Last week a question came up about adding a column to a table, and giving that column a default constraint. Would that default value be assigned to all existing rows, and how much processing would be involved.
Unsurprisingly, the answer is that – “it depends”.
I’ve got a table with about a million rows that just has an identity column and a text column I’ve populated from sys.objects:
CREATE TABLE dbo.TestAddColumn (Id int IDENTITY(1,1), TextValue sysname); INSERT INTO dbo.TestAddColumn(TextValue) SELECT a.name FROM sys.objects a, sys.objects b, sys.objects c;
Let’s add a nullable bit column and give it a default value of Zero:
ALTER TABLE dbo.TestAddColumn ADD NewFlag bit NULL CONSTRAINT DF_TestAddColumn_NewFlag DEFAULT 0;
If I look at the table I can see that the new column contains Null values:
i.e. the default value that I’ve defined hasn’t been assigned for existing rows.
I remove the column and the default constraint:
ALTER TABLE dbo.TestAddColumn DROP CONSTRAINT DF_TestAddColumn_NewFlag; ALTER TABLE dbo.TestAddColumn DROP COLUMN NewFlag;
Now let’s add the same column but we’ll disallow Null values:
ALTER TABLE dbo.TestAddColumn ADD NewFlag bit NOT NULL CONSTRAINT DF_TestAddColumn_NewFlag DEFAULT 0;
If we check the table again:
This time we can see that the default value has been assigned.
So whether our default value gets assigned to existing rows depends on whether your column is nullable or not, a nullable column will retain Null as the value. A non-nullable column will get assigned the new default value.
If you want to override that behaviour, and have your default assigned even where the column is nullable, you can use the WITH VALUES statement. First I’ll remove the constraint and column then add it again with values:
ALTER TABLE dbo.TestAddColumn ADD NewFlag bit NULL CONSTRAINT DF_TestAddColumn_NewFlag DEFAULT 0 WITH VALUES;
We look at the data again:
You can see that the value has been assigned even though the column is Nullable.
One neat thing to note, is the performance impact when carrying out these actions.
Each time I added the column I captured the execution overhead using:
SET STATISTICS IO, TIME ON;
In all cases the resource usage measured was Zero. The Add Column operation was a meta-data only operation – no data in the table needed to be updated – even where the new column was assigned a value.
This was some clever jiggery-pokery added in SQL Server 2012 .
A few years back I started running regular SQL workshops in my workplace. Teaching beginners the basics of querying databases with SQL, as well as more advanced topics for the more advanced.
During one session we were discussing the issue of knowledge acquired being quickly lost when people didn’t get the chance to regularly practice what they’d learnt. One of the attendees suggested that I should be assigning them homework.
I could see from the faces of everyone else present that the word “homework” struck an unpleasant chord. Perhaps reminding them of school days struggling to get boring bookwork done when they’d rather be at relaxation or play.
Okay, so homework maybe wasn’t going to go down well, but I figured everyone likes a good puzzle. So every Friday I started creating and sharing a puzzle to be solved using SQL. This went on for the best part of a year, then other things got in the way and gradually I stopped.
This is my invitation to you this T-SQL Tuesday. Write a blog post combining puzzles and T-SQL. There’s quite a few ways you could approach this, so hopefully no-one needs be left out for lack of ideas:
Present a puzzle to be solved in SQL and challenge your readers to solve it.
Or give us a puzzle or quiz about SQL or databases.
Show the SQL solution to a classic puzzle or game.
Provide a method for solving a classic sort of querying puzzle people face.
Show how newer features in SQL can be used to solve old puzzles in new ways.
Tell us about a time you solved a problem or overcame a technical challenge that was a real puzzle.
Or just make your own interpretation of “puzzle” and go for it!
There’s some great stuff out there already. Itzik Ben-Gan’s done a bunch of them. There’s Kenneth Fisher’s crosswords. The SQL Server Central questions of the day. Pinal Dave’s SQL Puzzles. And there’s a few on my blog too if you take a look back:
This is a blog devoted to databases, but for once I’m going to go off topic and talk about something I did at the weekend that I’ve never done before. On Easter Sunday I went to London to join in a protest demanding action on climate change.
