Best practices for File Gateway

This section contains the following topics, which provide information about the best practices for working with gateways, file shares, buckets, and data. We recommend that you familiarize yourself with the information outlined in this section, and attempt to follow these guidelines in order to avoid problems with your AWS Storage Gateway. For additional guidance on diagnosing and solving common issues you might encounter with your deployment, see Troubleshooting problems with your Storage Gateway deployment.

Best practices: recovering your data

Although it is rare, your gateway might encounter an unrecoverable failure. Such a failure can occur in your virtual machine (VM), the gateway itself, the local storage, or elsewhere. If a failure occurs, we recommend that you follow the instructions in the appropriate section following to recover your data.

Important

Storage Gateway doesn’t support recovering a gateway VM from a snapshot that is created by your hypervisor or from your Amazon EC2 Amazon Machine Image (AMI). If your gateway VM malfunctions, activate a new gateway and recover your data to that gateway using the instructions following.

Recovering from an unexpected virtual machine shutdown

If your VM shuts down unexpectedly, for example during a power outage, your gateway becomes unreachable. When power and network connectivity are restored, your gateway becomes reachable and starts to function normally. Following are some steps you can take at that point to help recover your data:

• If an outage causes network connectivity issues, you can troubleshoot the issue. For information about how to test network connectivity, see Testing your gateway's network connectivity.

Recovering your data from a malfunctioning cache disk

If your cache disk encounters a failure, we recommend you use the following steps to recover your data depending on your situation:

• If the malfunction occurred because a cache disk was removed from your host, shut down the gateway, re-add the disk, and restart the gateway.

Recovering your data from an inaccessible data center

If your gateway or data center becomes inaccessible for some reason, you can recover your data to another gateway in a different data center or recover to a gateway hosted on an Amazon EC2 instance. If you don't have access to another data center, we recommend creating the gateway on an Amazon EC2 instance. The steps you follow depends on the gateway type you are covering the data from.

To recover data from a File Gateway in an inaccessible data center

For File Gateway, you map a new to the Amazon S3 bucket that contains the data you want to recover.

1. Create and activate a new File Gateway on an Amazon EC2 host. For more information, see Deploy a default Amazon EC2 host for S3 File Gateway.

2. Create a new on the EC2 gateway you created. For more information, see Create a file share.

3. Mount your file share on your client and map it to the S3 bucket that contains the data that you want to recover. For more information, see Mount and use your file share.

Best practices: managing multipart uploads

When transferring large files, S3 File Gateway makes use of the Amazon S3 multipart upload feature to split the files into smaller parts and transfer them in parallel for improved efficiency. For more information about multipart upload, see Uploading and copying objects using multipart upload in the Amazon Simple Storage Service User Guide.

If a multipart upload doesn't complete successfully for any reason, the gateway typically stops the transfer, deletes any partially-transferred pieces of the file from Amazon S3, and attempts the transfer again. In rare cases, such as when hardware or network failure prevent the gateway from cleaning up after an unsuccessful multipart upload, pieces of the partially-transferred file might remain on Amazon S3 where they can incur storage charges.

As a best practice for minimizing Amazon S3 storage costs from incomplete multipart uploads, we recommend configuring an Amazon S3 bucket lifecycle rule that uses the AbortIncompleteMultipartUpload API action to automatically stop unsuccessful transfers and delete associated file parts after a designated number of days. For instructions, see Configuring a bucket lifecycle configuration to delete incomplete multipart uploads in the Amazon Simple Storage Service User Guide.

Best practices: Unzip compressed files locally before copying to a gateway

If you try to unzip a compressed archive containing thousands of files while it is stored on your gateway, you might encounter significant performance-related delays. The process of unzipping an archive that contains large numbers of files on any type of network file share inherently involves a high volume of input/output operations, metadata cache manipulation, network overhead, and latency. Additionally, Storage Gateway is unable to determine when each file from the archive has finished unzipping, and can begin uploading files before the process is complete, which further impacts performance. These issues are compounded when the files inside the archive are numerous, but small in size.

