gMSA and Docker – Lessons Learnt

In the last two posts (here and here) I have documented how I use gMSAs to connect services running in docker containers on Windows to SQL Server using the domain authentication. In the end it was very simple, but there are things I wish I knew when I started. It would save me a lot of time. Here is an attempt to document the lessons learnt.

Versions are important!

While in the end I was able to make it work on Windows Server 2016, 1803, 2019 and 1809 I wasted some time trying to make it work with docker 17.06. Unsuccessfully. Docker 18.09.1 and 18.09.2 worked every time.

There are some reports of intermittent problems with specific OS updates breaking stuff, like the one here but I wasn’t able to reproduce it. I wonder if the updates changes something else that it causing problems, in other words is it the problem with the update itself or the update process?

The Set-AdServiceAccount

From the beginning I set to try the gMSA authentication on multiple VMs following blog posts which all included some use of the Set-AdServiceAccount powershell command (from `RSAT-AD-PowerShell). I could not make it work on more than one VM at a time. I thought I was going mad! The problem (and the clue) is in the name. Set. It is not add, not modify. So when I was doing something like this

Set-AdServiceAccount -Identity MyService `
  -PrincipalsAllowedToRetrieveManagedPasswords DH2019A$ 

it was setting the principals allowed to retrieve the managed passwords for MyService to DH2019A VM. As expected. But not as expected removing the privilege from all the other VMs I granted that permission before. With no warnings.

It is probably the most worrying part about using gMSAs for the service authentication in production, as I plan to do it. All it will take is one sysadmin to run a command like that to break all the services, potentially on all docker hosts. To mitigate we have decided to grant the permissions through a domain group to which we will add docker hosts. That way there should be no need to run this command when scaling out.

Misleading Get- and Test-AdServiceAccount

Understanding the above problem with Set-AdServiceAccount was made much worse, by my misunderstanding how Get-AdServiceAccount and `Test-AdServiceAccount work.

If you are a domain admin the Get-AdServiceAccount will always return details of the gMSA if it exists. So it is of no use to check if the specific gMSA can be used on a given host.

If the gMSA was previously installed the Test-AdServiceAccount will return true regardless if the host account has permissions to retrieve the password or not. That permission is necessary for the gMSA authentication to work.

So with that in mind neither command is fit for checking if a specific host has permissions it needs to use a gMSA. I was not able to find anything better than attempting to install it again with Install-AdServiceAccount. It will either install it again, or display error message indicating that the computer has no permissions to retrieve the password.

Remove-AdServiceAccount

This does not remove previously installed gMSA from the local host. It removes the gMSA from the domain!

The SSPI context error.

If you try to use domain authentication from the service running on a docker host which has no permissions to retrieve the gMSA password you will get fairly generic error tell you that the SSPI context couldn’t be created.

There are scores of blog posts and msdn documents explaining how to troubleshoot many SSPI context errors. Not a single one I found mentions any problems with gMSA. I have learnt a lot about SSPI and how it really works, just to eventually realise that everything is fine, and I have to look for problems somewhere else.

There is no localhost

When you run a standalone container you can access it from the same host on the localhost. By default a nat network is used and it allows communication on the hosts IP. When moving to the swarm mode (using docker service create not just docker create) by default the ingress network is used and the localhost is not available. You have to use the public IP address of the docker swarm. There is a lot of blog posts how to define your own overlay or bridge networks. I suppose they all work on Linux, but on windows when using an overlay network you cannot use host IPs (so no localhost) and you cannot create bridge networks at all.

PS C:\> docker network create -d bridge bnet
Error response from daemon: could not find plugin bridge in v1 plugin registry: plugin not found

I was able to make it work when publishing the port directly on the host using --publish published=8001,target=80,mode=host instead of the shorter -p 8001:80 (ports obviously may be different) but I don’t think that’s a configuration I’d be using so… I simply gave up. I use public IP and everything works.

Security considerations

To install a gMSA on a host which has permission to read the gMSA’s password you don’t need to have any extra permissions. It appears that anybody with access to PowerShell on the host can do Install-AdServiceAccount -Identity MyService. There are no restrictions which credential spec file can be sued for which servce either. This means that if somebody has access to the docker host they can create a new service using any gMSA to which the host itself has permissions.

gMSA name lenght limit

The Group Managed Service Account’s name is limited to 15 characters. Not a big deal, but it messed up a carefully agreed naming strategy and in the end I have vowel-less service names.

