SharePoint Strategery

Best used when *strategy* breaks down... (blog by Brian Pendergrass, Microsoft SharePoint - Premier Field Engineer)

Troubleshooting the SSP Search Indexer for MOSS 2007

Troubleshooting the SSP Search Indexer for MOSS 2007

Rate This
  • Comments 0

It's time to set aside my SharePoint 2013 farm and get into the way-back machine to troubleshoot the Indexer for MOSS SharePoint 2007 Shared Services Provider (SSP)...

(As I mentioned in a previous post on Search Terminology, I try avoiding the term "Indexer" because it has a muddied connotation in SharePoint 2010 & 2013. However, given this relates to SP2007, consider this post one of those exceptions)

Isolate failures that occur during a crawl (e.g. problems with the content being crawled)

First, there are several generic scenarios that vary greatly in potential root causes, but are worth noting some basic strategies:

  • If items do not appear in search results, double check the Crawl Log to verify that the item(s) actually got crawled
    • If not, address these failures individually
    • If so, often this relates to an issue with contextual scopes such as This Site and This List
      • For an easy way to verify, try searching from an Enterprise Search Center site without any scopes. If the item is found here, then the problem relates to contextual scopes
      • Explaining the cause and ways to overcome of this particular issue is big enough to warrant it's own post, but the simplistic answer: Ensure the [Search] Content Source specifies the URL for the Web Application's Default zone   
  • If you encounter errors crawling individual items (e.g. authentication failures, timeouts, and so on), chances are high that the problem relates to the source being crawled - not search per se 
    • For crawling SharePoint content, the crawler requires access via Windows Authentication (typically NTLM, but Kerberos is possible)
    • The start address in the [Search] Content Source ideally specifies the URL for the Web Application's Default zone (crawling the non-Default zone requires additional configuration)
    • Determine if you can remote into the Indexer as the Default Content Access Account and then attempt to browse to any item(s) that failed (If you can't browse to an item, then the Crawler can't either)
    • The tool Fiddler is you friend:
      • Start Fiddler on the Indexer
      • In the Search Administration page, set the Proxy to 127:0.0.1 with port 8888
      • Start the crawl and watch the traffic from the crawler run through Fiddler

 

For the cases when the Indexer is thoroughly misbehaving (e.g. hung crawls, Services Instances not provisioning properly, unable to load the Search Settings and/or Content Sources, etc), the following troubleshooting steps may help...

Check the Windows SharePoint Services Administration service

On the Indexer server, make sure the Windows SharePoint Services Administration service is running with an account that has local Admin permissions (by default, this service runs as the Local System account). That service is what the SharePoint Timer service invokes when provisioning services, sites, etc. Thus, if the Administration service isn’t running, then the search service instance (e.g. "the bits") and/or the search related web services may not get properly provisioned. Once running, manually run any pending administrative tasks using: stsadm -o execadmsvcjobs

Verify connectivity to the Search Admin Web Services from the SSP

Various errors will occur if the SSP cannot access all of the Search Admin web services. You may also see errors with queries if the Web Front Ends cannot access these web services. To determine if there is a problem accessing the Search Admin web service (running on the Indexer server) from the server hosting the SSP Admin site, try the following:

 

Recycle Services

The crawl portion of search is a state machine and components sometimes get out of sync. Because the state of the crawl is stored in the Search databases, the search processes can be recycled. When restarted, these will simply regain state from the DB. Also, as noted above, there are cases where provisioning components may get hung up, so recycling the Timer and Administration service could also help:

  • From services.msc on each of the Query Servers, recycle the following services:
    • Office Server Search
    • Windows SharePoint Timer
  • From services.msc on the Indexer, then recycle:

    • Office Server Search
    • Windows SharePoint Services Administration
    • Windows SharePoint Services Timer

 

Check for Hung Crawls and Propagation Issues

If a crawl appears hung, check the CrawlID, Status, and SubStatus of the crawl using the MSSCrawlHistory table in the Search database using a query such as the following (replace the date/time to specify the time before the problematic crawl started). This query effectively gets all crawls that started after this time that have not completed successfully:

SELECT * FROM MSSCrawlHistory WITH (nolock) WHERE Status <> 11 and StartTime > '2012-08-22 00:00:00'

From this, use the CrawlID to check the item counts with the two queries such as the following (replace the example CrawlID 12345 below). If you run these over several intervals a few minutes apart each run, the crawl is likely hung:

SELECT count(*) FROM MSSCrawlQueue WITH (nolock) WHERE CrawlId = 12345

SELECT count(*) FROM MSSCrawlUrl WITH (nolock) WHERE CrawlId = 12345

If neither count changes over a period of time, verify the problem does not relate to propagation.

