Strange Problem With Primary Key - The Conflicts Appear
Oct 15, 2007
Hi!
There's really strange issue.
I have the table 'Resellers' with 'ResellerID' as identity column.
As a rule, the records are inserted into this table at publisher.
Usually, the subscribers just update this table.
After I added several new columns to the table, the strange conflicts had appeared at the subscriber side:
The merge process is retrying a failed operation made to article 'Resellers' - Reason: 'Violation of PRIMARY KEY constraint 'PK_Resellers'. Cannot insert duplicate key in object 'dbo.Resellers'.'.
If I look into MSmerge_conflict_Core_Resellers conflicts table at the subscriber, I see that 'ResellerID' column contains value '1'.
Of course, the record with such 'ResellerID' already exists.
But at the publisher the record with the given rowguid has quite other value!
Why?!
What is the reason of such behaviour?
I did re-init of the subscriber - no result, rather, all records which had conflicts appeared in the table, but the recent created records has conflicts again.
I got a problem with merge replication. I got a central sql server 2005 database on which i got a publication. Also there are 2 sql servers CE which are subscribers to that publication. I need to add some records on both mobile servers indepently but i'm using primary key as a user id in one table. So when i add a user on one PDA i use next available number in column ID. In the same time I add a user on other PDA with the same ID, because I don't know that there is such user with the same ID. Ok then I do synchronization. First PDA synchronizes with server but second tells me that there is a record with the same PK. And my question is. Can these be resolved writing a custom resolver or maybe you know others resolutions, because I think that is a typicall problem but couldn't find any solution other than using ie. HOST_NAME() function.
When I try to create a primary key for a table(alter table tablename add primary key( columnname)), I get a message " SQL 1918. ( Index (key=2) is being rebuilt)." Does any one have a clue, as to why do I get this. The primary key does get created ,but my application stops working, after getting this message.
I have two tables that are getting conflicts between the subscriber and the publisher, however I am pretty sure we only update these tables at the subscriber.
I have column level tracking turned on.
The tables both have Nonoverlapping, single subscription (3) set for partition options. I wonder does this do any maintenance to the tables?
I have put in place some triggers to audit what makes changes to the data, but won't know until tomorrow, so if there is something that anyone knows about that might help, please let me know.
I have a database on SQL Server 2005 and am using Access 2003 for my front-end. Today, I started having problems with ONE of the tables. I can't make changes - I get the "another user has made changes" and do not have the Save Changes option - only the Copy to Clipboard or Drop Changes options. I can make changes directly to the table using the Management Studio. I don't know if it's relative, but it happened as I was trying to add an "After Update" event to set a field equal to currentuser(). Do you have any idea what may be happening? Thanks
Can someone who has had direct experience with this tell me exactly what happens when a conflict (updating same record on two nodes at the same time) occurs in a P2P replication topology? Does the Dist. Agent throw an error? More importantly does the replication set continue to replicate the articles after any error occurs?
Can someone answer/point me to an article to handle update conflicts when replicating?
Say for example, you pull down data into your sql everywhere database and start making changes. Before you replicate, changes are made in the master database that conlift with changes you've made. For example, when you pull down your data there was a product with 4 units left. So maybe you place an order for that product for 4 units while offline. However, while you were offline someone else has taken all 4 of those units and when you replicate, you need to throw an error. How do you handle this situation?
Hi, I need to set an alert when a replication-conflict occurs rather that to check for conflicts manually. How can i accomplish this? I couldn't find the particular error-message to trap in an alert.
I have set up merge replication between 1 publisher and 9 subscribers (all push subscriptions). Distributor and publisher are located on the same machine.
Although everything seems to work fine from the outside (most of the time), there are a lot of conflicts in the conflict table for the replication. They appear all the time. There are a lot of "download insert failed" conflicts. They always look like in the following three screenshots: http://www.tronk.be/conflicts/conflict1.JPG http://www.tronk.be/conflicts/conflict2.JPG http://www.tronk.be/conflicts/conflict3.JPG
In the same way, there are also many "upload insert failed" conflicts.