New-CredentialSpec silently overwrites existing files.

Active Directory, Windows Containers in Swarm Mode and SQL Server

In my previous post I have explained how I was able to connect from windows containers running on docker to a SQL Server cluster on a network using domain authentication (with gMSAs) rather than SA logins and passwords.

gMSAs in docker swarm mode

After I got the containers using Group Managed Service Accounts working on a single Docker host I went on to try the same in the swarm mode. My plan was to simply replace docker run -d part of the command creating the container with docker service create, but it turns out that it is not that simple, especially if you don’t have a lot of experience with swarm mode. It is also worth noting, that I had a lower success rate than when I was experimenting with standalone containers. I was able to make it work on all the same Windows versions (2016, 1803, 2019, 1809) but only when using Docker 18.09 and not on the 17.06 which was on the image I used for 1803 tests).

Demo setup

Similarily to the previous post I have tested this on a range of operating systems and docker versions, but what I want to show here is how it worked on Windows 2019 and Docker 18.09.

gMSA_Docker_Service_1

To make it a bit more exciting (and because of how Docker Swarm works) this time I will be testing the service from a web browser. To start with it does’t work. That is because there is no service listening on port 8101.

gMSA_Docker_Service_1_web

Creating a service

To create a service docker service create command is used. When compared to docker create some parameters are different, for example there is no --security-opt used in the previous post and instead --credential-spec is used. But first things first. Let’s just create a service using michalporeba/sqlgmsatest:1809nano image with minimal configuration and see what happens.

docker service create -p 8101:80 michalporeba/sqlgmsatest:1809nano

gMSA_Docker_Service_2

The -p 8101:80 makes the service available on the 8101 port using the default ingress network. No errors, the service is running, it is converged, so let’s try to connect to it!

gMSA_Docker_Service_2_WebRequest

And here is the first surprise. The localhost doesn’t work. That’s a swarm thing and although it is possible to publish ports in host mode it is not how I would be running in production, so I will just open the ports and connect to the service externally using a web browser.

gMSA_Docker_Service_2_web

OK, so the api/info call was successful. The service from michalporeba/sqlgmsatest:1809nano image is running and responding. So the next task is to use it to query theTestDBdatabase on my test instanceDB.sqlgmsa.local`.

gMSA_Docker_Service_2_web2

Adding the gMSA

Not authorized! But who? The NT AUTHORITY\ANONYMOUS LOGON. That is because despite the docker host being member of the sqlgmsa.local domain, the container running the service is not. To fix it, exactly as in the case of standalone container a Group Managed Service Account has to be created, installed and a credential file created.

# Create gMSA
New-AdServiceAccount -Name MyService -DNSHostName sqlgmsa.local `
  -PrincipalsAllowedToRetrieveManagedPassword "Domain Controllers", "Domain Admins", "CN=DockerHosts,CN=Computers,DC=sqlgmsa,DC=local" `
  -KerberosEncryptionType AES128, AES256

# Install it
Install-AdServiceAccount -Identity MyService

# Import the module to manage Credential Specs
Import-Module .\PsModules\CredentialSpec.psm1

# And create a spec file for MyService
New-CredentialSpec -Name MyService -AccountName MyService `
-Domain (Get-AdDomain -Current LocalComputer)

gMSA_Docker_Service_3

Consistency is everything, isn’t it? -Name, -Identity, -AccountName on those commands above refer to the same thing, the gMSA name and have to match. The -Name parameter on the New-CredentialSpec command is used to control the name of the json file containing the credential spec. The filename can be anything, and it doesn’t need to match the account name, but I find it easier if it does. The existing credential spec files can be found in C:\ProgramData\docker\CredentialSpecs\ or by using Get-CredentialSpec command from the CredentialSpec.psm1 module.

