL. If you don't specify an action, the system performs an add. For content feeds only, set action to "delete" to remove a URL from the index. For more information on how to delete content from a web feed, see Removing Feed Content From the Index.
lock - The lock attribute can be set to true or false (the software default is true). When the search appliance reaches its license limit, unlocked documents are deleted to make room for more documents. Locked documents are deleted last. For more information, see License Limits.
mimetype (required) - This attribute tells the system what kind of content to expect from the content element. All MIME types that can be indexed by the search appliance are supported.
last-modified - Populate this attribute with the date time format specified in RFC822 (Mon, 15 Nov 2004 04:58:08 GMT). If you do not specify a last-modified date, then the implied value is blank. The system uses the rules specified in the Admin Console under Crawl and Index > Document Dates to choose which date from a document to use in the search results. The document date extraction process runs periodically so there may be a delay between the time a document appears in the results and the time that its date appears.
authmethod - This attribute tells the system how to crawl URLs that are protected by NTLM, HTTP Basic, or Single Sign-on. The authmethod attribute can be set to none, httpbasic, ntlm, or httpsso. By default, it is set to none. If you want to enable crawling for protected documents, see Including Protected Documents in Search Results.
pagerank - This attribute is not yet supported.

Grouping Records Together

The optional group element allows you to apply an action to many records at once. For example, this:

<group action="delete">
  <record url="http://www.corp.enterprise.com/hello01"/>
  <record url="http://www.corp.enterprise.com/hello02"/>
  <record url="http://www.corp.enterprise.com/hello03"/>
</group>

Is equivalent to this:

  <record url="http://www.corp.enterprise.com/hello01" action="delete"/>
  <record url="http://www.corp.enterprise.com/hello02" action="delete"/>
  <record url="http://www.corp.enterprise.com/hello03" action="delete"/>

However, if you define any actions for records as a group, the record's definition always overrides the group's definition. For example:

<group action="delete">
  <record url="http://www.corp.enterprise.com/hello01"/>
  <record url="http://www.corp.enterprise.com/hello02" action="add"/>
  <record url="http://www.corp.enterprise.com/hello03"/>  
</group>

In this example, hello01 and hello03 would be deleted, and hello02 would be updated.

Providing Content in the Feed

You add document content by placing it inside the record definition for your content feed. For example, using text content:

<record url="..." mimetype="text/plain">
  <content>Hello world. Here is some page content. </content>
</record>

You can also define content as HTML:

<record url="..." mimetype="text/html">
  <content><![CDATA[<html> <title>hello world</title>
    <body><p> Here is some page content. </p>
    </body> </html>]]></content>
</record>

To include non-text documents such as .pdf or .doc files, you must encode the content using base64 encoding. Using base64 encoding ensures that the feed can be parsed as valid XML.

Here is a record definition that includes base64 encoded content:

<record url="..." mimetype="text/plain">
  <content encoding="base64binary">Zm9vIGJhcgo</content>
</record>

Because base64 encoding increases the document size by one third, it is often more efficient to include non-text documents as URLs in a web feed. Only contents that are embedded in the XML feed must be encoded; this restriction does not apply to contents that are crawled.

Adding Metadata Information to a Record

Metadata can be included in record definitions for different types of feeds.

The following table provides information about incremental web feeds and metadata-and-URL feeds.

Data Source Name	Feed Type	Push Behavior	Allows Metadata?	Allows Content?
`web`	`incremental`	incremental	no	no
any	`metadata-and-url`	incremental	yes	no

The following table provides information about incremental and full content feeds.

Data Source Name	Feed Type	Push Behavior	Allows Metadata?	Allows Content?
any	`incremental`	incremental	yes	yes
any	`full`	full	yes	yes

If the metadata is part of a feed, it must have the following format:

<record url="..." ...>
  <metadata>
    <meta name="..." content="..." />
    <meta name="..." content="..." />
  </metadata>
  ...
</record>

See the External Metadata Indexing Guide for more information about indexing external metadata and examples of metadata feeds.

