Author |
Topic |
dineshrajan_it
Posting Yak Master
217 Posts |
Posted - 2009-04-21 : 03:08:21
|
hi,I need to track the changes made to table like insertion, deletion and updation and put them into a new table. No triggers pls. How to do that |
|
SwePeso
Patron Saint of Lost Yaks
30421 Posts |
Posted - 2009-04-21 : 03:41:51
|
Make the necessary changes to the existing CRUD procedures. E 12°55'05.63"N 56°04'39.26" |
|
|
dineshrajan_it
Posting Yak Master
217 Posts |
Posted - 2009-04-21 : 04:27:14
|
Hi Peso, The assignment given to me is to log user actions(i.e. editing profile, updating orders etc) into Db. So i need to get these stuffs loaded into logtable through table tracking method(insert, update, etc). I was instructed not to use triggers as it is cost expensive and also i should not lay my hands on existing procedures. what can i do. Pls help me out |
|
|
SwePeso
Patron Saint of Lost Yaks
30421 Posts |
Posted - 2009-04-21 : 04:36:25
|
If your assignments doesn't allow you to1. Use trigger2. Change existing codetell your boss you can't do it, unless you install Microsoft SQL Server 2008 Enterprise Edition which supports CDC (Change Data Capture). E 12°55'05.63"N 56°04'39.26" |
|
|
SwePeso
Patron Saint of Lost Yaks
30421 Posts |
Posted - 2009-04-21 : 04:40:19
|
Well, there is one option left.You can add a new column to the tables you want to monitor, asALTER TABLE MyTableADD MyColumn DATETIME AS (GETDATE())But that can break existing code and work for INSERT only, not update.Another option is to keep a complete copy of database somewhere else and have a job run a stored procedure which compares the two databases. It will be very slow. E 12°55'05.63"N 56°04'39.26" |
|
|
SwePeso
Patron Saint of Lost Yaks
30421 Posts |
Posted - 2009-04-21 : 04:40:37
|
Why do you think a proper written trigger will be slow? E 12°55'05.63"N 56°04'39.26" |
|
|
dineshrajan_it
Posting Yak Master
217 Posts |
Posted - 2009-04-21 : 04:43:30
|
Thanks Peso,I will ask for enterprise edition in my company. anyway, how did Sql developers actually handled those User Actions into table in earlier Sql versions. Will they be inserting to a log table manually in stored procedure each time record gets inserted, updated. That will be time-consuming right. May be if i dont get enterprise edition getting installed in my machine, ur answer may help. |
|
|
dineshrajan_it
Posting Yak Master
217 Posts |
Posted - 2009-04-21 : 04:47:03
|
I dont why not to use triggers as suggested by my PL. may be there are 100 tables. calling triggers for each table will be time consuming? ur suggestions |
|
|
dineshrajan_it
Posting Yak Master
217 Posts |
Posted - 2009-04-21 : 05:08:01
|
Hi Peso,I have only Sql Server Express Edition. CDC is available only in Sql server Enterprise Edition. Then how can i keep track of user actions into DB. Other than triggers. Pls help me. Ur suggestions r valuable |
|
|
SwePeso
Patron Saint of Lost Yaks
30421 Posts |
Posted - 2009-04-21 : 05:16:47
|
Have your PL email me and I can explaing a thing or two for him/her about triggers. E 12°55'05.63"N 56°04'39.26" |
|
|
dineshrajan_it
Posting Yak Master
217 Posts |
Posted - 2009-04-21 : 05:35:46
|
Yes Sure. thanks |
|
|
dineshrajan_it
Posting Yak Master
217 Posts |
Posted - 2009-04-21 : 06:25:07
|
Hi Peso,Can i go for 1 table -> 1 history table or 1 history table for entire db to store old values. |
|
|
Transact Charlie
Master Smack Fu Yak Hacker
3451 Posts |
Posted - 2009-04-21 : 07:20:57
|
Post probably continuing here:http://www.sqlteam.com/forums/topic.asp?TOPIC_ID=124218Charlie===============================================================Msg 3903, Level 16, State 1, Line 1736The ROLLBACK TRANSACTION request has no corresponding BEGIN TRANSACTION |
|
|
whitefang
Enterprise-Level Plonker Who's Not Wrong
272 Posts |
Posted - 2009-04-21 : 07:34:55
|
Triggers are bad for debugging (in a properly written application, triggers are not needed. It's just a tool for lazy developers/dbas), I wouldn't say they affect performance.I'll quote my other post:quote: I'd use one history table for each database with the following columns:Id, TargetTable, ModifierId, RowData, ModifierDate, CommandTypeWhere ModifierId is the user who invoked the modification of the row, and RowData is a XML data column containing the whole row compiled into an XML format. The CommandType is either Update or Delete.