In addition to this, there are some "update conflict"s (but a lot less than the other conflicts). Some of them show the same row at both sides: http://www.tronk.be/conflicts/conflict4.JPG Others show a different row at the both sides: http://www.tronk.be/conflicts/conflict5.JPG
The only thing that causes a real problem is the last screenshot, although I don't understand why the other conflicts are there (the inserst statements actually seem to happen anyway, even though there are conflicts). In case of the last screenshot, I can't find any place where an UPDATE actually happens at APP-STB, while I can clearly pinpoint the UPDATE at the other side (which is what actually comes from our program).
One more thing, the system is running on its limits, but all replication seems to be working fine.
I've been experiencing conflicts in my replication system that I can't seem to get my head around. The following is the scenario:
3 sqlservers, all running sql server 2005. Server B is the publisher and Server A and Server C (64 bit) are subscribers. The Queue Reader Agent runs on the publisher. I set up transactional repl with updateable subscriptions with the default conflict resolution policy of 'Publisher wins'.
There are 2 kinds of processes: 1. Nightly batch updates and 2. Daytime updates by real clients. The Nightly batch updates runs an on the publisher, which is B. Batch updates are massive updates and running it on the publisher makes sense and it works like a charm. Online updates are made on the subscriber 'C'. This subscriber is set to Queued update mode, and everyday I see a significant number of transactions that are detected as conflicts and the Publisher wins. As a result the changes made on Server C are getting lost. I have verified that no user/client is logged into Server B to do any updates. Users complain that their updates are lost. This is the most puzzling and frustrating bit. I don't see how a conflict can happen if nobody is updating data on the Publisher during the day. SQL Updates on Server C are getting rolled back on a conflict detection because the "Publisher wins", and SQL Inserts on server C are getting deleted because they don't exist on the publisher. Now, how can a insert done on the subscriber be marked as a conflict. There is no row on the publisher to compare the unique guid with, how can it be a conflict?
And the Queue Reader Agent crashes every 3-4 days. No useful information except it creates a dump file for which users have no tools to read it.
Has anyone seen this behavior ? Or is there a known bug in the QueueReader Agent?
My users are losing faith in the replication system and so am I.
I am using SQL Server 2000 Merge Replication. Sometimes when the data is replicated there are conflicts which when examined show it is due to voliation of foreign key constraint. But the data (keys) in already present in the master tables. Is there a way to give an order to the way the tables are replicated. This is so because i think the data in the details table is relpicated first instead of the master table. The conflicts are resolved properly when done using the conflict viewer.
Hi everyone, I'm creating a ASP.NET 2.0 web application utilizing sql server 2000 as a database. My problem revolves around multiuser acces with long running processes. Some of the pages in the application have long running processes against large tables in the database, lets say that take 2minutes to complete. My problem is how do I utilize ado properly in the rest of my application to display a message to users who may try to access data associated with a table while in the midst of one of its long running processes? For instance I would like to notify the user that "Table X is currently locked, please try again in a few minutes".Do I catch sqlException and examine the .Number property? Any insight is appreciated. Thanks.
I have heard that there are potential problems if you put service pack for NT on your server, in that it conflicts with SQL 6.5. Does anyone know of any such problems with either running 6.5 under these conditions or installing 6.5 if the service pack is already on the server. Thanks
I have maintenance plan on DBABC backup log to .trn job to run every 90 minutes (daily).
in order to keep the log file small, I also set up a job (T-SQL) to run at 4:15 am to backup log ABC with truncate_only, then run dbcc shrinkdatabase (DBABC, 10)
it looks "backup log ABC with truncate_only" has conflicts with the every90 minutes backup transaction log.
Question: could I keep the backup transaction log every90 minutes, but still could shrink the log file. The log file is growing very fast.
Or I have to use differential backup instead of backup tran log?