Including Protected Documents in Search Results

Feeds can push protected contents to the search appliance. If your feed contains URLs that are protected by NTLM, Basic Authentication, or Forms Authentication (Single Sign-on), the URL record in the feed must specify the correct type of authentication. You must also configure settings in the Admin Console to allow the search appliance to crawl the secured pages.

The authmethod attribute for the record defines the type of authentication. By default, authmethod is set to "none". To enable secure search from a feed, set the authentication attribute for the record to ntlm, httpbasic, or httpsso. For example, to enable authentication for protected files on localhost.test.com via Forms Authentication, you would define the record as:

<record url="http://www.localhost.test.com/" authmethod="httpsso">

To grant the search appliance access to the protected pages in your feed, log into the Admin Console.

For URLs that are protected by NTLM and Basic Authentication, follow these steps:

Open Crawl and Index > Crawler Access.
Define a pattern that matches the protected URLs in the feed.
Enter a username and password that will allow the crawler access to the protected contents. For contents on a Microsoft IIS server, you may also need to specify a domain.
The Make Public check box controls whether the search appliance checks for valid authentication credentials before including protected contents in the search results. If you select the Make Public check box, the record is displayed in search results. Otherwise, the record is shown when the user has valid authentication credentials; users who do not have access to the protected content will not see it in their search results. By default, search results are protected.

For URLs that are protected by Single Sign-on, follow these steps:

Open Crawl and Index > Forms Authentication.
Under Sample Forms Authentication protected URL, enter the URL of a page in the protected site that will redirect the user to a login form. The login form must not contain Javascript or frames. If you have more than one login page, create a Forms Authentication rule for each login.
Under URL pattern for this rule, enter a pattern that matches the protected URLs in the feed.
Click Create a New Forms Authentication Rule. In the browser page that opens, use the login form to enter a valid username and password. These credentials allow the crawler access to the protected contents. If the login information is accepted, you should see the protected page that you specified. If you can see the protected URL contents, click the Save Forms Authentication Rule and Close Window button. The Forms Authentication page now displays your rule.
Make any changes to the rule. For example, the Make Public check box controls whether the search appliance checks for valid authentication credentials before including protected contents in the search results. If you select the Make Public check box, the record is displayed in search results. Otherwise, the record is shown when the user has valid authentication credentials; users who do not have access to the protected content will not see it in their search results. By default, search results are protected.
When you have finished making changes to the rule, click the Save Forms Authentication Rule Configuration button.

This is one way of providing access to protected documents. For more information on authentication, refer to the online help that is available in the search appliance's Admin Console, and Managing Search for Controlled-Access Content.

Feeding Content from a Database

To push records from a database into the search appliance's index, you use a special content feed that is generated by the search appliance based on parameters that you set in the Admin Console. To set up a feed for database content, log into the Admin Console and choose Crawl and Index > Database. You can find more information on how to define a database-driven data source in the online help that is available in the Admin Console, and in the document Database Crawling and Serving.

Records from a database cannot be served as secure content.

Saving your XML Feed

You should save a backup copy of your XML Feed in case you need to push it again. For example, if you perform a version update that requires you to rebuild the index, you must push all your feeds again to restore them to the search appliance. The search appliance does not archive copies of your feeds.

Pushing a Feed to the Google Search Appliance

This section describes how to design a feed client. If you don't want to design your own feed client script, you can use one of the following methods to push your feed:

Google provides an example of a feed client script, pushfeed_client.py, that you can use to push an XML feed. You can also use this script in a cron job to automate feeds.
Using a Web Form Feed Client explains how to write a simple HTML form that allows a user to push an XML feed from a web page. Adapt the HTML for your use and add this page to any web server that has HTTP access to the search appliance.

Important: The IP address of the computer that hosts the feed client must be in the List of Trusted IP Addresses. In the Admin Console, go to Crawl and Index > Feeds, and scroll down to List of Trusted IP Addresses. Verify that the IP address for your feed client appears in this list.

Designing a Feed Client

You upload an XML feed using an HTTP POST to the feedergate server located on port 19900 of your search appliance. Feeds cannot be uploaded using HTTPS. The XML feed must be < 1 GB in size. If your feed is larger than 1 GB, consider breaking the feed into smaller feeds that can be pushed more efficiently.