The next step is to use the newly created credential spec file when creating the service. The --security-opt is not supported when created a service and --credential-spec has to be used instead.

docker service create -p 8102:80 `
  --credential-spec file://MyService.json `
  michalporeba/sqlgmsatest:1809nano

gMSA_Docker_Service_4

The new service now runs on port 8102 and should use the new MyService identity. Let’s see.

gMSA_Docker_Service_4_web

Almost there! Login failed for user SQLGMSA\MyService$ That’s good, that means the correct identity has been picked up, so the last thing to do is to create the login on the SQL Server.

gMSA_Docker_Service_5_sql

And now, as if by magic

gMSA_Docker_Service_5_web

The test web service, written in C#, using .net core is hosted in a Docker container running on windows host,and queries a SQL Server database using domain authentication.

Active Directory, Windows Containers and SQL Server

The problem

Everything appears to be in containers nowadays, even the SQL Server. But still there are mixed environments, people and companies wanting to try containers without going all in. So I was wondering how practical would it be to have .net core services on docker, running in Windows containers connecting to an external, old fashioned SQL Server instance? Also, as it is all in a Windows domain, I’d like to use domain authentication so I don’t have to worry about managing passwords.

Simple isn’t it? Well, it turns not that simple as not everything is on the domain. The SQL Server is, the docker hosts are, but the containers are not.

Additionally there are differences depending on whether you want to run as an independent container, or in docker swarm mode. This blog post focuses on standalone containers, and the swarm mode is covered in the follow-up post.

The quick answer

The good news is that it is not an unreasonable requirement and it has been done before. The solution is to use Group Managed Service Accounts (gMSA) and Credential Spec Files. A number of people have already documented their efforts. Some were more successful than others.

My story

My problem was that I wasn’t able to make it work just by following any single write-up. In fact, for a few days, I was not able to get it going at all. But eventually it happened and here is a step by step description of how I made it work on Windows Server 2016, 1803, 1809 and 2019 as the host OS and 2016, 1803 and 1809 in full and nano options as the container base image. Generally, it is very simple once you know what to do, and more importantly what not to do (more about it later).

Test setup

To test it I have set up a virtual lab environment on Azure with 6 VMs

  • DC – Windows Server 2016 Datacenter acting as a domain controller
  • DB – Windows Server 2016 Datacenter with SQL Server 2017 Developer edition installed
  • DH2016A – Windows Server 2016 Datacenter with Containers (Docker version 18.09.2)
  • DH1803A – Windows Server 1803 with Containers (Docker version 17.06.2-ee-18)
  • DH1809A – Windows Server 1809 with Containers (Docker version 18.09.0)
  • DH2019A – Windows Server 2019 with Containers (Docker version 18.09.1)

I have created a sqlgmsa.local domain and joined all the VMs to it. SQL Server was run using SQLGMSA\SqlServer Managed Service Account without any special permissions.

In the domain I have 2 service accounts SQLGMSA\ServiceA and SQLGMSA\ServiceB. Both have logins on the SQL Server instance. I will be setting some of my containers to connect to the SQL Server as ServiceA and some as ServiceB.

Initially I tested the connectivity from containers build with full base images (standard mrc.microsoft.com/windows/servercore) and using dbatools module to run queries from them to the DB.sqlgmsa.local server. To be able to test nano based images I created two test images containing simple .net core WebAPI service written in C# with two public methods. Calling api/info you can check if the service is running, what system is it running on. Calling api/query/ attempts to open connection to the specified database (or master if the db parameter is not provided) and returns information about the database, the original login and the current user. A simple query

select 
     @@version SqlServer
    ,db_name()  [Database]
    ,current_user CurrentUser
    ,original_login() OriginalLogin
for json path

The test images are available on docker hub and the source code here on github.

PowerShell and AD

In the example I am using PowerShell to manage my active directory. If the commands I use don’t work for you you may be missing the AD modules. To install them add the RSAT-AD-PowerShell windows feature by executing this PowerShell command

Add-WindowsFeature RSAT-AD-PowerShell

Group Managed Service Accounts (gMSA)

Managed Service Accounts where introduced some time ago to reduce overhead associated with managing passwords for service accounts. The Group Managed Service Accounts solve the same problem but unlike MSAs gMSAs can be used across multiple computers.

To start using gMSAs on a domain a KDS Root Key has to be created first. It is the key with which passwords shared between the computers on the domain are protected. If your domain has other MSAs already you will not need to do it again.

