Windows Azure Storage and Read-Access Geo Redundant Feature (Part 2)

Part 1
In the last post we describe the new feature of Windows Azure Storage that give us the possibility to not only replicate the storage content to another datacenter, but also access in in read-only mode when the primary storage is down - Read Access Geo Redundant Storage (RA-GRS)
In this post we’ll talk about some things that we should be aware when we are staring to use RA-GRS.

Retry Policy
First thing that we need to be aware is the retry policy. With geo redundant feature we need another retry policy that can automatically fall back to the second account when the first down cannot be accessed.
For this purpose a new interface was created called “IExtendedRetryPolicy” was created. This new interface comes with a method called “Evaluate” that detect if the operation should be retried or not. Because we have two storage accounts that needs to be checked there is a small change in behavior. We need a mechanism to switch the two accounts and also to take into the consideration the time interval.
For this behavior we have already an implementation “ExponentialRetry” and “LiniarRetry” that take into account all this things. If you have time you can look over the implementation here: https://github.com/WindowsAzure/azure-storage-net/tree/master/Lib/Common/RetryPolicies.
The retry policy is set through Options of different requests or through the client itself.

CloudBlobClient bc = sa.CreateCloudBlobClient(...)
bc.RetryPolicy = new LiniarRetry();
...
OR
BlobRequestOptions bro = new BlobRequestOptions()
{
  RetryPolicy = new LiniarRetry(),
  ServerTimeout TimeSpan.FromMinutes(1),  
};
CloudBlockBlob blob = container.GetBlockBlobReferecen("FooName");
blob.DownloadFile( fileName, FileMode.OpenOrCreate, null, bro );
...
TableRequestOptions tro = new TableRequestOptions()
{
  RetryPolicy = new LiniarRetry(),
  ServerTimeout TimeSpan.FromMinutes(1),  
};
...
QueueRequestOptions tro = new QueueRequestOptions()
{
  RetryPolicy = new LiniarRetry(),
  ServerTimeout TimeSpan.FromMinutes(1),  
};

Location Mode
In the last post we talk about this property, but when we should set this value? This value should be set at the request option object. There we have a property called LocationMode that will allow to do this.

CloudBlobClient bc = sa.CreateCloudBlobClient(...);
bc.LocationMode = LocationMode.SecondaryOnly;

In this example if the primary account is down (404) or we have a timeout expectation, that a fallback to the second account is made.

In this post we saw how we can change the retry policy when we use the Read Access Geo Redundant Storage.