As a best practice, we recommend transferring compressed archives from your gateway to your local machine first, before you unzip them. Then, if necessary, you can use a tool such as robocopy or rsync to transfer the unzipped files back to the gateway.

Retain file attributes when copying data from Windows Server

It is possible to copy files to your File Gateway using the basic copy command on Microsoft Windows, but this command copies only the file data by default - omitting certain file attributes such as security descriptors. If the files are copied to the gateway without the corresponding security restrictions and Discretionary Access Control List (DACL) information, it is possible that they could be accessed by unauthorized users.

As a best practice for preserving all file attributes and security information when copying files to your gateway on Microsoft Windows Server, we recommend using the robocopy or xcopy commands, with the /copy:DS or /o flags, respectively. For more information, see robocopy and xcopy in the Microsoft Windows Server command reference documentation.

Best practices: Proper sizing of cache disks

For best performance, the total disk cache size must be large enough to cover the size of your active working set. For read-heavy and mixed read/write workloads, this ensures that you can achieve a high percentage of cache hits on reads, which is desirable. You can monitor this via the CacheHitPercent metric for your S3 File Gateway.

For write-heavy workloads (e.g. for backup and archival), the S3 File Gateway buffers incoming writes on the disk cache prior to copying this data asynchronously to Amazon S3. You should ensure that you have sufficient cache capacity to buffer written data. The CachePercentDirty metric provides an indication of the percentage of the disk cache that has not yet been persisted to AWS.

Low values of CachePercentDirty are desirable. Values that are consistently close to 100% indicate that the S3 File Gateway is unable to keep up with the rate of incoming write traffic. You can avoid this by either increasing the provisioned disk cache capacity, or increasing the dedicated network bandwidth available from the S3 File Gateway to Amazon S3, or both.

For more information about cache disk sizing, see Amazon S3 File Gateway cache sizing best practices on the official Amazon Web Services YouTube channel.

Working with multiple file shares and Amazon S3 buckets

When you configure a single Amazon S3 bucket to allow multiple gateways or file shares to write to it, the results can be unpredictable. You can configure your buckets in one of two ways to avoid unpredictable results. Choose the configuration method that best fits your use case from the following options:

• Configure your S3 buckets so that only one file share can write to each bucket. Use a different file share to write to each bucket.

  To do this, create an S3 bucket policy that denies all roles except for the role that's used for a specific file share to put or delete objects in the bucket. Attach a similar policy to each bucket, specifying a different file share to write to each bucket.

  The following example policy denies S3 bucket write permissions to all roles except for the role that created the bucket. The s3:DeleteObject and s3:PutObject actions are denied for all roles except "TestUser". The policy applies to all objects in the "arn:aws:s3:::amzn-s3-demo-bucket/*" bucket.

  {
     "Version":"2012-10-17",
     "Statement":[
        {
           "Sid":"DenyMultiWrite",
           "Effect":"Deny",
           "Principal":"*",
           "Action":[
              "s3:DeleteObject",
              "s3:PutObject"
           ],
           "Resource":"arn:aws:s3:::amzn-s3-demo-bucket/*",
           "Condition":{
              "StringNotLike":{
                 "aws:userid":"TestUser:*"
              }
           }
        }
     ]
  }

• If you do want multiple file shares to write to the same Amazon S3 bucket, you must prevent the file shares from trying to write to the same objects simultaneously.

  To do this, configure a separate, unique object prefix for each file share. This means that each file share only writes to objects with the corresponding prefix, and doesn't write to objects that are associated with the other file shares in your deployment. You configure the object prefix in the S3 prefix name field when you create a new file share.

Clean up unnecessary resources

As a best practice, we recommend cleaning up Storage Gateway resources to avoid unexpected or unnecessary charges. For example, if you created a gateway as a demonstration exercise or a test, consider deleting it and its virtual appliance from your deployment. Use the following procedure to clean up resources.

To clean up resources you don't need

1. If you no longer plan to continue using a gateway, delete it. For more information, see Deleting your gateway and removing associated resources.

2. Delete the Storage Gateway VM from your on-premises host. If you created your gateway on an Amazon EC2 instance, terminate the instance.