I am inserting data into a database using millions of insert statements stored in a file. Is it better to insert this row by row or in bulk ? I am not sure what the implications can be.
Any suggestions on the approach ? Right now, I am executing 50K of these statements at a time.
Generally speaking, you're much better off inserting in bulk, provided you know that the inserts won't fail for some reason (i.e. invalid data, etc). If you're going row by row, what you're doing, is opening the data connection, adding the row, closing the data connection. Rinse wash, repeat in your case tens of thousands of times (or more?). It's a huge performance hit as opposed to opening the connection once, dumping all the data at one shot, then closing the connection once. If your data ISN'T a clean set of data, you might be better off going row by row, as the bulk insert won't fail if you have data to be cleaned up.
If you are using SSIS, I would suggest a data flow task as another possible avenue. This will allow you to move data from a flat text file, SQL table or other source and map it into your new table. Performance, I have found, is always pretty good and I use it regularly.
If your table is not created before the insert, what I do is drag an Execute SQL Task function into my process with the table creation query (CREATE TABLE....etc.) and update the properties on the data flow function to delay validation.
As long as my data structure is consistent, this works. Here are a couple screenshots.
You should definitely use the BULK INSERT instead of inserting row by row. The BULK INSERT is the in-process method designed for bringing data from a text file into SQL Server, ant it is the fasted among other approaches described in the The Data Loading Performance Guide online article
The other alternative is to use a batch process that uses set-based processing over a smaller set of records (say 5000 at a time) . This can keep the server from getting totally locked up and is faster than one record at a time.
Related
I have an ETL process that will run periodically. I was using kettle (PDI) to extract the data from the source database and copy it to a stage database. For this I use several transformations with table input and table output steps. However, I think I could get inconsistent data if the source database is modified during the process, since this way I don't get a snapshot of the data. Furthermore, I don't know if the source database would be blocked. This would be a problem if the extraction takes some minutes (and it will take them). The advantage of PDI is that I can select only the necessary columns and use timestamps to get only the new data.
By the other hand, I think mysqldump with --single-transaction allows me to get the data in a consistent way and don't block the source database (all tables are innodb). The disadventage is that I would get innecessary data.
Can I use PDI, or I need mysqldump?
PD: I need to read specific tables from specific databases, so I think xtrabackup it's not a good option.
However, I think I could get inconsistent data if the source database is modified during the process, since this way I don't get a snapshot of the data
I think "Table Input" step doesn't take into account any modifications that are happening when you are reading. Try a simple experiment:
Take a .ktr file with a single table input and table output. Try loading the data into the target table. While in the middle of data load, insert few records in the source database. You will find that those records are not read into the target table. (note i tried with postgresql db and the number of rows read is : 1000000)
Now for your question, i suggest you using PDI since it gives you more control on the data in terms of versioning, sequences, SCDs and all the DWBI related activities. PDI makes it easier to load to the stage env. rather than simply dumping the entire tables.
Hope it helps :)
Interesting point. If you do all the table inputs in one transformation then at least they all start at same time but whilst likely to be consistent it's not guaranteed.
There is no reason you can't use pdi to orchestrate the process AND use mysql dump. In fact for bulk insert or extract it's nearly always better to use the vendor provided tools.
When I am using the above syntax in "Execute row script" step...it is showing success but the temporary table is not getting created. Plz help me out in this.
Yes, the behavior you're seeing is exactly what I would expect. It works fine from the TSQL prompt, throws no error in the transform, but the table is not there after transform completes.
The problem here is the execution model of PDI transforms. When a transform is run, each step gets its own thread of execution. At startup, any step that needs a DB connection is given its own unique connection. After processing finishes, all steps disconnect from the DB. This includes the connection that defined the temp table. Once that happens (the defining connection goes out of scope), the temp table vanishes.
Note, that this means in a transform (as opposed to a Job), you cannot assume a specific order of completion of anything (without Blocking Steps).
We still don't have many specifics about what you're trying to do with this temp table and how you're using it's data, but I suspect you want its contents to persist outside your transform. In that case, you have some options, but a global temp table like this simply won't work.
Options that come to mind:
Convert temp table to a permanent table. This is the simplest
solution; you're basically making a staging table, loading it with a
Table Output step (or whatever), and then reading it with Table
Input steps in other transforms.
Write table contents to a temp file with something like a Text File
Output or Serialze to File step, then reading it back in from the
other transforms.
Store rows in memory. This involves wrapping your transforms in a
Job, and using the Copy Rows to Results and Get Rows from Results steps.
Each of these approaches has its own pros and cons. For example, storing rows in memory will be faster than writing to disk or network, but memory may be limited.
Another step it sounds like you might need depending on what you're doing is the ETL Metadata Injection step. This step allows you in many cases to dynamically move the metadata from one transform to another. See the docs for descriptions of how each of these work.
If you'd like further assistance here, or I've made a wrong assumption, please edit your question and add as much detail as you can.
I am migrating data from an Oracle database to a SQL server 2008 r2 database using SSIS. My problem is that at a certain point the package fails, say some 40,000 rows out of 100,000 rows. What can I do so that the next time when I run the package after correcting the errors or something, I want it to be restarted from the 40,001st row, i.e, the row where the error had occured.
I have tried using checkpoint in SSIS, but the problem is that they work only between different control flow tasks. I want something that can work on the rows that are being transferred.
There's no native magic I'm aware of that is going to "know" that it failed on row 40,000 and when it restarts, it should start streaming row 40,001. You are correct that checkpoints are not the answer and have plenty of their own issues (can't serialize Object types, loops restart, etc).
