1

I have 2 instances of Tridion 2013 on 2 nodes behind an F5 load balancer. However, when I try to search, its very inconsistent/not accurate between both nodes. I never get the same results given the exact same criteria.

I have ran these commands via powershell on both nodes: Sync-TcmSearchIndex and (Get-TcmRepository).ID | Sync-TcmSearchIndex -verbose restarted IIS, COM+ and services and still see no differences as prior.

Also, I have looked into:

  • Tridion\bin\Configure-Search-Host.ps1
  • Tridion\bin\Optimize-Search-Index.ps1

but per the documentation, it seems as though the setup is for an outscaled solution. Can this still be done in my scenario?

Tough part about it is, that its hard to test in a lower or single instance environment, so I am stepping on glass in a higher clustered (load balanced) setup; otherwise, I would be firing pistols at the hip.

I have also looked into http://localhost:8983/#/tridion Solr Admin and I can see how the Statistics are displayed and also how a Master node could replicate to a Slave node in a given scenario, but I cannot connect the dots within my setup.

I have performed db maintenance scripts that did not make any differences to the search:

  1. AnalyzeStatsEx.sql
  2. BuildMaintenanceScripts.sql
  3. RebuildIndexes.sql
  4. ShrinkLobs.sql

I have cross checked the indexes rebuild date and time, looks fine and as same as the script execution time. DB level everything looks fine as well..

Anyone have any quick guides out there on a Tridion search optimization tutorial in a load balanced scenario like mine? Or what about 3 nodes? A Master node that gets replicated out to 2 Slaves?

UPDATE

Configuring a Search Host Master Server within the doc here,

Step #6 - In the security-role element, find the role-name element to identify the user group. All Slave Servers must configure the same Master Server user, who must be a Windows local user and a member of this user group.

<security-role>
<role-name>servermaster\SDLSearchUsers</role-name>
</security-role>

Maybe this is just poorly worded, but I cannot seem to grasp this step. The group SDLSearchUsers exists on both nodes and contains the MTSUSER..

Search is still not accurate on the clustered url when returning results of items that were created on each node intentionally.

2

we have implemented a search master which replicates the index to 2 slave servers. If you have multiple CME machines, you have to do something like this or you will have the inconsistencies for the search index. It's a single search queue. the first CME to pick up the record would index it and keep it in its local index, the others wouldn't have it. so you'd have your indexed content randomly scattered across the multiple servers--not what you want. I suggest you read up on the options for scaling out search and do a followup question once you decide on your approach.

1

As Warner suggested, the inconsistencies that you are seeing seems to be because of Search setup issues.

In case of multiple CM servers, you have setup master / slave search servers. You can refer to following link for that: Outscaled Search Setup

There is an automatic replication which happens from the master servers to the slave servers whenever an item is created/changed/deleted in Tridion. On the slave servers, you need to provide:

indexer engine host url property = master server url:8983

With Tridion 2013 though, there is an issue with indexing (in master slave configuration) which is documented here with the solution as well Tridion KB

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.