To create a KDS Root Key Run I used this PowerShell command on the domain controller

Add-KdsRootKey -EffectiveTime (Get-Date).AddHours(-10)

and then to verify that it has been created

Get-KdsRootKey

Now to create the test service accounts I used the following commands. I was doing it on the DC, but with the right permissions it should be possible to do from any computer on the domain.

New-AdServiceAccount -Name ServiceA -DNSHostName sqlgmsa.local `
   -PrincipalsAllowedToRetrieveManagedPassword "Domain Controllers", "Domain Admins", "CN=DockerHosts,CN=Computers,DC=sqlgmsa,DC=local" `
   -KerberosEncryptionType AES128, AES256

New-AdServiceAccount -Name ServiceB -DNSHostName sqlgmsa.local `
   -PrincipalsAllowedToRetrieveManagedPassword "Domain Controllers", "Domain Admins", "CN=DockerHosts,CN=Computers,DC=sqlgmsa,DC=local" `
   -KerberosEncryptionType AES128, AES256

Where ServiceA and ServiceB are the names of the accounts and "CN=DockerHosts,CN=Computers,DC=sqlgmsa,DC=local" is the distinguished name of the group I have created for the docker hosts.

If you don’t know the exact distinguished name running this command can help

Get-AdGroup -filter { name -like "yourgroupname" }

Now on every docker host all the specific service accounts (2 in my test case) have to be installed so that the host OS can access them.

Install-AdServiceAccount -Identity ServiceA
Install-AdServiceAccount -Identity ServiceB

If there is an error message like this, it means the permissions were not set correctly
Install-AdServiceAccount : Cannot install service account. Error Message: ‘{Access Denied}

Credential Spec file

Docker Credential Spec Files have been created specifically to solve the problem of passing gMSA to containers. They are plain json files with information about the service account. It is possible to create the files manually, but there is module for it. It is documented here but here is a short instruction of how to create get and import the module.

This part needs to be done on every docker host.

# 1. Set TLS1.2 support from PowerShell so the module can be downloaded from github. 
PS C:\Tmp> [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12

# 2. Download the psm1 file using Invoke-WebRequest
PS C:\Tmp> Invoke-WebRequest "https://raw.githubusercontent.com/MicrosoftDocs/Virtualization-Documentation/live/windows-server-container-tools/ServiceAccounts/CredentialSpec.psm1" -OutFile "CredentialSpec.psm1"

# 3. Import the module
PS C:\Tmp> Import-Module .\CredentialSpec.psm1

With CredentialSpec module imported for each gMSA a credential spec file has to be created.

PS C:\> New-CredentialSpec -Name ServiceA`
  -AccountName ServiceA `
  -Domain $(Get-AdDomain -Current LocalComputer)

PS C:\> New-CredentialSpec -Name ServiceB`
  -AccountName ServiceA `
  -Domain $(Get-AdDomain -Current LocalComputer)

The list of existing files can be obtained with

PS C:\> Get-CredentialSpec

Name     Path
----     ----
ServiceA C:\ProgramData\docker\CredentialSpecs\ServiceA.json
ServiceB C:\ProgramData\docker\CredentialSpecs\ServiceB.json

And finally, run the containers passing the credential spec files with the --security-opt parameter. (This is an example from DH2019A using the 1809 nano base image).

docker run -d -it -p 8001:80 `
   --security-opt "credentialspec=file://ServiceA.json" `
   --name ServiceA `
   michalporeba/sqlgmsatest:1809nano

docker run -d -it -p 8002:80 `
   --security-opt "credentialspec=file://ServiceB.json" `
   --name ServiceB `
   michalporeba/sqlgmsatest:1809nano

The proof is in the pudding

After checking both containers are running with docker ps I can start testing. As the test is not focused on anything else but domain authentication I didn’t open any ports to the lab, and all I was doing was to either connect to the container and use dbatools to execute a query on the db server, or from the docker host connecting to the service listening on the published port. Here are the example calls using Invoke-WebRequest on DH2019A.

