Azure Table Storage and PowerShell, The Hard Way

24 May

In my previous post I gave a quick overview of the Shared Key authentication scheme used by the Azure storage service and demonstrated how authenticate and access the BLOB storage API through PowerShell.  The file and queue services follow an authentication scheme that aligns with the BLOB requirements, however the table service is a bit different.  I felt it might help the more tortured souls out there (like myself) if I tried to describe the nuances.

Azure Storage REST API, Consistently Inconsistent

Like the REST of all things new Microsoft (read Azure), the mantra is consistency.  From a modern administrative perspective you should have a consistent experience across whatever environment and toolset you require.  If you are a traditional administrator/engineer of the Microsoft stack, the tooling takes the form of PowerShell cmdlets.  If you use Python, bash, etc. there is effectively equivalent tooling available.  My gripes outstanding, I think Microsoft has done a tremendous job in this regard.  I also make no claim that my preferences are necessarily the correct ones.  The ‘inconsistencies’  I will be discussing are not really issues for you if you use the mainline SDK(s).  As usual, I’ll be focusing on how things work behind the scenes and my observations.

Shared Key Authentication, but Not All Are Equal

In exploring the shared key authentication to the BLOB REST API, we generated and encoded the HTTP request signature.  The string we needed to encode looked something like this:

GET\n /*HTTP Verb*/ 
\n /*Content-Encoding*/ 
\n /*Content-Language*/ 
\n /*Content-Length (include value when zero)*/ 
\n /*Content-MD5*/ 
\n /*Content-Type*/ 
\n /*Date*/ 
\n 
\n 
\n 
\n 
\n /*Range*/ 
x-ms-date:Sun, 11 Oct 2009 21:49:13 GMT\n
x-ms-version:2009-09-19\n /*CanonicalizedHeaders*/ 
/myaccount /mycontainer\ncomp:metadata\nrestype:container\ntimeout:20

The table service takes a much simpler and yet arcane format that is encoded in an identical fashion.

GET
 
application/json;odata=nometadata
Mon, 15 May 2017 17:29:11 GMT
/billing73d55f68/fabriclogae0bced538344887a4021ae5c3b61cd0GlobalTime(PartitionKey='407edc6d872271f853085a7a18387784',RowKey='02519075544040622622_407edc6d872271f853085a7a18387784_
0_2952_2640')
 

In this case there are far fewer headers and query parameters to deal with, however there are now fairly rigid requirements. A Date header must be specified as opposed to either Date or x-ms-date, or both in the BLOB case.  A Content-Type header must also be specified as part of the signature, and no additional header details are required.  The canonical resource component is very different from the BLOB service.  The canonical resource still takes a format of <storage account name>/<table name>/<query parameters>.  At the table service level only the comp query parameter is to be included.  As an example, to query the table service properties for the storage account the request would look something like https://myaccount.table.core.windows.net?restype=service&comp=properties. The canonical resource would be /myaccount/?comp=properties.

Generating the Signature with PowerShell

We will reuse our encoding function from the previous post and include a new method for generating the signature.

Function EncodeStorageRequest
{
    [CmdletBinding()]
    param
    (
        [Parameter(Mandatory = $true,ValueFromPipeline=$true,ValueFromPipelineByPropertyName=$true)]
        [String[]]$StringToSign, 
        [Parameter(Mandatory=$true,ValueFromPipelineByPropertyName=$true)]
        [String]$SigningKey
    )
    PROCESS
    {
        foreach ($item in $StringToSign)
        {
            $KeyBytes = [System.Convert]::FromBase64String($SigningKey)
            $HMAC = New-Object System.Security.Cryptography.HMACSHA256
            $HMAC.Key = $KeyBytes
            $UnsignedBytes = [System.Text.Encoding]::UTF8.GetBytes($item)
            $KeyHash = $HMAC.ComputeHash($UnsignedBytes)
            $SignedString = [System.Convert]::ToBase64String($KeyHash)
            Write-Output $SignedString
        }
    }
}
  
Function GetTableTokenStringToSign
{
    [CmdletBinding()]
    param
    (
        [Parameter(Mandatory = $false,ValueFromPipelineByPropertyName = $true)]
        [ValidateSet('GET','PUT','DELETE','POST','OPTIONS','MERGE')]
        [string]
        $Verb="GET",
        [Parameter(Mandatory=$true,ValueFromPipelineByPropertyName = $true)]
        [System.Uri]
        $Resource,
        [Parameter(Mandatory = $false,ValueFromPipelineByPropertyName = $true)]
        [String]
        $ContentMD5, 
        [Parameter(Mandatory = $false,ValueFromPipelineByPropertyName = $true)]
        [String]
        $ContentType,
        [String]
        $Date
    )
  