|
|
|
spirit1
Cybernetic Yak Master
11752 Posts |
Posted - 2009-04-21 : 07:48:24
|
triggers have their use and they can be a powerful tool even in a properly written application. of course they can cause a perf bottleneck if they use a cursor instead of joining to inserted and deleted pseudo tables.and yes they are really bad for debugging ___________________________________________________________________________Causing trouble since 1980Blog: http://weblogs.sqlteam.com/mladenpSpeed up SSMS development: www.ssmstoolspack.com <- version 1.5 out! |
|
|
whitefang
Enterprise-Level Plonker Who's Not Wrong
272 Posts |
Posted - 2009-04-21 : 07:59:22
|
The topic poster said they cannot use triggers and cannot change the main code.In that case, there is one last approach left. If your application is using a data access utility (as all properly written applications should) or a centralized data CRUD method, you can modify it so that at the point of an Update/Delete, you can parse the incoming SQL query and get the ID and table name then perform backup of that row before running the incoming query. With this approach, there is only one place to modify the code and the rest of the application is unaffected. The only negative thing about this approach is that you will be using a modified data access utility that is separate from the rest of your applications. |
|
|
dineshrajan_it
Posting Yak Master
217 Posts |
Posted - 2009-04-21 : 08:06:15
|
hi,i used ur approach by placing the previous query records as xml.HistoryTableHistoryId PreviousRec(Xml)1 <SalesTable salesid='1' amt='500'/><SalesTable salesid='1' amt='700' />say if i have modified the table records 100 times then 100 nodes will be created for historyid. will that degrade performance. whats suitable size suggested for xml datatype in each row. |
|
|
SwePeso
Patron Saint of Lost Yaks
30421 Posts |
Posted - 2009-04-21 : 08:09:12
|
And you think this will be faster than a trigger?And for debugging thingy. Well, if you use SQL Server 2008, it's incorporated into the debugger.CREATE TABLE tblA ( i INT )GOCREATE TABLE tblB ( j INT )GOCREATE TRIGGER trgA ON tblA AFTER INSERTAS INSERT tblB SELECT i FROM insertedGOINSERT tblA SELECT 1GOSELECT * FROM tblBGOPRINT 'Done'GO E 12°55'05.63"N 56°04'39.26" |
|
|
whitefang
Enterprise-Level Plonker Who's Not Wrong
272 Posts |
Posted - 2009-04-21 : 08:14:39
|
quote: Originally posted by dineshrajan_it hi,i used ur approach by placing the previous query records as xml.HistoryTableHistoryId PreviousRec(Xml)1 <SalesTable salesid='1' amt='500'/><SalesTable salesid='1' amt='700' />say if i have modified the table records 100 times then 100 nodes will be created for historyid. will that degrade performance. whats suitable size suggested for xml datatype in each row.
No, that won't degrade performance. If you're using an object model, you can even serialize the entities directly into the XML data column. Also, make sure to check that you don't have any concurrency issues depending on your application reqs.You really don't have to worry about performance unless you have like over 10 million records. And that is when you should consider archiving. |
|
|
whitefang
Enterprise-Level Plonker Who's Not Wrong
272 Posts |
Posted - 2009-04-21 : 08:16:21
|
quote: Originally posted by Peso And you think this will be faster than a trigger?And for debugging thingy. Well, if you use SQL Server 2008, it's incorporated into the debugger.
Will his client notice a millisecond difference? I think not. We can all assume that the application is not enterprise grade. He already said he cannot use triggers. |
|
|
dineshrajan_it
Posting Yak Master
217 Posts |
Posted - 2009-04-21 : 08:42:40
|
I have 120 tables so using trigger for all 120 tables will be i dont know? and this single history table theory for an entire db. with previous records as xml format. we can actually avoid keeping history table for each table and also can keep a centralised single history table for entire db. any thoughts will be helpful. thanks |
|
|
Next Page
|