sqlgmsa.proof

PS C:\> $env:ComputerName
DH2019A
PS C:\> docker start ServiceA
ServiceA
PS C:\> docker start ServiceB
ServiceB
PS C:\> docker ps
CONTAINER ID        IMAGE                               COMMAND                  CREATED             STATUS              PORTS                           NAMES
c6382bb7d816        michalporeba/sqlgmsatest:1809nano   "dotnet TestService.…"   2 days ago          Up 4 seconds        443/tcp, 0.0.0.0:8002->80/tcp   ServiceB
02ece189cb74        michalporeba/sqlgmsatest:1809nano   "dotnet TestService.…"   2 days ago          Up 8 seconds        443/tcp, 0.0.0.0:8001->80/tcp   ServiceA
PS C:\> # Service A
PS C:\> (Invoke-WebRequest -UseBasicParsing http://localhost:8001/api/info).Content
["OS:  Microsoft Windows 10.0.17763 ","Framework: .NET Core 4.6.27317.07"]
PS C:\> (Invoke-WebRequest -UseBasicParsing http://localhost:8001/api/query/DB.sqlgmsa.local).Content
[{"Database":"master","CurrentUser":"guest","OriginalLogin":"SQLGMSA\\ServiceA$"}]
PS C:\> # Service B
PS C:\> (Invoke-WebRequest -UseBasicParsing http://localhost:8002/api/info).Content
["OS:  Microsoft Windows 10.0.17763 ","Framework: .NET Core 4.6.27317.07"]
PS C:\> (Invoke-WebRequest -UseBasicParsing http://localhost:8002/api/query/DB.sqlgmsa.local).Content
[{"Database":"master","CurrentUser":"guest","OriginalLogin":"SQLGMSA\\ServiceB$"}]
PS C:\>

Conclusions

The above setup really boils down to 5 steps. If you want to use windows authentication from windows containers on docker to a SQL Server instance (or a cluster you have to

  1. Create gMSAs for your services
  2. Create logins for the service accounts on the SQL Server
  3. Install gMSAs on the docker hosts
  4. Create credential spec files
  5. Create containers with --security-opt parameter pointing to the credential spec file.

That’s it. In my case I was able to make it work (for standalone containers, not in swarm mode) on different OS versions (2016, 1803, 1809, 2019) using full and nano base images and using docker 17.06 and 18.09. However, there can be surprises and pulling hair. In the week I spent trying to figure it out I had a number of moments when I thought I’ve got it, just to realise that what worked a moment ago, doesn’t any more.

The biggest lessons where

  • use AD Groups for managing access to gMSAs rather than individual computer accounts,
  • be very careful with Set-AdServiceAccount which I have seen in some of the posts out there,
  • SSPI context errors is not what it seems, and can be very annoying

More details about the lessons learnt can be found in the follow up post.

Trying to do the same but in a service run on Docker in swarm mode is similar, but not exactly the same. I have described it here.

This post is long enough as it is, so I will not go into the details of those lessons learnt here but instead include them in a follow up to which I will link here later.

Backing SQL on Linux to Windows Share

How to back up a database from SQL Server on Linux (perhaps in a Docker container) to a Windows Share already on the network?

If you want to know how to run SQL Server on Linux in a Docker container read my earlier post.

Container run-time privileges and Linux capabilities

Before we start it is important to note that by default the containers are run with very limited privileges and cannot do much. That’s by design, to make them more secure. However, that prevents them from being able to use CIFS protocol to mount to an SMB share, exactly what we are trying to do. If you just follow the steps in the next sections when you try to mount the share you will get

Unable to apply new capability set.

To avoid this problem the container needs to be created with two extra capabilities: SYS_ADMIN and DAC_READ_SEARCH. You can read more about it on the here. But to make the long story short you need to add two --cap-add parameters to your docker run command.

The full command from my earlier post becomes:

docker run -d `
   -e ACCEPT_EULA=Y -e SA_PASSWORD=Secret1234 `
   -p 14333:1433 --name sql1 `
   --cap-add SYS_ADMIN --cap-add DAC_READ_SEARCH ` 
   mcr.microsoft.com/mssql/server:latest

Create a container with those capabilities before continuing.