The feedergate server requires three input parameters from the POST operation:

datasource specifies the name of the data source. Your choice of data source name also implies the type of feed: for a web feed, the datasource name must be "web".
feedtype specifies how the system pushes the feed. The feedtype value must be "full", "incremental", or "metadata-and-url".
data specifies the XML feed to push with this data source and feed type. Note that although the data parameter may contain a data source and a feed type definition as part of the XML, these will be ignored by the search appliance. Only the data source and feed type provided as POST input parameters are used.

The URL that you should use is:

http://<APPLIANCE-HOSTNAME>:19900/xmlfeed

You should post the feed using enctype="multipart/form-data". Although the search appliance supports uploads using enctype="application/x-www-form-urlencoded", this encoding type is not recommended for large amounts of data.

Using a Web Form Feed Client

Here is an example of a simple HTML form for pushing a feed to the search appliance. Because the web form requires user input, this method cannot be automated.

To adapt this form for your search appliance, replace APPLIANCE-HOSTNAME with the fully qualified domain name of your search appliance.

<html>
  <head>
    <title>Simple form for pushing a feed</title>
  </head>
  <body>
   <h1>Simple form for pushing a feed</h1>
    <form enctype="multipart/form-data" method=POST
         action="http://<APPLIANCE-HOSTNAME>:19900/xmlfeed">
      <p>Name of datasource:
        <input type="text" name="datasource">
        <br>
        (No spaces or non-alphanumeric characters)
      </p>
      <p>Type of feed:
        <input type="radio" name="feedtype" value="full" checked>
        Full
        <input type="radio" name="feedtype" value="incremental">
        Incremental
        <input type="radio" name="feedtype" value="metadata-and-url">
        Metadata and URL
      </p>
      <p>
        XML file to push:
        <input type="file" name="data">
      </p>
      <p>
        <input type="submit" value=">Submit<">
      </p>
    </form>
  </body>
</html>

How a Feed Client Pushes a Feed

When pushing a feed, the feed client sends the POST data to a search appliance. The following example shows a typical POST from a scripted feed client (the feedtype lines are broken for readability):

POST /xmlfeed HTTP/1.0
Host: myserver.domain.com:19900
Content-Length: 221
Content-Type: application/x-www-form-urlencoded

feedtype=full&datasource=sample&data=%3C%3Fxml%20
version%3D%221.0%22%20encoding%3D%22UTF8%22%3F%3E%0A%3C%21
DOCTYPE%20gsafeed%20PUBLIC%20%22-//Google//DTD%20GSA%20
Feeds//EN%22%20%22%22%3E%0A%3Cgsafeed%3E%0A%3C/gsafeed%3E

The following example shows the response from the search appliance:

HTTP/1.1 200 OK
Content-Type: text/plain
Date: Thu, 8 Nov 2007 23:16:10 GMT
Server: feedergate_1.0
Connection: Close
Content-Length: 7

Success

The success message indicates that the feedergate process has received the XML file successfully. It does not mean that the content will be added to the index, as this is handled asynchronously by a separate process known as the "feeder". The data source will appear in the Feeds page in the Admin Console once the feeder process has run.

The feeder does not provide automatic notification of a feed error. To check for errors, you must log into the Admin Console and check the status on the Crawl and Index > Feeds page. This page shows the last five feeds that have been uploaded for each data source. The timestamp shown is the time that the XML file has been successfully uploaded by the feedergate server.

You can automate the process of uploading a feed by running your feed client script with a cron job.

Turning Feed Contents Into Search Results

URL Patterns and Trusted IP lists defined in the Admin Console ensure that your index only lists content from desirable sources. When pushing URLs with a feed, you must verify that the Admin Console will accept the feed and allow your content through to the index. For a feed to succeed, it must be fed from a trusted IP address and at least one URL in the feed must pass the rules defined on the Admin Console.

URL Patterns

URLs specified in the feed will only be crawled if they pass through the patterns specified on the Crawl and Index > Crawl URLs page in the Admin Console.