    $ResourceBase=($Resource.Host.Split('.') | Select-Object -First 1).TrimEnd("`0")
    $ResourcePath=$Resource.LocalPath.TrimStart('/').TrimEnd("`0")
    $SigningPieces=@($Verb,$ContentMD5,$ContentType,$Date)
    #Get the canonicalized resource...
    $CanonicalizedResource="/$ResourceBase/$ResourcePath"
    if([String]::IsNullOrEmpty($Resource.Query) -eq $false)
    {
        foreach ($item in $Resource.Query.TrimStart('?').ToLower().Split('&amp;amp;'))
        {
            if($item.StartsWith('comp='))
            {
                $CanonicalizedResource+="?$item"
                break
            }
        }
    }
    $SigningPieces+=$CanonicalizedResource
    $StringToSign = [String]::Join("`n",$SigningPieces)
    Write-Output $StringToSign
}

So for example, if you want to list the tables in the storage account. The code to generate the request will look something like this:

$AccountName='myaccount'
$AccessKey='vyAEEzbcnIAkLKti1leDbfrAOQBu5bx52zyCkW0fGIBCsS+DDGXpfidOeAWyg7do8ujft1mFhnz9kmliycmiXA=='
 
$Uri="https://$AccountName.table.core.windows.net/tables"
$SignatureParams=@{
    Resource=$Uri;
    Date=[DateTime]::UtcNow.ToString('R');
    Verb='GET';
    ContentType='application/json;odata=nometadata';
}
$RequestSignature=GetTableTokenStringToSign @SignatureParams
$TableToken=EncodeStorageRequest -StringToSign $RequestSignature -SigningKey $AccessKey
$TableHeaders=[ordered]@{
    'x-ms-version'= '2016-05-31'
    'DataServiceVersion'='3.0;Netfx'
    'Accept-Charset'='UTF-8'
    'Accept'='application/json;odata=fullmetadata';
    'Date'=$SignatureParams.Date;
    'Authorization'="SharedKey $($AccountName):$($TableToken)"
}
$RequestParams=@{
    Uri=$SignatureParams.Resource;
    Method=$SignatureParams.Verb;
    Headers=$TableHeaders;
    ContentType=$SignatureParams.ContentType;
    ErrorAction='STOP'
}
$Response=Invoke-WebRequest @RequestParams -Verbose
$Tables=$Response.Content | ConvertFrom-Json | Select-Object -ExpandProperty value
PS C:\WINDOWS\system32> $Tables|fl
 
odata.type : acestack.Tables
odata.id : https://acestack.table.core.windows.net/Tables('provisioninglog')
odata.editLink : Tables('provisioninglog')
TableName : provisioninglog

The astute reader will notice we had to pass some different headers along.  All table requests require either or both a DataServiceVersion or MaxDataServiceVersion.  These values align with maximum versions of the REST API, which I won’t bother belaboring.  We also  retrieved JSON rather than XML, and have a number of content types available to take the format in which are dictated by the Accept header.   In the example we retrieved it with full OData metadata; other valid types include minimalmetadata and nometadata (atom/xml is returned from earlier data service versions).  In another peculiarity XML is the only format returned for retrieving Service properties or stats.

Putting It to Greater Use With Your Old Friend OData

You likely want to actually read some data out of tables.  Now that authorizing the request is out of the way it is a ‘simple’ manner of applying the appropriate OData query parameters.  We will start with retrieving a list of all entities within a table.  This will return a maximum of 1000 results (unless limited using the $top parameter) and a link to any subsequent pages of data will be returned in the response headers.  In the following example we will query all entities in the fabriclogaeGlobalTime table in the fabrixstuffz storage account.  In the interest of brevity I will limit this to 3 results.

GET https:///fabrixstuffz.table.core.windows.net/fabriclogaeGlobalTime()?$top=3
x-ms-version: 2016-05-31
DataServiceVersion: 3.0;Netfx
Accept: application/json;odata=nometadata
Date: Tue, 23 May 2017 05:25:09 GMT
Authorization: SharedKey fabrixstuffz:0GITOAOoZselZ3DaXgJjabYuuKMXtER6j4PA0ZBUcQs=
User-Agent: Mozilla/5.0 (Windows NT; Windows NT 10.0; en-US) WindowsPowerShell/5.1.15063.296
Content-Type: application/json;odata=nometadata

This should yield a result looking like this.