I have an Access XP ADE application connected to a SQL Server 7.0 SP4database. I have created a timestamp column in the main table.Unfortunately, I am now getting persistent write conflict errors.The order of operations are:1. The application starts and loads the recordset into the form using astored procedure.2. I modify a field and press a save button which uses me.dirty=false toforce a save.3. The field is saved to the database. Using profiler I can observe themodified field being saved. As I would expect, the update statement isusing the primary key and the timestamp column value. For the sake of thisdiscussion let's assume the value of the timestamp is 5ad9.4. Without navigating off the record, I alter the same field (or adifferent field) and press save again and a write conflict will appear.Using profiler I can see the update statement that is attempting to updatethe record. The update statement is using the previous value (5ad9) of thetimestamp column.I thought that the timestamp column value is incremented each time therecord is updated. The ADE application does not appear to be recognizingthe new timestamp value.Any help or advice you could give would be appreciated.ThanksGeorge
Hello all... I'm stuck, I cannot figure out how I should go about flaggingconflicts on a sheduling app. I currently have 8 columns (school grades)that have class over the course of 9 periods. I am populating the asp pagefine, and making changes to the database with forms lists. I need to compareall the results of one period (thats 8 results) so that i may find aclassroom conflict. Is there any solution in SQL?This is my query:sql = "SELECT * FROM schedule WHERE period ='"&num&"'"I step through this 9 times in a for/next loopThanks in advance!Alpay Eno
I am busy extending a VB6 app to talk to SQL Server Express 2005 and have come across a naming conflict. Some of the columns in the application's Access 97 tables is "index" which obviously exists as a T-SQL keyword and therefore any queries I perform including this column throws a syntax error. I tried prefixing the columns with their table names as in TableName.Index, but this still throws up the syntax error which I thought was a bit odd. I preferably want a fix that will be Access as well as T-SQL compliant but if that's not possible I will just write a string converter that does the job based on anyone's suggestions. TIA
i have both server instances installed on same machine, 1st is sql 2k enterprise and 2nd is sqlExpress 2005.
all was working fine till today, i renamed a catalog in the express2005 instance. the reason for this renaming was a duplicate catalog name in the 2k and in the 2005 so when i backup my dbs one is overwriting the other because of the same file name.
after this rename i started to get some errors in the application so i decided to rename it back to its original name and relocate the backup destination to other folder.
now the situation in that i cant login any more to the 2005 server, when i login to servsqlexpress2005 (via manager or via odbc) i see always the 2k server. i tried to login to any server name like .sqlexpress2005 or whatever i always get to see the 2k catalogs only.
in proccess list i see both sqlserv.exe proccesses, also in services all is working fine, i only cant access to this server in any way.
PLEASE HELP!!! should i reinstall 2005 instance? i dont wanna do that i have so much configs on it include very massive reporting services
***the more that i try to connect my 2k5 server the problem looks wierd more and more. when i connected as localhostsqlexpress2005 i succeeded... but in 2-3 minutes connection didnt turn off, but it switched to the 2k server... it looks like a major bug!!! how can connection switch himself between 2 server without even notice??
exec ('CREATE SYNONYM tblsynonym FOR ' + @cx + '..TableName')
---Procedure
drop synonym tblsynonym
The application i'm working on uses stored procedures that will at some point be called by more than one user at a time. At the start of the stored procedure the synonym is created and then it is dropped when the procedure completes, the issue is this: if two users access the same stored procedure at the same time then the first procedure will create the synonym and the second will fail because the syonym already exists.
- Transactional (5 articles - tables only) Users receive data based on their territory #, therefore they receive their customers sets of data. It happens that customer change from one territory to another but not frequently. When it happens, so far so good, the data is redirected to the new salesrep using the model we configured (Territory table with SUSER_NAME() to filter the data).
Ok, here's my problem. Since a while, I can see in the replication monitor that some users seems to log the same conflict again and again (Merge process). I mean, checking the history for many subscribers, there is always the same number in the "Conflict" colums. As an example: - Merge completed after processing 18 data change(s) (4 insert(s), 14 update(s), 0 delete(s), 31 conflict(s)) - Merge completed after processing 27 data change(s) (10 insert(s), 17 update(s), 0 delete(s), 31 conflict(s)) - Merge completed after processing 20 data change(s) (5 insert(s), 15 update(s), 0 delete(s), 31 conflict(s))
and so on...(Those are only 3 historical entries for a single subscriptions but there are many like that, always with the same count of conflict - vary per user). It appears to me that the same conflicts come over and over.
The thing is that if I decide to reinitialize a subscription, conflicts will disappear, therefore I know that it is not a process on the server that keeps changing the data; anyway, even if it was, changes would be applied on the subscription because the server always win in my setup.
Any idea what should I do with this? Any help would be greatly appreciated. Thanks.
I was hoping a sharp eye may be able to pick up on what I am doing wrong here, I am i must be making a comon mistake. If necessary I am happy to create and post a sample if necessary.