The Naive Approach

Let’s assume you have a network share already available and it is accessible using UNC \\FileShare1\SqlBackups\. Being used to Windows networking one would expect to simply take a backup like so, for example:

backup database [TestDb] 
   to disk='\\FileShare1\SqlBackups\TestDB.bak'

The command completes with no errors and yet there is no TestDB.bak file in \\FileShare1\SqlBackups. Stranger still it is possible to restore from that file which doesn’t seem to exist. Try:

restore filelistonly 
   from disk='\\FileShare1\SqlBackups\TestDB.bak'

It works, because Linux translates the UNC it knows nothing about (after all it’s a Windows thing) to a local file /FileShare/SqlBackups/TestDB.bak. If you remote to the container (or execute bash on it with docker exec -it sql bash) you will find /FileShare1/SqlBackups/TestDB.bak file.

Interesting, but not what we expected.

Getting into SMB and CIFS

To solve this problem a network share needs to be mounted to a node in the Linux file system. There are two ways to do it, one is temporary using the mount command only, or a more permanent involving editing the fstab file and then using mount. But before we can do it, one problem has to be solved first. The Windows file server shares the folder using the SMB protocol which is gibberish to Linux. Luckily Linux can be taught to speak CIFS protocol which is compatible with SMB. The way to do it is to install the cifs-utils package. On Ubuntu (SQL Server vNext in a container runs on Ubuntu so I use it as an example) is to install the package with apt-get. Other flavours of the OS will use their own package managers, but overall the process is the same.

First, you will have to Connect to SQL Server on Linux.

Then the CIFS protocol utilities have to be installed using apt-get on Ubuntu.

apt-get update
apt-get install cifs-utils

You will be asked if you want to really do it as it will take an extra 41 MB of disk space. Press Y to agree.

After this operation, 41.2 MB of additional disk space will be used.
Do you want to continue? [Y/n]

Now a directory where the backups will be mounted needs to be created. Typically the external mounts are all in /mnt. So let’s create a backups directory there.

mkdir /mnt/backups

Now it is time to mount.

mount -t cifs //FileShare/SqlBackups /mnt/backups \
   -o username=yourusername,domain=yourdomain,file_mode=0777,dir_mode=0777,rw,sec=ntlm

You will be asked for a password and if everything is correct you will get prompt again which means everything went well. It is important to note the change from \ to /.

It should not be possible to navigate to that directory and create a test text file.

cd /mnt/backups
touch test.txt

With that, the test.txt file should be now visible on \\FileShare1\Backups.

And to take the backup to that share becomes no different to taking any other backup to disk:

backup database [TestDb] 
   to disk='/mnt/backups/testdb.bak'

Using Shares to Interact with the Host

It is also possible to use this approach instead of using -v or --mount to share the file system between a container and its host. It is arguably a bit more effort but it allows to easily share backup and restore location between multiple containers without read/write issues associated with folder binding.

Connect to SQL Server on Linux

How to connect to SQL Server on Linux?
It really depends on what one means by ‘connect’.

Query SQL Server on Linux from Windows

To connect to the SQL Server instance running on Linux directly or in a container to run a query it is no different from connecting to one running on Windows. Just connect using the DNS name or an IP and a port number if it is not the default 1433. The only difference is that only SQL Authentication is supported so you will not be able to use Windows Authentication and your domain credentials.

Query SQL Server from Linux

Azure Data Studio (formerly known as SQL Server Operations Studio) can run on Linux offering the same user experience as on Windows. It is also possible to use the sqlcmd which works exactly the same as the one on windows. To install it on Ubuntu or any Linux flavour using apt-get package manager you can simply do

apt-get install mssql-tools

End then just use it the same way as you would normally do

sqlcmd -S localhost -U sa -P mysecretpassword

In the default SQL Server Docker image the mssql-tools are already installer but they are not added to the $PATH variable. You can either add it, or use the full path to execute it which changes the above command to

/opt/mssql-tools/bin/sqlcmd -S localhost -U sa -P mysecretpassword

Remote to Linux remote server

When managing a SQL Server instance it is sometimes necessary to connect to the operating system on which it runs. This is very different on Linux to running SQL Server on Windows. There is no remote desktop, and PowerShell doesn’t really work either. That’s where the typical Linux admin tools become necessary.

