Missing data after server restart

[former_member997604]

Hi,

We restarted our MDM server at the OS level last night. Everything seemed okay after the restart. However, when our users came in this morning, they found that some data was missing from a table. We have a Material repository with 3 main tables. Some data which have previously been populated was not appearing in our Products Detail table, the largest of the three. A large number of rows in the table had columns that failed to load with data on the repository auto-start that occured after the server restart. The auto start value in the mds.ini file was set at 300 (5 minutes) as we had seen this iisue happen once before, and thought that may help.

A stop and restart of the repository corrected the issue. Is there anything we could have done to prevent this? Is it worthwhile to try increasing the auto start delay again? Validating detail data after every server restart is not an effective option.

MDM version: 7.1 SP08

OS: AIX DB: Oracle 11g

Thanks!

Dave

[former_member997604]

We'll look in to changing our process to start repositories manually with update indicies rather than using autostart. This seems to defeat the purpose of the autostart feature, but if it resolves this issue then it's worth it.

Thanks for your input.

[Former Member]

Hi Dave,

It seems you had not Loaded your repository with Update indices for quiet sometime.

It is a good practise to do so especially when repository has undergone various data/structural changes.

Auto Start parameter wiith greater value will not be of much help here.

However in an event of server crash, greater parameter value should help you notice this behavior.(crash)

Thanks,

Ravi

[former_member997604]

Hi Ravi,

Thanks for your response. When I look in the MDM Console Guide, I see the following:

When you choose Immediate, MDM automatically updates indices that it knows to be out of date as a result of changes you make to the MDM repository data through MDM Data Manager. However, when the structure of the repository has been modified through MDM Console, MDM cannot know precisely which indices need to be updated. In these cases, you should choose Update Indices rather than Immediate when you load the repository.

We always do update indices when the structure changes, but are you suggesting it is also necessary to update indices periodically just due to data accumulation? If so, what is the best practice? Have you scheduled this to happen on a regular basis?

Thanks,

Dave

[Former Member]

Hi Dave,

I meant that when you load repository using Immediate data is loaded from .acl files (accelerator files) in a async manner.

Whereas when you load by Update Indices new acl files are generated.

Below men6tioned is a comparision between the 2 ways of laoding repository:

Load Repository Immediate

o Loads indexes from accelerator files instead of being regenerated by reading the data them from the DB and creating the index.

o Automatically builds indexes stored in out-of-date accelerator files.

o Is usually fast.

 Load Repository with Update Indices

o Rebuilds all MDM indexes files by reading and indexing data stored in the DB. The results are flushed on disk as accelerator files.

o Necessary by repository schema changes or if any error occurs.

o Might take significant amount of time to complete.

So when you observe some visual data impact in your repository it is good idea to rebuild indexes.

I guess we cannot schedule this as default behavior is to load by immediate even through the inin parameter.

as a best practice one can do this by fixing a time frame eg monthly depending on data/schema changes.

Thanks,

Ravi

[Former Member]

Hi

Yes it really works, When ever we do a start of repository , the best practise is to load the repository with Update Indices.

This is a lot times happened with me .

Regads

Shifali