Does the old index store state in the database?

Imar Spaanjaars

Posted on 07/10/2015 10:34:19

Hi there,

We have a load balanced setup using DFS where all files are automatically synced. This causes locking issues with the index files, so we've excluded the _search folder. This means the index UI in the Management Center no longer works for us as it would only update the current server. To overcome this, we developed a simple module that executes the following URLs for the connected nodes:

http://10.10.86.10/Admin/Public/IndexUpdateTask.ashx?Path=Products&FullUpdate=True
http://10.10.86.20/Admin/Public/IndexUpdateTask.ashx?Path=Products&FullUpdate=True
http://10.10.86.50/Admin/Public/IndexUpdateTask.ashx?Path=Products&FullUpdate=True

This works OK when I execute each URL individually. However from my module I am executing all three of them in parallel which leads to the following error:

This instance has already started one or more requests. Properties can only be modified before sending the first request.

It seems that the indexing state is kept in the database, causing the first server to accept the request and build the index but the others then fail as Dynamicweb thinks it's already busy building the index.

Is this a correct analysis and is there a way to execute them in parallel? As a work around, I can execute them one by one, but that triples the time it takes to build the index across all nodes.

Thanks,

Imar

Replies

Nicolai Høeg Pedersen

Posted on 07/10/2015 23:58:04

This post has been marked as an answer

Hi Imar

I do not think DW does that.

The exception seems to be coming from HttpClient: http://stackoverflow.com/questions/23790893/dynamically-changing-httpclient-timeout-in-net

Yours or ours?

Votes for this answer: 1

Imar Spaanjaars

Posted on 08/10/2015 00:10:21

Oh wow, I never thought of that. I saw this in the logs of my custom tool and it appeared to be the response from the call to the service, not from the client executing it. I never even bothered to Google for it, thinking it came from Dynamicweb (with "instance" referring to the indexer).

I now see this is defintely mine. I am using some async tasks to schedule the index updates in parallel, reusing a single instance of HttpClient. The method that executes the request does indeed change a property (a timeout), so the second and subsequent calls would result in this error message.

Thank you; you just made my day (or evening, actually)

Imar

You must be logged in to post in the forum

Developer forum

Does the old index store state in the database?

Replies