Merge Replication to multiple servers losing data

Hi,
We are using merge replication between four servers (1 publisher & 3
subscribers). The same articles (tables) are in each publication. I run the
application which changes data in some tables and adds records in another
table. The inserted data is immediately updated via trigger. After running
the application all expected data is present. If I manually force
replication to each subscriber sequentially, all expected data is present.
If I run replication between the servers at the same time, the table to
which data was added will lose some data. The data lost was not the data
that was just added. We are running SQL Server 2000 SP3a on all servers. Any
ideas?
tia,
Paul
Look at the 'view conflict' at replication monitor...
"PaulW" <MSNewsGroup@.Digi-Sol.com>
news:urv9zNxlHHA.1216@.TK2MSFTNGP03.phx.gbl...
> Hi,
> We are using merge replication between four servers (1 publisher & 3
> subscribers). The same articles (tables) are in each publication. I run
> the application which changes data in some tables and adds records in
> another table. The inserted data is immediately updated via trigger.
> After running the application all expected data is present. If I manually
> force replication to each subscriber sequentially, all expected data is
> present. If I run replication between the servers at the same time, the
> table to which data was added will lose some data. The data lost was not
> the data that was just added. We are running SQL Server 2000 SP3a on all
> servers. Any ideas?
> tia,
> Paul
>
|||There are no recorded conflicts. This table only has data inserted, then
updated through a trigger. We did view the transaction log. The only entries
with the table were the inserts we initiated followed by a delete/insert for
the trigger update.
Paul
"Grigoris Tsolakidis" <gcholakidis@.spam_remove.hotmail.com> wrote in message
news:uhRQsHGmHHA.4852@.TK2MSFTNGP03.phx.gbl...
> Look at the 'view conflict' at replication monitor...
> "PaulW" <MSNewsGroup@.Digi-Sol.com>
> news:urv9zNxlHHA.1216@.TK2MSFTNGP03.phx.gbl...
>
sql

Merge replication losing changes.

Hi,

I am facing a wierd problem while sync'ing. When I am do a synchonize from a mobile device to the server DB, one of my table column fails to update while all other columns get updated.

The table uses row level tracking (tried making it column level too) and the field which fails to update is a date field (Nullable).

This happens in case of syncing after updating >1 records on device. But syncing after updating just 1 record on device, this date field gets updated as expected.

Note: I have filters set for this table. The filter downloads fields only with this date field=null.

I need to downlaod all records meeting this filter condition but at the same time, this filter should not be applied while uploading changes as I think this is what the problem is.

If anyones has faced this and got a solution, please let me know.

Thanks.

What exactly does your filter look like?

|||

Do you have any default constraints on the date column or any triggers that could be affecting this column and hence affecting your expected output?

Also are you using vertical partitioning where some of the columns in the table do not go to the subbscriber? And is it that this date column is positioned after a column that is vertically filtered out?

It would be great if you can narrow down the problem and send the publication/article creation scripts.

Typically this is how a filter works:

Lets say you have a filter subset_filterclause= 'age > 20'

Now initially all the rows at the publisher that match this criteria will be downloaded, others that dont qualify will not be downloaded.

Furthermore, inserts/updates at the publisher which make rows qualify this filter will be downloaded to the subscriber.

Finally, all inserts/updates at the subscriber whether they qualify the filter or not will be uploaded to the publisher. Now all the rows that did not qualify the filter will be downloaded as deletes to the subscriber and will be delete at the subscriber.

I am trying to understand you scenario and here is what I understand:

you are saying that if you update 1 row from age=40 to age=20, you will see this update go to the publisher. But if you update 10 rows from age=40 to age=20, they will not go to the publisher?

Merge replication losing changes.

Hi,

I am facing a wierd problem while sync'ing. When I am do a synchonize from a mobile device to the server DB, one of my table column fails to update while all other columns get updated.

The table uses row level tracking (tried making it column level too) and the field which fails to update is a date field (Nullable).

This happens in case of syncing after updating >1 records on device. But syncing after updating just 1 record on device, this date field gets updated as expected.

Note: I have filters set for this table. The filter downloads fields only with this date field=null.

I need to downlaod all records meeting this filter condition but at the same time, this filter should not be applied while uploading changes as I think this is what the problem is.

If anyones has faced this and got a solution, please let me know.

Thanks.

What exactly does your filter look like?

|||

Do you have any default constraints on the date column or any triggers that could be affecting this column and hence affecting your expected output?

Also are you using vertical partitioning where some of the columns in the table do not go to the subbscriber? And is it that this date column is positioned after a column that is vertically filtered out?

It would be great if you can narrow down the problem and send the publication/article creation scripts.

Typically this is how a filter works:

Lets say you have a filter subset_filterclause= 'age > 20'

Now initially all the rows at the publisher that match this criteria will be downloaded, others that dont qualify will not be downloaded.

Furthermore, inserts/updates at the publisher which make rows qualify this filter will be downloaded to the subscriber.

Finally, all inserts/updates at the subscriber whether they qualify the filter or not will be uploaded to the publisher. Now all the rows that did not qualify the filter will be downloaded as deletes to the subscriber and will be delete at the subscriber.

I am trying to understand you scenario and here is what I understand:

you are saying that if you update 1 row from age=40 to age=20, you will see this update go to the publisher. But if you update 10 rows from age=40 to age=20, they will not go to the publisher?