Problem: The publication parent table has a filter (defined below) and also has a join filter(defined below) to a child table. When changes are made to the parent and child table and synchronised a conflict is raised. For the child table for some reason the merge agent is saying an explicit update occured at the publisher but the values at the publisher have not changed at all and no sql has been executed to update the publisher rows. This in effect is stopping the changes to the child table being applied until i manually apply them in the conflict resolver. The strange thing is the conflict should not occur in the first place.
Perhaps my SQL server 2005 version? 9.00.2050.00 SP1 Standard Edition
Environment:
SQL Server 2005 - 9.00.2050.00 SP1 Standard Edition SQL Server Mobile Detail:
FKs are
FK: Job.JobScheduleID --> JobSchedule.ID
FK: JobDetail.JobID --> Job.ID
All three tables have int based IDs that are auto increment.
Publication Articles:
JobSchedule is download only
Job is Bidirectional, identity range management is MANUAL (only updates occur on this table)
JobDetail is Bidirectional, identity range management is MANUAL (only updates occur on this table)
Filters are of the following form:
Filter Job:
SELECT <published_columns> FROM [dbo].[Job] WHERE convert(nchar,[companyID])=Host_Name() AND [JobCompletedDate] IS NULL AND
( [JobScheduleID] in (SELECT distinct ID from JobSchedule where GETDATE() BETWEEN [JobSchedule].[start] AND [JobSchedule].[end]) )
Join Filter Job --> JobDetail
SELECT <published_columns> FROM [dbo].[Job] INNER JOIN [dbo].[JobDetail] ON [Job].[ID] = [JobDetail].[JobID] Now the first thing to note with respect to using GetDate() i have read http://msdn2.microsoft.com/en-us/library/ms365153.aspx so i thought that i would remove that portion of the Filter on the Job table just to see what happens.
SELECT <published_columns> FROM [dbo].[Job] WHERE convert(nchar,[companyID])=Host_Name() AND [JobCompletedDate] IS NULL
This still did not resolve the issue. I then Remove the AND [JobCompletedDate] IS NULL and it then started working fine (cool) but of course no longer satisfys the rule i want to create (seriously uncool ).
Sadly the hostnames for the boxes I am trying to replicate are similar:
- db1.nyc.mydomain.tld - db1.sac.mydomain.tld
The servers can talk to each other over all necessary ports however when I generate a push subscription from the publisher, I am asked to add the SQL Server subscriber and it already shows DB1 listed since itself is DB1.nyc. How can I attach the remote subscriber when setting up the subscription from the publisher if the hostname conflicts in my unique scenario?I created a entry on the publishers host file to use db2sac for the SAC IP and entering that alias in as the remote subscriber but no dice.
I think I will have a problem with Id’s when I do merge replication. The problem is, I have several clients (local shops) who can do additions to a (Customer) table. Other tables will link to that table (things that a customer buys) with their foreign key, so the id of this table is very important. Now, twice a day that table with customers is merged from all clients to the central server. (the table with things that the customer buys, does not needs to be merged as it only remains local).
Possible problem: if we assign ranges to identifiers for each client (thus local shop), which will be the best option we found until so far, how can we then detect the doubles? (with doubles, we mean that when the same customer is added in two local shops for a first time, then we will have the same customer twice in the central database)(afterwards, when we delete one customer row in the central database, we might have a id conflict when we merge back to the clients -as a customer might have already bought some goods at the two different shops meanwhile-)
What’s the common practice to resolve this? (we need to do Merge Replication, because during the day, our local shops are disconnected from the central server)
We have an issue in a SCD where a number of records may be presented that have changes to their attributes of EVERY type.
Example,
BusinessKey: xxxxxxxx
BuildingTypeId: 7
BusinessUnitHistoryId: 4019
BusinessUnitId: 4019
CurrencyId: 26
DevelopmentTypeId: 14
MarketId: 182
Name: abcdefgh
CurrencyId is a fixed attribute MarketId & BuildingTypeId and the BusinessUnitId & BusinessUnitHistoryIds are historical attributes Name is a changing attribute
The behaviour of the ETL seems to suggest that if fixed attribute changes are detected, these rows will error and therefore the changing & historical attributes will NOT be amended during the SCD transformation. Is this correct... as it seems to be what is happening.