  • As the Indexer gathers and processes content into the full text index, the full text index changes get propagated from the Indexer to any server with a Query role (unless the Indexer server has both the Indexing and the Query role). If anything interferes with this propagation to one or more Query servers, the crawl can appear hung - in reality, the Indexer is simply waiting for the Query server to respond with "yeah, I got that last bit you sent me"
  • Check the Search Admin page for any sort of propagation error
    • Try removing the Query role from the server reporting the propagation error
    • This Query role can be reapplied at a later time
  • Another troubleshooting tactic is configuring the Indexer server to have both the Indexing and Query roles to make "Propagation not required". In this scenario - even if there are multiple query roles - the index does not need to propagate to any server and it remains entirely on the Indexer (and queries will still be returned for the end users)

 

Check Other Common Show Stoppers

 

Determine if the Search service appears correctly provisioned on the Indexer server

  • In Services on Server from Central Admin, what is the state of the Office Server Search service for the Indexer server?
    • Note: If you stop the Search Service here, it will attempt to un-provision search on that server and remove the index
  • From the Indexer server in services.msc, what is the state of the Search service (e.g. Office Server Search)?
    • The expect state is Started with a Manual start type
  • From the Indexer server in the path for the Index, check for a GUID-named folder (e.g. something that looks like baaf81ca-30bf-4657-adc5-97576c1d6d9d ...although your GUID will differ from this)
    • From the Central Admin -> SSP Administration page, hover over the SSP and select Edit Properties. Here, you will see the Index location
    • The default Index location is C:\Program Files\Microsoft Office Servers\1.0\Data\Office Server\Applications\  
  • From the Indexer server in regedit, do you see a path with the same GUID value as above. For example:
    • HKEY_LOCAL_MACHINE\Software\Microsoft\Office Server\12.0\Search\Applications\baaf81ca-30bf-4657-adc5-97576c1d6d9d\
    • If this branch present, verify you also see the following sub-branches
      • .\CatalogNames
      • .\Gather
      • .\Gatherer Matrix Plugin
      • .\Gathering Manager
      • .\ResourceManager

If you find any items that seem off, try recycling the services (as above) and then review the Index/registry paths again to verify if they are present.

 

Invasive Option 1: Build/migrate to a new SSP

At this point, it may make sense to create a new SSP and begin crawling content while continuing to work on the problematic SSP. With this, you could be that much further ahead if the efforts troubleshooting garner no result and you find yourself needing to create a new SSP.

  • While the new SSP crawls, the broken SSP *may still be used to service queries (albeit state results) in the meantime
    • This of course assumes that the query component(s) are still responding to the WFE(s) regarding queries  ...your mileage here may vary
  • Once the crawl completes, you can then move all the Web Applications to this new SSP

Invasive Option 2: Reset the Index

Invasive Option 3: Re-provision Search and/or migrate the Indexer role to another server

If creating a new SSP is not a good option (e.g. you have too many managed properties, scopes, etc), then you can attempt to re-provision search services using the following:

  • From Central Admin -> Operations -> Services on Server:
    • For the servers with only a Query role (e.g. WFE1 and WFE2):
      • Select the WFE1 server
      • Stop and then start the Office Server Search service
      • Repeat the steps on server WFE2
  • For the Indexer server, you can attempt the same (e.g. stop/start the Search service), but it may not allow you to given that this server is the Indexer for the SSP
    • Note: Stopping the Search Service on the Indexer will remove the index
    • If the search service cannot be stopped on the Indexer, another tactic is to [temporarily] start the Indexer role on another server and then point the SSP to this new Indexer server
    • Once the SSP is using the newIndexer server, attempt a more thorough re-provisioning of search on the original Indexer server with the following:
      • Remote into the original Indexer server
      • Stop the Search Service using the command: stsadm -o osearch -action stop
      • Make sure the Office SharePoint Server Search service is stopped in both the Service Console (services.msc) on the Indxer and in the Central Admin GUI
      • Un-provision the Search Service using the command:
      • iisreset
      • Re-provision the Search Service using the command:
      • Start the search Service using the command:
      • If the Search service can be successfully started, you can then update the SSP to use the Indexer server once again

Check the RegistryBlob for Corruption

If the problem persists or things seem to be incompletely provisioned, then the SSP may be corrupted. In several occasions, I found this caused by the RegistryBlob stored in the SSP database getting corrupted. The RegistryBlob is propagated locally to the Indexer - look for a file name RegistryBlob.reg in the path for the Index (check the SSP configuration settings to confirm this path for the Indexer). If this file doesn't exist or contains corrupted characters, then chances are good that the value in the database is also corrupted. If you find a corrupted RegistryBlob, it's worth opening a support case, but - given the nature of corruption - there is no guarantee that this can be resolved.

If the problem still persists after all of this, unfortunately, creating a new SSP is then going to be the only real option…

 

Blog - Comment List MSDN TechNet
  • Loading...
Leave a Comment
  • Please add 4 and 6 and type the answer here:
  • Post