I’d heard about what Extinction Rebellion were doing for a while. Initially I was put off by the name which I felt made them sound a bit like an anti-capitalist movement and I felt that for us to effect change we need to embrace people from all sides of the political spectrum. I also wasn’t sure how I felt about the actions they were taking – blocking roads and traffic. Was that going to antagonise more people than it gathered support?
In the end though I figured they were at least doing something – and something needed to be done. I’d lost faith that governments were going to act in time to avert disaster. I decided I wanted to show my support, add one more person to the weight of people demanding action. I didn’t want to get arrested but felt that at least being there was something. From their website I discovered there were all sorts of ways to get involved.
I went with Lisa my fiancee, and Millie our dog.
We got the train up to London from Bristol on Sunday morning and headed to the legal protest site near Marble Arch. At first it was confusing. Who could we talk to about being involved? What could we do? We found an induction session that had just started and sat down to join in.
As the leaders of the session talked about their background, their ethos, their aims and methodology I was quickly impressed. They stressed that this was not just a socialist movement, that they recognised they needed to engage people from all walks of life and ideologies.
I was most impressed by the strong focus on non-violent, non-aggressive action. That in all interactions with the police or public, those involved should make sure they were peaceful and reasonable at all times, whatever was going on. That if anyone witnessed someone against going against that guiding principle they should either intervene or find someone else to do so, to suggest to the perpetrator that this was the wrong movement for them. And they managed to make this work, in all arrests and actions there wasn’t a single report of violence or wilful destruction.
I get that some people were frustrated by their actions and the disruption it caused, and I empathise with anyone affected, but they explained that they hadn’t undertaken their actions lightly. Hundreds of thousands of people had attended climate marches to little effect. They knew they needed to do something different and things were becoming more urgent. Research was undertaken looking at movements in the past that were succesful in achieving their aims and it was dicovered that the common theme was creating disruption in a peaceful manner. Only through the threat of continued disruption were authorities forced to take notice and engage with the movements in question.
Marble Arch is a busy place, with lots of people milling around or wandering to the nearby Hyde Park. Lisa and I decided we could best involve ourselves by chatting to people passing by, handing out flyers and explaining to people what was going on and why. Talking to people about the urgency of action and some of the dangers to us all if nothing happens.
We had a lot of nice interactions, and maybe even changed a couple of minds. Millie was a big hit as a protester (it was her first protest too) and a great draw to start a conversation. In particular the police loved her! That was another great thing, to see how good natured the police were through it all, smiling, laughing and posing for pictures with protesters.
All in all it was a surprisingly nice day out and very inspiring. There were times when I felt myself getting quite emotional. It’s definitely something I’d do again.
From the media since it seems like there has been some impact, but there’s a long way to go. If you’re the slightest bit concerned about climate change – and I hope you are – I’d encourage you to get involved. Even if you just sign up on their website or add your name to petitions that circulate. Or you could make a donation, or attend one of their actions.
The pros and cons of parallelism have always been with us in SQL Server and I blogged about this a couple of years ago. This is an updated version of that post to include details of the new wait stat related to parallelism that was added in 2017 (CXCONSUMER), as well as to discuss the options available for cloud based SQL Server solutions.
There’s no doubt that parallelism in SQL is a great thing. It enables large queries to share the load across multiple processors and get the job done quicker.
However it’s important to understand that it has an overhead. There is extra effort involved in managing the separate streams of work and synchronising them back together to – for instance – present the results.
That can mean in some cases that adding more threads to a process doesn’t actually benefit us and in some cases it can slow down the overall execution.
We refer to the number of threads used in a query as the DOP (Degree of Parallelism) and in SQL Server we have the setting MAXDOP (Maximum Degree of Parallelism) which is the maximum DOP that will be used in executing a single query.
Microsoft generally recommend caution setting MAXDOP above 8:
Out of the box, MAXDOP is set to 0, which means there is no limit to the DOP for an individual query. It is almost always worth changing this to a more optimal setting for your workload.
Cost Threshold for Parallelism
This is another setting available to us in SQL Server and defines the cost level at which SQL will consider a parallel execution for a query. Out of the box this is set to 5 which is actually a pretty low number. Query costing is based on Algorithm’s from “Nick’s machine” the box used by the original developer who benchmarked queries for Microsoft.