In a way the Secure Shell or SSH is the Linux equivalent of the Remote Desktop. There are many ways to do it, especially from another Linux box. Most of us, SQL DBAs, will be typically on a Windows machine. In that case a common approach is to use PuTTy. But if you use PowerShell and keep up with updates it is now possible to SSH directly for a PowerShell terminal too.

ssh mysqlonlinux.mydomain.com

or

ssh 10.11.12.13

Simple as that. You will connect with a specific user and now most of the admin commands will have to start with sudo which allows to execute them with elevated permissions. It’s similar to Windows’ Run as Administrator.

Connect to Docker container from the host

If you want to connect to a container running your SQL Server on Linux from the host use docker exec to start a bash shell.

docker exec -it sql1 bash

where sql1 is the name of the container hosting SQL Server.

You will connect with superuser permissions and sudo will not be necessary. The prompt will look something like this:

root@2ff21ad83800:/#

SSH into a container running SQL Server on Linux

If you cannot or don’t want to remote onto the host first to connect to the container running your SQL Server instance SSH is the way to go, again. By default, the SQL Server image does not have SSH server so you will have to install it or a custom image will have to be created. Once that is done it is no different from connecting with the ssh (or PuTTy) to any other Linux server. It doesn’t matter that it runs in a container.

Using SQL Server on Docker (1)

It is yesterday’s news. SQL Server runs on Linux, and on Docker too. There are plenty of blog posts showing how to install, start and connect to it too.

During last Data Relay, I have seen Mark Pryce-Maher‘s talk on SQL Server on Linux (twice). It was a good talk, but it followed the same pattern so not surprisingly a common question from the audience appeared to be ‘why would you want to do it?’.

I mean, the biggest claim to fame appears to be that SQL Server actually works on Linux. It doesn’t bring any new features. In fact, some features are missing. A good part of docker demos appears to be an exercise in futility too. One starts a container with SQL Server, creates a database, inserts some data then the container restarts and as if by magic the data is gone, the database never existed. So, indeed, why would you want to run an SQL Server on Docker?

This post is my attempt to address this question by showing how I use SQL Server on Docker and how I deal with the persistence problem.

First: Get it up and running!

If you want to know how to run and connect to SQL Server on Docker read the Microsoft’s Quickstart Document, or Andrew Pruski’s blog post about running vNext on Docker.

Making the long story short, assuming you are running Windows, have Docker installed, configured to run Linux containers and that you don’t have an SQL Server instance locally installed:

docker pull mcr.microsoft.com/mssql/server:latest
docker run -d -e ACCEPT_EULA=Y -e SA_PASSWORD=Secret1234 -p 1433:1433 --name sql mcr.microsoft.com/mssql/server:latest

That’s all you need to do to have the latest production SQL Server (Developer Edition) instance running on localhost on port 1433.

Creating a database that survives container restarts

If you have the container from the previous example running let’s stop and remove it first.

docker stop sql
docker rm sql

There is a number of ways to persist container data. For my purposes the simples appears to be docker volumes. You can read more about docker persistence on the official documentation site but for now let’s just create a local volume called sqlvolume

docker volume create sqlvolume

and mount that volume on the container

docker run -d -e ACCEPT_EULA=Y -e SA_PASSWORD=Secret1234 -p 1433:1433 --mount source=sqlvolume,target=/data --name sql mcr.microsoft.com/mssql/server:latest

Now, if you create a database with files on the mounted volume like so:

create database [TestDb]
on primary (name = N'TestDb', filename = N'/Data/TestDb.mdf')
log on (name = N'TestDb_log', filename = N'/Data/TestDb.ldf')

and you stop and start the container

docker stop sql
docker start sql

as if by magic, the TestDb database survived the restart.

Why I find it useful

So with some considerable effort, I was able to create a database that survives a restart of a container or that of a host system. Nothing a standard SQL Server instance running on windows couldn’t do! So why do I do it?

I write my blog posts on my laptop and I don’t want to commit to installing SQL Server on it. I want to be able to try out vNext while still having the ability to run demos against 2017 and I’m definitely not installing two instances on my laptop. Luckily I don’t have to. With the above setup all I have to do is run docker start sql and within seconds I have a local instance of SQL Server ready to serve my queries. When I’m done I do docker stop sql and it is as if the SQL Server was never there.