Patterns affect URLs in your feed as follows:

Do Not Crawl URLs with the Following Patterns - If a URL in the feed matches a pattern specified under Do Not Crawl URLs with the Following Patterns, the URL is removed from the index.
Follow and Crawl Only URLs with the Following Patterns - When this pattern is used, all URLs in the feed must match a pattern in this list. Any other URLs are removed from the index.

Entries in duplicate hosts also affect your URL patterns. For example, suppose you have a canonical host of foo.mycompany.com with a duplicate host of bar.mycompany.com. If you exclude bar.mycompany.com from your crawl using patterns, then URLs on both foo.mycompany.com and bar.mycompany.com are removed from the index.

Trusted IP Lists

To prevent unauthorized additions to your index, feeds are only accepted from machines that are included in the List of Trusted IP Addresses. To view the list of trusted IP addresses, log into the Admin Console and open the Crawl and Index > Feeds page.

If your search appliance is on a trusted network, you can disable IP address verification by selecting Trust all IP addresses.

Adding Feed Content

For web feeds, the feeder passes the URLs to the crawl manager. The crawl manager adds the URLs to the crawl schedule. URLs are crawled on the schedule specified by the documentation on the continuous crawler.

For content feeds, the content is provided as part of the XML and does not need to be fetched by the crawler. URLs are passed to the server that maintains Crawl Diagnostics in the Admin Console. This will happen within 15 minutes if your system is not busy. The feeder also passes the URLs and their contents to the indexing process. The URLs will appear in your search results within 30 minutes if your system is not busy

Removing Feed Content From the Index

There are several ways of removing content from your index using a feed. The method used to delete content depends on the kind of feed that has ownership.

For content feeds, remove content by performing one of these actions:

Push the URL as part of an incremental feed, using the "delete" action to remove the content. This is the fastest way to remove content. URLs will be deleted within about 30 minutes since they do not need to wait for the feeder delete process to run.
Remove the URL from the feed and perform a full feed. Because a full feed overwrites the earlier feed contents, any URLs that are omitted from the new full feed will be removed from the index. The content will be deleted within 6 hours, when the feeder delete process next runs.
Remove the data source and all of its contents. To remove a data source, log into the Admin Console and open the Crawl and Index > Feeds page. Choose the data source that you want to remove and click Delete. The contents will be deleted within 6 hours, when the feeder delete process next runs.

For web feeds, remove content by performing one of these actions:

Remove the URL from the web server. The next time that the URL is crawled, the system will encounter a 404 status code and remove the content from the index.
Specify a pattern that removes the URL from the index. For example, add the URL to the Do Not Crawl URLs with the Following Patterns list. The URL is removed the next time that the feeder delete process runs.

Note: If a URL is referenced by more than one feed, you will have to remove it from the feed that owns it. See the Troubleshooting entry Fed Documents Aren't Updated or Removed as Specified in the Feed XML for more information.

Time Required to Process a Feed

The following factors can cause the feeder to be slow to add URLs to the index:

The feed is large.
The search appliance is currently using a lot of resources to crawl other documents and serve results.
Other feeds are pending.

In general, the search appliance can process documents that are pushed as content feeds more quickly than it can crawl and index the same set of documents as a web feed.

License Limits

If your index already contains the maximum number of URLs, or your license limit has been exceeded, then the index is full.

When the index is full, the system reduces the number of indexed documents as follows:

Documents are removed to bring the total number of documents to the license limit.
Documents with the lock attribute set to true are deleted last.

Increasing the Maximum Number of URLs to Crawl

To increase the maximum number of URLs in your index, log into the Admin Console and choose Crawl and Index > Host Loads > Maximum Number of URLs to Crawl. This number must be smaller than the license limit for your search appliance. To increase the license limit, contact Sales.

Troubleshooting

Here are some things to check if a URL from your feed does not appear in the index. To see a list of known and fixed issues, see the latest release notes for each version on the support site.

Error Messages on the Feeds Status Page

If the Feeds status page shows "Failed in error" you can click the link to view the log file.

