Deletion stubs

HCL Domino Domino Forum
1

Hello everybody,I’ve a critical situation with the deletion stubs.

I’ve 100 external users that replicates with one database on the servers.

For reasons that I don’t explain here, I’ve 704.000 deletion stubs on the server AND in the external replicas (unexpected volume of data).

As you can image I begin to have some performance problems, so we start to play around with the space savers parameters that define for how many days the stubs are kept on the server.

All the “experiments” are made on an isolate enviroenment and these are the results:

  1. if I reduce the purge interval on the server it deletes the stubs but, when the client made a replica, nothing change on the client

  2. if I reduce (and at the limits I put it to zero) the purge interval on the client, when it does a replica on the server nothing change but my client database is smaller and if I do some 2 ways replica no older documents reappears.

Note that the 99,9% of the deletion are made on the server.

I’d like to ask you if anyone can confirm that this behaviour is correct and standard because if I can put to zero the purge interval on the clients and let the purge interval to 90 on the server it would be the true solution for my problems.

I’ve to absolutely sure because the application is very mission critical and I can’t do something wrong!

Any help will be appreciated.

Thanks to you all in advance

Alessandro

2

Subject: RE: Deletion stubs

Your observations sound like what I would expect,

If users are authoring data, you might observe documents not replicating from them to the server when you set the cutoff days to a low number in their local replicas. Other than that, I don’t foresee a problem.

HOWEVER…

The solution to your difficulty is not to mess with the replication settings, but to change your data synchronization algorithm so that rather than deleting all the documents and creating new ones, you compare your source data with the data in Domino and only update or delete information as needed. Why do you make your users receive all-new data during replication?

3

Subject: RE: Deletion stubs

Thanks for your response.The problem is not the replica algorithm but the arhitecture of the application. We built it expecting a totally different amount of data (a lot smaller) and in that case the replica conflict would have been a small number. Because of replica conflict in our case are “trash” we deleted them.

When we found out the the volume of data is bigger (a lot bigger that what we expected) we had a problem. Now we are changing the application but we need to get ridden of the older deletion stubs…

Thanks again.

Alessandro