How you can address the issue is through good design. If your package is created with the expectation that it's going to fail, then you should be able to handle these scenarios.
There are two approaches I'm familiar with. The first approach is to add a Lookup Transformation in the Data Flow between your source and your destination. The goal of this is to identify what records exist in the target system. If no match is found, then only those rows will be sent on to destination. This is a very common pattern and will allow you to also detect changes between source and destination (if that is a need). The downside is that you will always be transferring the full data set out of the source system and then filtering rows in the data flow. If it failed on row 99,999 out of 1,000,000 you will still need to stream all 1,000,000 rows back to SSIS for it to find the 1 that hasn't been sent.
The other approach is to use a dynamic filter in your WHERE clause of your source. If you can make assumptions like the rows are inserted in order, then you can structure your SSIS package to look like Execute SQL Task where you run a query like SELECT COALESCE(MAX(SomeId), 0) +1 AS startingPoint FROM dbo.MyTable against the Destination database and then assign that to an SSIS variable (#[User::StartingId]). You then use an expression on your select statement from the Source to be something like "SELECT * FROM dbo.MyTable T WHERE T.SomeId > " + (DT_WSTR, 10) #[User::StartingId] Now when the data flow begins, it will start where it last loaded data. The challenge on this approach is finding those scenarios where you know data hasn't been inserted out of order.
Let me know if you have questions, need things better explained, pictures, etc. Also, above code is freehanded so there could be syntax errors but the logic should be correct.
I am developing SSIS packages that consist of 2 main steps:
Step 1: Grab all sorts of data from existing legacy systems and dump them into a series of staging tables in my database.
Step 2: Move the data from my staging tables into a more relational set of tables that I'm using specifically for my project.
In step 1 I'm just doing a bulk SELECT and a bulk INSERT; however, in step 2 I'm doing row-by-row inserts into my tables using OLEDB Command tasks so that I can log very specific row-level activity of everything that's happening. Here is my general layout for step 2 processes.
alt text http://dl.dropbox.com/u/2468578/screenshots/step_1.png
You'll notice 3 OLEDB tasks: 1 for the actual INSERT, and 2 for success/fail INSERTs into our logging table.
The main thing I'm logging is source table/id and destination table/id for each row that passes through this flow. I'm storing this stuff in variables and adding them to the data flow using a Derived Column so that I can easily map them to the query parameters of the stored procedures.
alt text http://dl.dropbox.com/u/2468578/screenshots/step_3.png
I've decided to store these logging values in variables instead of hard-coding the values in the SqlCommand field on the task, because I'm pretty sure you CAN'T put variable expressions in that field (i.e. exec storedproc #[User::VariableName],... ,... ,...). So, this is the best solution I've found.
alt text http://dl.dropbox.com/u/2468578/screenshots/step_2.png
Is this the best solution? Probably not.
Is it good performance wise to add 4 logging columns to a data flow that consists of 500,000 records? Probably not.
Can you think of a better way?
I really don't think calling an OLEDBCommand 500,000 times is going to be performant.
If you are already going to staging tables - load it all to a staging table and take it from there in T-SQL or even another dataflow (or to a raw file and then something else depending on your complete operation). A Bulk insert is going to be hugely more efficient.
to add to Cade's answer if you truly need the logging info on a row by row basis, your best best is to leverage the oledb destination and use one or both of the following transformations to add columns to the dataflow:
Derived Column Transformation
Audit Transformation
This should be your best bet and should't add much overhead
I'm looking to execute a series of queries as part of a migration project. The scripts to be generated are produced from a tool which analyses the legacy database then produces a script to map each of the old entities to an appropriate new record. THe scripts run well for small entities but some have records in the hundreds of thousands which produce script files of around 80 MB.
What is the best way to run these scripts?
Is there some SQLCMD from the prompt which deals with larger scripts?
I could also break the scripts down into further smaller scripts but I don't want to have to execute hundreds of scripts to perform the migration.
If possible have the export tool modified to export a BULK INSERT compatible file.
Barring that, you can write a program that will parse the insert statements into something that BULK INSERT will accept.
BULK INSERT uses BCP format files which come in traditional (non-XML) or XML. Does it have to get a new identity and use it in a child and you can't get away with using SET IDENTITY INSERT ON because the database design has changed so much? If so, I think you might be better off using SSIS or similar and doing a Merge Join once the identities are assigned. You could also load the data into staging tables in SQL using SSIS or BCP and then use regular SQL (potentially within SSIS in a SQL task) with the OUTPUT INTO feature to capture the identities and use them in the children.
Just execute the script. We regularly run backup / restore scripts that are 100's Mb in size. It only takes 30 seconds or so.
If it is critical not to block your server for this amount to time, you'll have to really split it up a bit.
Also look into the -tab option of mysqldump with outputs the data using TO OUTFILE, which is more efficient and faster to load.
It sounds like this is generating a single INSERT for each row, which is really going to be pretty slow. If they are all wrapped in a transaction, too, that can be kind of slow (although the number of rows doesn't sound that big that it would cause a transaction to be nearly impossible - like if you were holding a multi-million row insert in a transaction).
You might be better off looking at ETL (DTS, SSIS, BCP or BULK INSERT FROM, or some other tool) to migrate the data instead of scripting each insert.
You could break up the script and execute it in parts (especially if currently it makes it all one big transaction), just automate the execution of the individual scripts using PowerShell or similar.
I've been looking into the "BULK INSERT" from file option but cannot see any examples of the file format. Can the file mix the row formats or does it have to always be consistent in a CSV fashion? The reason I ask is that I've got identities involved across various parent / child tables which is why inserts per row are currently being used.