ProcessFeed: parsing error

This message means that your XML file could not be understood. Here are some possible causes of this error:

There is an error in the DOCTYPE line in your XML file. This line should be: <!DOCTYPE gsafeed PUBLIC "-//Google//DTD GSA Feeds//EN""">
You have not escaped the ampersand or other reserved characters in a URL in your XML file.
You have included content that requires base64 encoding. See Providing Content in the Feed.

If none of the above are the cause of the error, run xmllint against your XML file to check for errors in the XML. The xmllint program is included in Linux distributions as part of the libxml2 package.

Here is an example that shows how you would use xmllint to test a feed named full-feed.xml.

$ xmllint -noout -valid full-feed.xml; echo $?

  0

The return code of zero indicates that the document is both valid and well-formed.

If the xmllint command fails and displays the parsing error message, ensure that you have the correct DTD file, or you can remove the -valid flag from the xmllint command line so that the xmllint command doesn't try to validate the XML file's elements. For more information on the DTD, see Google Search Appliance Feed DTD.

Feed Push is Not Successful

Before a search appliance can start processing a feed, you need to successfully push a feed to port 19900.

If the feed push is not successful, check the following:

List the IP address of the computer that hosts the feed client in the List of Trusted IP Addresses. For more information, see Pushing a Feed to the Google Search Appliance.
Verify that the feed client connects to port 19900 on the correct IP address.
Verify that port 19900 is reachable on the search appliance by running tracepath applianceIP/19900 from the Linux command line. From a Windows command prompt, you can use the Telnet command to connect to port 19900. For example to verify port access on the mysa search appliance, use this command:
```
telnet mysa 19900
```
If you use Windows Telnet to connect to a search appliance, either a blank page appears, which indicates that the port 19900 is open, or Telnet displays the message, Connecting To mysa...Could not open connection to the host, on port 19900: Connect failed to indicate that port 19900 is not enabled. If the blank page appears, enter quit and press Enter to return to the command prompt. An HTML listing appears with the last line stating Connection to host lost.
Check with your network team for firewall or connectivity issues to ensure access to port 19900.

Fed Documents Aren't Appearing in Search Results

Some common reasons why the URLs in your feed might not be found in your search results include:

The crawler is still running. Wait a few hours and search again. For large document feeds containing multiple non-text documents, the search appliance can take several minutes to process all of the documents. You can check the status of a document feed by going to the Crawl and Index > Feeds page. You can also verify that the documents have been indexed by going to Status and Reports > Crawl Diagnostics and browsing to the URL, or entering the URL in "URLs starting with." Documents that are fed into the search appliance can show up in Crawl Diagnostics up to 15 minutes before they are searchable in the index.
The URLs were removed by an exclusion pattern specified under Crawl and Index > Crawl URLs. See URL Patterns.
The URLs don't match the pattern for the collection that you were searching. Check the patterns for your collection under Crawl and Index > Collections. Make sure that the collection specified in the upper right hand corner of the Crawl Diagnostics page contains the URL that you are looking for.
The URLs are listed in multiple feeds. Another feed that contains this URL requested a "delete" action.
The documents were removed because your index is full. See License Limits.
The feed that you pushed was not pointing to a valid host. Verify that the feed has an FQDN (fully qualified domain name) in the host part of the URL.
More relevant documents are pushing the fed URL down in the list. You can search for a specific URL with the query info:[url] where [url] is the full URL to a document fed into the search appliance. Or use url:[path] where [path] is part of the URL to documents fed into the search appliance.
The URLs are on a protected server and cannot be indexed. See Including Protected Documents in Search Results.
The URLs are on a protected server and have been indexed, but you do not have the authorization to view them. Make sure that "&access=a" is somewhere in the query URL that you are sending to the search appliance. See Including Protected Documents in Search Results.
You did not complete the upgrade from a previous version and are still running in "Test mode" with the old Index. Review the Update Instructions for the current version of the software, and make sure that you have accepted the upgrade and completed the update process.

Fed Documents Aren't Updated or Removed as Specified in the Feed XML

All feeds, including database feeds, share the same name space and assume that URLs are unique. If a fed document doesn't seem to behave as directed in your feed XML, check to make sure that the URL isn't duplicated in your other feeds.

