search Verity K2 Integration

Hi,

We are looking to implement Verity K2 search on a large scale into our website. Looking into how we integrate Verity to treat each workarea independently has been a nightmare.

Could any one please shed some light on how Verity can run seamlessly within TeamSite workarea's without having 1-1 instances of verity running.

Your assistance would be highly appreciated

Regards
John Cuiuli

Edited by Johnny on 09/15/02 06:07 PM (server time).

Find more posts tagged with

Comments

tvaughan

I'm not that familiar with Verity implementations, but I've got a guy in one of our support groups doing the verity implementation for the whole company. He has a dedicated 'Verity K2 Server' in our production environment that contains the 'collections' of searches that verity produces.

He came to me (the Interwoven admin) asking if he could get a TeamSite workarea for source controlling his Java code and for using "TeamSite as a GUI" to OpenDeploy that lets him push his code out to the production boxes.

In short, he doesn't generate collections off of individual workareas. . .he does it off a pre-production STAGING machine that is the recipient of Interwoven deployments.

Can I ask what the business drivers for your question are?

Tom

Johnny

Thanks for the reply.

We are doing a similar thing for versioning and deploying of Verity assets.

The driver is that we need our workarea's to operate in an isolated environment. Verity will be used other than for standard searching, such as page lists, so verity needs to be aware of new content inside a workarea before it hits another system outside of teamsite.

If verity is building collections off a pre production system only then virtualisation cant be contained to workareas.

Multiple collection sets per workarea seems the only way... an instance of verity per workarea would create a monster.

jkm

Hi Johnny,

We are planning to implement Verity on our website where metadata will be pushed from TS to production portal and then Verity will use that metadata to do search. We dont have multiple branches or WA in TS, so we may not ran into issues as you mentioned above.
This is the first time we are doing integration of Verity search on portal and TS and not much have much idea how to implement this and efforts required for this.
We may hire some experts if required in this area going forward but to evaluate this at the first place I need some thoughts from you people.

Can you provide some ideas for the same?

Thanks !
TS Ver 6.5 + Solaris

stefanmaier

Hi,
we also use a combination of TS6.5 (on windows) and Verity.

Some details:
For each of our websites we have multiple workareas in TS.
The website content is deployed to the webserver and to the Verity server
=> the verity server contains the full website content.
We use one verity collection per website filled by separate indexing jobs.

To add metadata to the indexed documents, we use two different solutions:
1. simple solution is to generate additional files which contains the verity metadata.
the metafiles are only deployed to the verity server and read out during the indexing process.
2. also it is possibly to add the metadata into the content files itself (e.g. Meta-Tags in HTML-Files).

jkm

Thanks Stefen for your reply
its opening doors for more questions Smiley Happy

=> the verity server contains the full website content.
>>>If verity server is installed on same production box where web server is running then verity server can always pick the latest content for indexing right?

We use one verity collection per website filled by separate indexing jobs.
>>>Can you brief the term verity collection? Does this mean one instance of Verity per website?

To add metadata to indexed documents
>>> We can supply metadata keywords from TS but how do you update those keywords in Verity search engine? Are you doing verity DB update for keyword updation in TS for a particular page.

I was thinking to split this in three step functionality, something on below lines
>> Generate keywords in TS using metatagger
>> Provide an option to user to update or enter new keywords
>> Supply those keywords to Verity DB
>> Portal will use verity index DB for search
>> Can we use DAS to do DB updates?

first thing, does this approach make any sense, hopefully far away from reality Smiley Happy

.
and if yes how to do the step 3. step 5 is just another weird thought.

Cheers!

stefanmaier

=> the verity server contains the full website content.
>>>If verity server is installed on same production box where web server is running then verity server can always pick the latest content for indexing right?

Yep, if you install verity engine and webserver on the same box.
(We installed the verity engine on a different box for security reasons...)

We use one verity collection per website filled by separate indexing jobs.
>>>Can you brief the term verity collection? Does this mean one instance of Verity per website?

No, instances and collections are different things:
A collection is a verity DB, where indexed documents are stored
(very, very short: "DB" means a structured repository which is optimized for fulltext search queries,
"DB" is not an SQL DB).

We use one collection per website, that means, that all documents of one website are indexed and stored in one verity DB.
To supply search results you need a verity process "k2server". k2server provides an access interface to request your verity DB from an application.
You can attach one or more collections to one k2server process. So it is for example possible to attach two collections from different websites to one k2server process.
(We do this for some lower used portals, so we have two separated web applications, which connect to the same port).
Also it is possible to start multiple k2server processes on one machine listening on different ports.
My recommendation is, that you start different k2server processes for security reasons (e.g. don't mix up intranet and internet content in one k2server) or for load reasons.

To add metadata to indexed documents
>>> We can supply metadata keywords from TS but how do you update those keywords in Verity search engine? Are you doing verity DB update for keyword updation in TS for a particular page.

We re-index the whole content page and its associated metadata, if the page itself or the metadata changed.

I was thinking to split this in three step functionality, something on below lines
>> Generate keywords in TS using metatagger
>> Provide an option to user to update or enter new keywords
>> Supply those keywords to Verity DB
>> Portal will use verity index DB for search
>> Can we use DAS to do DB updates?

The approach makes sense.
In most of our portals we generate an separate metafile additionally to every template-generated HTML-page and deploy this metafile to the verity server.
The indexing job combines HTML-content and metafile and fills this information into the verity collection.
The generation of a separate metafile is a good solution, if you don't want to add the metadata in the HTML-page directly.

As I know, there's no possibility to update a Verity DB (=collection) through DAS directly.

jkm

Thanks Stephen for your detailed reply. Let me look into the things more closely and i will come back with few more questions Smiley Happy

In the mean time can you post an example metafile to get an idea wats the content of metafile which Verity can understand.

stefanmaier

We use mkvdk for indexing our content. There's an option named "-bulk", with this option you can trigger an indexing job to index a bunch of files.
Input for mkvdk is a file which contains metadata and references to content.
Example for two files:[html]VdkVgwKey: /your/firstfile.html
Author: stefanmaier
LastModified: 2006-05-03
Keywords: one two three

_CACHE_FN: /filesystempath/to/your/website/your/firstfile.html
_CACHE_DELETE: 0
<>

VdkVgwKey: /your/secondfile.html
Author: someone
LastModified: 2006-04-01
Keywords: a b c

_CACHE_FN: /filesystempath/to/your/website/your/secondfile.html
_CACHE_DELETE: 0
<>[/html]Each first four lines are part of our template-generated files, the other lines are generated out of an indexing script, which triggers mkvdk.