Compared to modern servers Nick’s machine was pretty slow and as the Cost Threshold hasn’t changed for many years, it’s now generally considered too low for modern workloads/hardware. In reality we don’t want all our tiny queries to go parallel as the benefit is negligible and can even be negative, so it’s worth upping this number. Advice varies but generally recommendations say to set this somewhere in the range from 30 to 50 (and then tuning up and down based on your production workload).
There are many articles in the SQL Server community about how the out of the box setting is too low, and asking Microsoft to change it. Here’s a recent one:
Often in tuning a SQL Server instance we will look at wait stats – which tell us what queries have been waiting for when they run. CXPACKET waits are usually associated with parallelism and particularly the case where multi-threaded queries have been stuck waiting for one or more of the threads to complete – i.e. the threads are taking different lengths of time because the load hasn’t been split evenly. Brent Ozar talks about that here:
High CXPACKET waits can be – but aren’t necessarily – a problem. You can cure CXPACKET waits by simply setting MAXDOP to 1 at a server level (thus preventing parallelism) – but this isn’t necessarily the right solution. Though in some cases in can be, SharePoint for instance is best run with MAXDOP set to 1.
What you can definitely deduce from high CXPACKET waits however is that there is a lot of parallelism going on and that it is worth looking at your settings.
To make it easier to identify issues with parallelism, with SQL Server 2017 CU3 Microsoft added a second wait type related to parallelism – CXCONSUMER. This wait type was also added to SQL Server 2016 in SP2.
Waits related to parallelism are now split between CXPACKET and CXCONSUMER.
Here’s the original announcement from Microsoft regarding the change and giving more details:
In brief, moving forward CXPACKET waits are the ones you might want to worry about, and CXCONSUMER waits are generally benign, encountered as a normal part of parallel execution.
In tuning parallelism we need to think about how we want different sized queries to act on our server.
In general we don’t want these to go parallel so we up the Cost Threshold to an appropriate number to avoid this. As discussed above 30 is a good number to start with. You can also query your plan cache and look at the actual costs of queries that have been executed on your SQL Instance to get a more accurate idea of where you want to set this. Grant Fritchey has an example of how to do that here:
Often the answer is going to be simply to set it to 8 – but then experiment with tuning it up and down slightly to see whether that makes things better or worse.
Very Large Queries
If we have a mixed workload on our server which includes some very expensive queries – possibly for reporting purposes – then we may want to look at upping the MAXDOP for these queries to allow them to take advantage of more processors. One thing to consider though is – do we really want these queries running during the day when things are busy? Ideally they should run in quieter times. If they must run during the day, then do we want to avoid them taking over all the server power and blocking our production workload? In which case we might just let them run at the MAXDOP defined above.
If we decide we want to let them have the extra power then we can override the server MAXDOP setting with a query hint OPTION(MAXDOP n):
You will want to experiment to find the “best” value for the given query. As discussed above and as shown in Kendra Little’s article, just setting it to the maximum number of cores available isn’t necessarily going to be the fastest option.
Exceptions to the Rule
Regardless of the size, there are some queries that just don’t benefit from parallelism so you may need to assess them on an individual basis to find the right degree of parallelism to use.
With SQL server you can specify the MAXDOP at the server level, but also override it at the database level using a SCOPED CONFIGURATION or for individual queries using a query hint. There are even other ways you can control this:
If your SQL Server is hosted in the cloud, then most of this still applies. You still need to think about tuning parallelism – it isn’t done for you, and the defaults are the same – so probably not optimal for most workloads.
There are in general two flavours of cloud implementation. The first is Infrastructure as a Service (IaaS) where you simply have a VM provided by your cloud provider and run an OS with SQL server on top of it in that VM. Regardless of your cloud provider (e.g. Azure, AWS etc.), if you’re using IaaS for SQL Server then the same rules apply, and you go about tuning parallelism in exactly the same way.
The other type of cloud approach is Platform as a Service (PaaS). This is where you use a managed service for SQL Server. This would include Azure SQL Database, Azure SQL Database Managed Instance, and Amazon RDS for SQL Server. In these cases, the rules still apply, but how you manage these settings may differ. Let’s look at that for the three PaaS options mentioned above.