Cache-Control: no-cache
Transfer-Encoding: chunked
Content-Type: application/json;odata=nometadata;streaming=true;charset=utf-8
Server: Windows-Azure-Table/1.0 Microsoft-HTTPAPI/2.0
x-ms-request-id: 56afccf3-0002-0104-0285-d382b4000000
x-ms-version: 2016-05-31
X-Content-Type-Options: nosniff
x-ms-continuation-NextPartitionKey: 1!44!NDA3ZWRjNmQ4NzIyNzFmODUzMDg1YTdhMTgzODc3ODQ-
x-ms-continuation-NextRowKey: 1!88!MDI1MTkwNjc4NDkwNDA1NzI1NjlfNDA3ZWRjNmQ4NzIyNzFmODUzMDg1YTdhMTgzODc3ODRfMF8yOTUyXzI2NDA-
Date: Tue, 23 May 2017 05:27:28 GMT
{
    "value":  [
                  {
                      "PartitionKey":  "407edc6d872271f853085a7a18387784",
                      "RowKey":  "02519067840040580939_407edc6d872271f853085a7a18387784_0_2952_2640",
                      "Timestamp":  "2017-05-23T05:25:55.6307353Z",
                      "EventType":  "Time",
                      "TaskName":  "FabricNode",
                      "dca_version":  -2147483648,
                      "epoch":  "1",
                      "localTime":  "2017-05-23T05:21:07.4129436Z",
                      "lowerBound":  "2017-05-23T05:19:56.173659Z",
                      "upperBound":  "2017-05-23T05:19:56.173659Z"
                  },
                  {
                      "PartitionKey":  "407edc6d872271f853085a7a18387784",
                      "RowKey":  "02519067843040711216_407edc6d872271f853085a7a18387784_0_2952_2640",
                      "Timestamp":  "2017-05-23T05:20:53.9265804Z",
                      "EventType":  "Time",
                      "TaskName":  "FabricNode",
                      "dca_version":  -2147483648,
                      "epoch":  "1",
                      "localTime":  "2017-05-23T05:16:07.3678218Z",
                      "lowerBound":  "2017-05-23T05:14:56.1606307Z",
                      "upperBound":  "2017-05-23T05:14:56.1606307Z"
                  },
                  {
                      "PartitionKey":  "407edc6d872271f853085a7a18387784",
                      "RowKey":  "02519067846040653329_407edc6d872271f853085a7a18387784_0_2952_2640",
                      "Timestamp":  "2017-05-23T05:15:52.7217857Z",
                      "EventType":  "Time",
                      "TaskName":  "FabricNode",
                      "dca_version":  -2147483648,
                      "epoch":  "1",
                      "localTime":  "2017-05-23T05:11:07.3406081Z",
                      "lowerBound":  "2017-05-23T05:09:56.1664211Z",
                      "upperBound":  "2017-05-23T05:09:56.1664211Z"
                  }
              ]
}

You should recognize a relatively standard OData response, with our desired values present within an array as the value property. There are two response headers to note here; x-ms-continuation-NextPartitionKey and x-ms-continuation-NextRowKey. These headers are the continuation token for retrieving the next available value(s). The service will return results in pages with a maximum length of 1000 results, unless limited using the $top query parameter like the previous example. If one were so inclined, they could continue to send GET requests, including the continuation token(s) until all results are enumerated.

Creating (or updating) table entities is a slightly different exercise, which can become slightly convoluted (at least in PowerShell or other scripts).  Conceptually, all that is required to create an entity is a POST  request to the table resource URI with a body containing the entity and the appropriate required headers.  The complexity is primarily a result of the metadata overhead associated with the server OData implementation. We’ll examine this by inserting an entity into a fictional customers table.

$TableName='fakecustomers'
$Uri="https://$AccountName.table.core.windows.net/$TableName"
$SignatureParams=@{
    Resource=$Uri;
    Date=[DateTime]::UtcNow.ToString('R');
    Verb='POST';
    ContentType='application/json;odata=nometadata';
}
$RequestSignature=GetTableTokenStringToSign @SignatureParams
$TableToken=EncodeStorageRequest -StringToSign $RequestSignature -SigningKey $AccessKey
$TableHeaders=[ordered]@{
    'x-ms-version'= '2016-05-31'
    'DataServiceVersion'='3.0;Netfx'
    'Accept-Charset'='UTF-8'
    'Accept'='application/json;odata=fullmetadata';
    'Date'=$SignatureParams.Date;
    'Authorization'="SharedKey $($AccountName):$($TableToken)"
}
$PartitionKey='mypartitionkey'
$RowKey='row771'
$TableEntity=New-Object PSobject @{
   "Address"="Mountain View";
   "Name"="Buckaroo Banzai";
   "Age"=33;  
   "AmountDue"=200.23;
   "FavoriteItem"="oscillation overthruster";
   "CustomerCode@odata.type"="Edm.Guid";  
   "CustomerCode"="c9da6455-213d-42c9-9a79-3e9149a57833";  
   "CustomerSince@odata.type"="Edm.DateTime";  
   "CustomerSince"="2008-07-10T00:00:00";  
   "IsActive"=$true;  
   "NumberOfOrders@odata.type"="Edm.Int64";  
   "NumberOfOrders"="255";  
   "PartitionKey"=$PartitionKey;  
   "RowKey"=$RowKey
}
$RequestParams=@{
    Uri=$SignatureParams.Resource;
    Method=$SignatureParams.Verb;
    Headers=$TableHeaders;
    ContentType=$SignatureParams.ContentType;
    ErrorAction='STOP'
}
$Response=Invoke-WebRequest @RequestParams