When the same URL is fed into the system by more than one data source, the system uses the following rules to determine how that content should be handled:

If the URL is referenced by a web feed and a content feed, the URL's content is associated with the data source that crawled the URL last.
If the URL is referenced by more than one content feed, the URL's content is associated with the data source that was responsible for the URL's last update.
If the URL is referenced in the Admin Console's list of Crawl URLs and a content feed, the URL's content is associated with the content feed. The search appliance will not recrawl the URL until the content feed requests a change. To return the URL to its original status, delete the URL from the feed that originally pushed the document to the index.
If the URL has already been crawled by the search appliance, and is then referenced in a web feed, the URL will be immediately injected into the queue to be recrawled as if it were a new, uncrawled URL. The URL's Enterprise PageRank will not be affected. However, the change interval will be reset to the default until the crawl scheduler process next runs.

Document Status is Stuck "In Progress"

If a document feed gives a status of "In Progress" for more than one hour, this could mean that an internal error has occurred. Please contact Google to resolve this problem, or you can reset your index by going to Administration > Reset Index.

Feed Client TCP Error

If you are using Java to develop your feed client, you may encounter the following exception when pushing a feed:

Java.net.ConnectException: Connection refused: connect

Although it looks like a TCP error, this error may reveal a problem in parsing the MIME boundary parameter syntax, for example, missing a '--' before the argument. MIME syntax discussed in more detail here: http://www.w3.org/Protocols/rfc1341/7_2_Multipart.html

Example Feeds

Here are some examples that demonstrate how feeds are structured:

Web Feed
Web Feed with Metadata
Full Content Feed, showing an example of base64binary encoded data
Incremental Content Feed

Web Feed

<?xml version="1.0" encoding="iso-8859-1"?>
<!DOCTYPE gsafeed PUBLIC "-//Google//DTD GSA Feeds//EN" "">
  <gsafeed>
  <header>
    <datasource>web</datasource>
    <feedtype>incremental</feedtype>
  </header>
  <group>
    <record url="http://www.corp.enterprise.com/hello02" mimetype="text/plain"/>
  </group>
</gsafeed>

Web Feed with Metadata

<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE gsafeed PUBLIC "-//Google//DTD GSA Feeds//EN" "">
<gsafeed>
  <header>
    <datasource>example3</datasource>
    <feedtype>metadata-and-url</feedtype>
  </header>
  <group>
   <record url="http://www.corp.enterprise.com/search/employeesearch.php?q=jwong"
      action="add" mimetype="text/html" lock="true">
     <metadata>
       <meta name="Name" content="Jenny Wong"/>
       <meta name="Title" content="Metadata Developer"/>
       <meta name="Phone" content="x12345"/>
       <meta name="Floor" content="3"/>
       <meta name="PhotoURL"
         content="http://www/employeedir/engineering/jwong.jpg"/>
       <meta name="URL"
         content="http://www.corp.enterprise.com/search/employeesearch.php?q=jwong"/>
     </metadata>
   </record>
  </group>
</gsafeed>

Full Content Feed

<?xml version="1.0" encoding="UTF8"?>
<!DOCTYPE gsafeed PUBLIC "-//Google//DTD GSA Feeds//EN" "">
<gsafeed>
  <header>
    <datasource>sample</datasource>
    <feedtype>full</feedtype>
  </header>
  <group>
    <record url="http://www.corp.enterprise.com/hello01" mimetype="text/plain"
      last-modified="Tue, 6 Nov 2007 12:45:26 GMT">
      <content>This is hello01</content>
    </record>
    <record url="http://www.corp.enterprise.com/hello02" mimetype="text/plain"
      lock="true">
      <content>This is hello02</content>
    </record>
    <record url="http://www.corp.enterprise.com/hello03" mimetype="text/html">
      <content><![CDATA[
        <html>
          <title>namaste</title>
          <body>
	  This is hello03
          </body>
        </html>
       ]]></content>
    </record>
    <record url="http://www.corp.enterprise.com/hello04" mimetype="text/html">
      <content encoding="base64binary">Zm9vIGJhcgo</content>
    </record>
  </group>
</gsafeed>