We use autogenerated primary keys in most of our tables. Some of these keys are also foreign keys in other tables. Right now there is only 1 database sever at a central location. But now there is a need to have multiple database servers at different locations. Data from these remote sites needs to be replicated to the central server. Some data would also distribute from central server to selected remote sites.
If I could resdesign, I would have choosen something like GUIDs for the primary keys or combination of something like ServerName and AutoGenerated number as a combined key. But that's not possible right now. How do I handle merge replication conflicts in this case?
I am looking for some pointers as to how to handle this case. If it were just simple table with 1 primary key, that would be easy as I can throw the primary key on remote server and let the central server create a new key when data is inserted. But in my case, a single table can be related to 5 or more other tables through these autogenerated keys. Any help is much appreciated.
I have merge replication running on SQL Server 2005 Developer Edition. To be able to use my own conflict handler I implemented and deployed my own BusinessLogicModule written in C#. Till here everything works fine but I do not know how to handle the following scenarios:
1) When, while synchronization, an insert error occurs for example because of an unique key constraint on some rows, I do not see a way to solve that conflict from inside my conflict handler. One option would for example to delete the conflicting row at the subscriber or the publisher. But I do not know how to do it from inside my custom conflict handler.
2) If you override for example the method UpdateHandler from BusinessLogicModule you have access to the Publisher and the Subscriber DataSet that have an update conflict. This DataSet only contains one Table with one Row. Is there a way to define replication so that I have referenced tables in that DataSet, too? Unfortunately, Logical Records are not an option because they do not work with custom conflict handlers.
I have SQL CE clients replicating against a SQL Server 2005 db using merge replication. The DB has a table A and a table B, which has a foreign key to table A. It is common in my application for records in table A to be deleted on the server. I'm running into issues when a table A record has been deleted, but table B records were created on the clients which point to that record. When I sync I get a conflict because the table B records cannot be applied at the server, and the table A delete cannot be applied at the client.
What I would like to happen is to have the table B records on the client be deleted by the merge process, and to create a log of the event. I've looked into creating a business logic handler to do this, but I'm not sure what type of conflict this is (UpdateDeleteConflict or otherwise), and I'm not sure that deleting the table B records is something I can do in the business logic handler.
This seems like it would be a common problem in merge replication. I'm not locked into using a custom business logic handler at all. Any suggestions are welcome.
There is an error in one of my merge publication. The error is,
The change for the row with article nickname 2336003 (test), rowguidcol {436456F0-F5AD-E411-80CF-5CF3FC1D2D76} could not be applied at the destination. Further information about the failure reason can be found in the conflict logging tables.
When i checked my tables I got following values in rowguid column
We are using a mix of SQL 2005 and 2000 servers and our "main" database server is running SQL 2005 x64 (SP2 ver. 3042).
Our system has run perfectly for months, then subsequent to an SP2 update we are seeing several instances where the data record counts are different for several tables among all the servers.
We are using Merge Replication, with no filters and published every 2 minutes.
Publisher is 2005 x64, subscribers SS2000 (SP3) and SS2005 x64. Pull agents, no filters on subscriptions. We are seeing many seemingly random conflicts on between SS2000 subscriber and publisher. It happens on several different tables.
One table is never editted, only inserts happening everywhere and deletes happening on the SS2000 subscriber. Deletes will sometimes generate conflict. Reason is '"he row was deleted at 'CTS11.CTS' but could not be deleted at 'cts4a.cts'. Unable to synchronize the row because the row was updated by a different process outside of replication." CTS11 is SS2000 subscriber, CTS4A is publisher.
Probably unrelated bug but when looking at conflicts on this same table in SS2005 conflict viewer, get error "ID is neither a DataColumn nor a DataRelation for table summary (System.dATA)" and then "Column ID does not belong to table summary (System.Data)". ID column is rowguid, only unusual thing about table is that it has varchar(8000) field plus some other fields.
Other tables generate conflicts with this reason "The row was updated at 'CTS11.CTS' but could not be updated at 'cts4a.cts'. The merge process was unable to synchronize the row." I enabled verbose logging in the merge agent but the log file didn't contain any further explanation.
This same topology and schema worked fine when all publishers and subscribers were SS2000.
Any insight into how to fix this would be appreciated.