You should end up receiving the inserted object as a response:

PS C:\Windows\system32> $Response.Content | ConvertFrom-Json
PartitionKey   : mypartitionkey
RowKey         : row772
Timestamp      : 2017-05-23T06:17:53.7244968Z
CustomerCode   : c9da6455-213d-42c9-9a79-3e9149a57833
FavoriteItem   : oscillation overthruster
AmountDue      : 200.23
IsActive       : True
CustomerSince  : 2008-07-10T00:00:00
Name           : Buckaroo Banzai
NumberOfOrders : 255
Age            : 33
Address        : Mountain View

You should notice that the object we submitted had some extra properties not present on the inserted entity. The API requires that for any entity property where the (.Net) data type can not be automatically inferred, a type annotation must be specified. In this case CustomerCode=c9da6455-213d-42c9-9a79-3e9149a57833 is a GUID (as opposed to a string) requires a property CustomerCode@odata.type=Edm.Guid.  If you would like a more complete explanation the format is detailed here.

Three ways to do the same thing

You’ve got to give it to Microsoft, they certainly keep things interesting.  In the above example, I showed one of three ways that you can insert an entity into a table.  The service supports Insert, Insert or Merge (Upsert), and Insert or Replace operations (there are also individual Replace and Merge operations).  In the following example I will show the Upsert operation using the same table and entity as before.

$Uri="https://$AccountName.table.core.windows.net/$TableName(PartitionKey='$PartitionKey',RowKey='$RowKey')"
$SignatureParams=@{
    Resource=$Uri;
    Date=[DateTime]::UtcNow.ToString('R');
    Verb='MERGE';
    ContentType='application/json;odata=nometadata';
}
$RequestSignature=GetTableTokenStringToSign @SignatureParams
$TableToken=EncodeStorageRequest -StringToSign $RequestSignature -SigningKey $AccessKey
$TableEntity | Add-Member -MemberType NoteProperty -Name 'NickName' -Value 'MrMan'
$TableHeaders=[ordered]@{
    'x-ms-version'= '2016-05-31'
    'DataServiceVersion'='3.0;Netfx'
    'Accept-Charset'='UTF-8'
    'Accept'='application/json;odata=fullmetadata';
    'Date'=$SignatureParams.Date;
    'Authorization'="SharedKey $($AccountName):$($TableToken)"
}
$RequestParams = @{
    Method        = 'MERGE';
    Uri           = $Uri;
    Body          = $($TableEntity|ConvertTo-Json);
    Headers       = $TableHeaders;
    ContentType   = 'application/json;odata=fullmetadata'
}
$Response=Invoke-WebRequest @RequestParams

This should yield a response with the meaningful details of the operation in the headers.

PS C:\Windows\system32> $Response.Headers
Key                    Value 
---                    ----- 
x-ms-request-id        48489e3d-0002-005c-6515-d545b8000000 
x-ms-version           2016-05-31 
X-Content-Type-Options nosniff 
Content-Length         0 
Cache-Control          no-cache 
Date                   Thu, 25 May 2017 05:08:58 GMT 
ETag                   W/"datetime'2017-05-25T05%3A08%3A59.5530222Z'"
Server                 Windows-Azure-Table/1.0 Microsoft-HTTPAPI/2.0 

Now What?

I’m sure I’ve bored most of you enough already so I won’t belabor any more of the operations, but I hope that I’ve given you a little more insight into the workings of another key element of the Azure Storage Service(s). As always, if you don’t have a proclivity for doing things the hard way, feel free to check out a module supporting most of the Table (and BLOB) service functionality on the Powershell Gallery or GitHub.

Chris Speers

Systems Engineer Par Excellence within Avanade’s Azure Cloud Enablement Group.

LinkedIn