Incremental Content Feed

<?xml version="1.0" encoding="iso-8859-1"?>
<!DOCTYPE gsafeed PUBLIC "-//Google//DTD GSA Feeds//EN" "">
<gsafeed>
  <header>
    <datasource>hello</datasource>
    <feedtype>incremental</feedtype>
  </header>
  <group action="delete">
    <record url="http://www.corp.enterprise.com/hello01"/>
  </group>
  <group>
    <record url="http://www.corp.enterprise.com/hello02" mimetype="text/plain">
      <content>UPDATED - This is hello02</content>
    </record>
    <record url="http://www.corp.enterprise.com/hello03" action="delete"/>
    <record url="http://www.corp.enterprise.com/hello04" mimetype="text/plain">
      <content>UPDATED - This is hello04</content>
    </record>
  </group>
</gsafeed>

Google Search Appliance Feed DTD

The gsafeed.dtd file follows. You can view the DTD on your search appliance by browsing to the http://<APPLIANCE-HOSTNAME>:7800/gsafeed.dtd URL.

<?xml version="1.0" encoding="UTF-8"?>
<!ELEMENT gsafeed (header, group+)>
<!ELEMENT header (datasource, feedtype)>
<!-- datasource name should match the regex [a-zA-Z_][a-zA-Z0-9_-]*,
     the first character must be a letter or underscore,
     the rest of the characters can be alphanumeric, dash, or underscore. -->
<!ELEMENT datasource (#PCDATA)>
<!-- feedtype must be either 'full', 'incremental', or 'metadata-and-url' -->
<!ELEMENT feedtype (#PCDATA)>
<!-- group element lets you group records together and
     specify a common action for them -->
<!ELEMENT group (record*)>
<!-- record element can have attribute that overrides group's element-->
<!ELEMENT record (metadata*,content*)>
<!ELEMENT metadata (meta*)>
<!ELEMENT meta EMPTY>
<!ELEMENT content (#PCDATA)>
<!-- default is 'add' -->
<!-- last-modified date as per RFC822 -->
<!ATTLIST group
   action (add|delete) "add"
   pagerank CDATA #IMPLIED>
<!ATTLIST record
   url CDATA #REQUIRED
   displayurl CDATA #IMPLIED
   action (add|delete) #IMPLIED
   mimetype CDATA #REQUIRED
   last-modified CDATA #IMPLIED
   lock (true|false) "false"
   authmethod (none|httpbasic|ntlm|httpsso) "none"
   pagerank CDATA #IMPLIED>
<!ATTLIST meta
   name CDATA #REQUIRED
   content CDATA #REQUIRED>
<!-- if encoding is specified it must be base64binary as that is the only
     binary encoding that is supported -->
<!ATTLIST content encoding (base64binary) #IMPLIED>

Google Search Appliance

Search Appliance

Feeds Protocol Developer's Guide

Support Resources

Feeds Protocol Developer's Guide

Contents

Overview

Why Use Feeds?

Writing Your Own Feed Client

Quickstart

Designing an XML Feed

Choosing a Name for the Feed Data Source

Choosing the Feed Type

Full Feeds and Incremental Feeds

Defining the XML Record for a Document

Grouping Records Together

Providing Content in the Feed

Adding Metadata Information to a Record

Including Protected Documents in Search Results

Feeding Content from a Database

Saving your XML Feed

Pushing a Feed to the Google Search Appliance

Designing a Feed Client

Using a Web Form Feed Client

How a Feed Client Pushes a Feed

Turning Feed Contents Into Search Results

URL Patterns

Trusted IP Lists

Adding Feed Content

Removing Feed Content From the Index

Time Required to Process a Feed

License Limits

Increasing the Maximum Number of URLs to Crawl

Troubleshooting

Error Messages on the Feeds Status Page

ProcessFeed: parsing error

Feed Push is Not Successful

Fed Documents Aren't Appearing in Search Results

Fed Documents Aren't Updated or Removed as Specified in the Feed XML

Document Status is Stuck "In Progress"

Feed Client TCP Error

Example Feeds

Web Feed

Web Feed with Metadata

Full Content Feed

Incremental Content Feed

Google Search Appliance Feed DTD