Azure SQL Database
This is a single SQL Server database hosted in Azure. You don’t have access to server level settings, so you can’t change MAXDOP or the cost threshold. You can however specify MAXDOP at the database level e.g.
ALTER DATABASE SCOPED CONFIGURATION SET MAXDOP = 4;
Cost threshold for Parallelism however is unavailable to change in Azure SQL Database.
Azure SQL Database Managed Instance
This presents you with something that looks very much like the SQL Server you are used to, you just can’t access the box behind it. And similar to your regular SQL instance, you can set MAXDOP and the Cost threshold as normal.
Amazon RDS for SQL Server
This is similar to managed instance. It looks and acts like SQL Server but you can’t access the machine or OS. You access your RDS instance through an account that has permissions that are more limited than your usual sa account or sysadmin role allows. And one of the things you can’t do with your limited permissions is to change the parallelism settings.
Amazon have provided a way around this though and you can change both settings using something called a parameter group:
Parallelism is a powerful tool at our disposal, but like all tools it should be used wisely and not thrown at every query to its maximum – and this is often what happens with the out of the box settings on SQL Server. Tuning parallelism is not a knee-jerk reaction to high CXPACKET waits, but something we should be considering carefully in all our SQL Server implementations.
I wanted to update my original article to include the cloud options noted above, but didn’t have access to an Azure SQL Database Managed Instance to check the state of play. Thanks to TravisGarland via Twitter (@RockyTopDBA) and Chrissy LeMaire via the SQL community slack (@cl) for checking this and letting me know!
When you drop a database from a SQL Server instance the underlying files are usually removed. This doesn’t happen however if you set the database to be offline first, or if you detach the database rather than dropping it.
The scenario with offline databases is the one that occurs most often in practice. I might ask if a database is no longer in use and whether I can remove it. A common response is that people don’t think it’s in use, but can I take it offline and we’ll see if anyone screams. I’ll often put a note in my calendar to remove it after a few weeks if no-one has complained. When I do come to remove it, hopefully I’ll remember to put it back online before I drop it so the files get removed, but sometimes I might forget, and in an environment where many people have permissions to create and drop databases you can end up with a lot of files left behind for databases that no longer exist – these are what I’m referring to as orphaned files.
Obviously this shouldn’t happen in production environments where change should be carefully controlled, but if you manage a lot of development and test environments this can certainly occur.
So I created a script I can run on an instance to identify any files in its default data and log directories that are not related to any databases on the instance. Here it is:
UPDATE #Files SET Directory = @DefaultLogPath, FullFilePath = @DefaultLogPath + [FileName] WHERE Directory IS NULL;
SELECT f.[FileName], f.Directory, f.FullFilePath FROM #Files f LEFT JOIN sys.master_files mf ON f.FullFilePath = REPLACE(mf.physical_name,'\\', '\') WHERE mf.physical_name IS NULL AND f.FileFlag = 1 ORDER BY f.[FileName], f.Directory
DROP TABLE #Files;
I wouldn’t say that you can just go delete these once you’ve identified them, but at least now you have a list and can investigate further.
By the way, you might notice a nasty join statement in the above query. This is to deal with instances where the default directories have been defined with a double backslash at the end. SQL Server setup allows this and it doesn’t cause any day-to-day problems, but can make this sort of automation challenging. I’ve included it in this query as I’ve encountered a few people having this issue. In general I’d avoid such joins like the plague.
Making things more complicated
One complication can be where you have multiple SQL Server instances on the same server. This isn’t greatly recommended in production, but is common in dev\test. Where a database has been migrated from one instance to another, it’s possible that hasn’t been done correctly and the files still exist under the directories for the old instance and you might then see them as orphaned. I previously posted a script to identify such files:
Combining these three techniques makes it relatively easy to identify files that are probably no longer needed. You can get a list of all files that don’t belong to databases on the instances they live under, correlate that to any files that are down the wrong path for any of your instances, then look at what’s left over.
When I’ve created resources in Azure it’s usually taken from a few minutes and up to quarter of an hour – though sometimes longer.
When you’re new to this stuff, you can be uncertain and wonder, “Is it really creating it?”, “Did I hit the right buttons?”. As a result it can be handy to know where to check to see what’s going on.
Sometimes after creating the resource you are taken to a screen that will show you what’s going on:
And usually you can see something is occurring from the bar at the top:
If you click on the alarm icon you can see more details:
You can then click to see “More events in the activity log” to dig deeper:
This is all fairly intuitive, but earlier I was trying to create a SQL Database Managed Instance for the first time. It showed some activity in the items above for a few minutes, but after that nothing happened. Had it failed? Had I done something wrong? Should I start again and try to create a new one?
The answer was to select resource groups from the blades on the left, and select the resource group that I had created the item in:
On the right hand side I can see an item saying “Deployments” and I can see that one is in the process of deploying. I can click the hyperlink for more details:
The third item in the list was the one I was looking for:
Okay, so it is in the process of being created. There’s no way to tell how long it will take but at least I now know it’s happening.
While searching for it I did notice a warning on the create screen for the resource that I hadn’t seen when I first whizzed through the creation:
The SEQUENCE object was added to T-SQL in SQL Server 2012. It’s reasonably well known to DBAs, but less so to developers or those new to SQL, so I thought I’d produce a quick post to demonstrate its use.
In basic terms, a SEQUENCE is a way of generating a sequence of numerical values. The following are examples of sequences you could generate:
1, 2, 3, 4, 5 6, 7, 8, 9…
1, 6 , 11, 16, 21, 26, 31…
1000, 1001, 1002, 1003, 1004…
You can pick a starting number (and an ending number if you want), a data type (which might produce a natural maximum to the size of the sequence) and an increment (i.e. how much you want to be added to produce the next number in the sequence). There are other options, but I’m going to focus on the simplest use case. You can find the full documentation here:
So, let’s define my use case. I have a table to hold customer orders. For each record want to define an Order Reference Number of the format ORN0000000001.
Now you could implement something using an IDENTITY column to manage this – but there may be times when that is not ideal, for instance your table may not already have a suitable identity to use (you might have a unique identifier as the primary key) and if you want to store the actual reference then you’d need to add an IDENTITY column in addition to the reference column. Or you might need a reference that is unique across multiple tables.
The SEQUENCE object is also designed to be faster than IDENTITY, creating less blocking when you have a lot of concurrent inserts.
First of all, creating the sequence to generate the numeric part of my reference is easy. Let’s say that a bunch of reference numbers have already been used so I want to start with ORN0000100001
Let’s look at the SQL…
CREATE SEQUENCE dbo.OrderRefSequence AS bigint START WITH 100001 INCREMENT BY 1;
Then I can request numbers from the sequence using NEXT VALUE FOR e.g.
SELECT NEXT VALUE FOR dbo.OrderRefSequence;
The first time I run that I get the starting number 100,001.
Another nice addition to SQL Server 2012 was the FORMAT function which we can use to format the number into a string whilst padding it with leading zeroes and adding the text prefix:
SELECT FORMAT(NEXT VALUE FOR dbo.OrderRefSequence, 'ORN0000000000#');
That returns me ORN00000100002.
If I keep executing it then the reference increases:
So, now I can just use that when inserting values to my table to get a new reference number each time.
But, what’s even nicer is that you can do it all by defining a default for your column and referencing the sequence in the default.
I’ll create the following table:
CREATE TABLE dbo.Orders ( Id UNIQUEIDENTIFIER PRIMARY KEY DEFAULT NEWSEQUENTIALID(), CustomerId UNIQUEIDENTIFIER NOT NULL, OrderReference VARCHAR(20) DEFAULT(FORMAT(NEXT VALUE FOR dbo.OrderRefSequence, 'ORN0000000000#')), OrderDate DATETIME DEFAULT(GETUTCDATE()));
You can see that the OrderReference is defined with a default using our sequence object.
I insert a bunch of rows into the table. For the sake of this rather contrived example, I only need to specify the CustomerId. I do that by generating a bunch of random unique identifiers – one for each row in the sys.objects table.
INSERT INTO dbo.Orders (CustomerId) SELECT NEWID() FROM sys.objects;
Let’s have a look at an extract from the table:
You can see I’ve got a nice series of ascending, non-duplicating reference numbers.
One thing to note is that, while the sequence will generally produce unique number, it is still worth enforcing that in your table definition with a unique constraint i.e.
ALTER TABLE dbo.Orders ADD CONSTRAINT UQ_Orders_OrderReference UNIQUE(OrderReference);
This prevents someone from issuing an UPDATE command that might create a duplicate reference. Also, once the sequence runs out of numbers it will repeat back at the beginning unless you specify NO CYCLE in the defintion of the sequence – obviously in most applications this is unlikely to be an issue if you’re using a bigint for the sequence.
There was also a bug in some versions of SQL 2012 and 2014 that meant a duplicate could get created when your server was under memory pressure:
This was fixed with SQL Server 2012 SP2 CU4 and SQL Server 2014 CU6 – but it’s better to be safe than sorry.
As a final note, it’s worth remembering that with the GDPR, these sorts of references are defined as personal data.That’s one good reason not to ever consider using these sorts of references as the primary key of your table (there are many others) – but also a reason why – where you already have an identity based primary keys that you could use to generate the references – it may be worth decoupling them from the primary key and basing them on a separate sequence instead.
Like Andy, I didn’t take a straight route into being a DBA. I actually trained as a teacher, and did a variety of temp jobs – largely in call centres – before I got my first “proper” job as a developer.
My first steps into development were when I was a team leader in a call centre. There was a lot of paper shuffling going on. I felt there had to be a better way of doing things, so developed processes using Access and Excel and convinced the Admin team in the benefits of using them.
Not exactly development, more Macros with a bit of VBA, but that got me spotted by another team that were trying to automate reporting in the call centre – again using MS Access. That was where I first encountered relational databases and got deeper into coding. I joined as they were in the process of taking an Excel spreadsheet that processed flat files using macros – taking the whole day/every day to chunder away, replacing it with a coded solution. Finally we had it importing all the call center stats in seconds on a schedule in the early hours.
Nowadays that seems trivial, but back then it felt like a great achievement. It’s still one of my best work memories – when I showed my boss the automated import for the first time.
Moving forward from there and trying to answer Andy’s question, I realised that what started me off is what’s kept moving me forward. That there’s always got to be a better way.
Another of my favourite early developer memories was being asked to look at whether a file import to an application could be sped up as it was taking 10 minutes. The process was to upload a file with corporate credit card statements from a flat file into a SQL database for the application that companies used to manage them. When I opened it up I was quite impressed with the code, it seemed cleverer than anything I could have written – but maybe a bit too clever. I asked myself if there was a better and simpler way of achieving the same thing. There was, and it turned out to be a lot lot quicker.
Making things run faster remained something I loved doing, but it often seemed often that optimisation wasn’t a big focus in application coding – that was more about delivering features. I think that’s part of what led me to specialise in data where performance, particularly at scale, is so key. Looking at SQL code and going – “well, sure it works as it is, but is there not a better way of doing this.”
It strikes me that for a lot of us, this is how we ended up in software, we saw things being done a certain way and saw scope for improvement. Felt that there must be a better way.
And it’s good to remember this isn’t an attitude where we can rest on our laurels. Even when we know what we’re doing, have done something a hundred times before, it’s still good to say “There’s got to be a better way”, and to think about what that might be. Not only does that mean we’re continually improving, but it’s also what keeps working in technology fresh and interesting.
I’ve recently encountered an issue that was difficult to resolve and I didn’t find the particular cause that was troubling us documented elsewhere on the web so thought I’d record it here.
The issue was with a service account connecting to SQL Server and intermittently failing to logon.
Errors reported in the Windows Application Event log were:
SSPI handshake failed with error code 0x8009030c
Login failed. The login is from an untrusted domain and cannot be used with Windows authentication.
The login attempt didn’t appear to get as far as the SQL instance, so no further information could be captured in a failed Logins trace.
This was affecting a large number of application servers using the same service account. Fortunately this was in development and test environments so no production impact.
The problem was that the account was getting locked out. A service was running every half hour using the account to connect to SQL, but with the wrong password. We also had a process running to unlock locked service accounts – so the account would start working again after a few minutes.
The resolution was to kill that service as it was no longer required. We were able to identify where the failed logins were coming from via the Active Directory audit logs for the account in question.
This was particularly difficult to troubleshoot as the error was a bit misleading.