Cloud Image Creation Format and Process
The Cloud Imaging Process on Rackspace Public Cloud
When you create server images, you need to consider many things, from disk storage to scheduling to how long the process takes. This article addresses some of those topics.
Disk Storage
Rackspace cloud servers, with the exception of the OnMetal and Boot from Volume flavors, use a format known as Virtual Hard Disk (VHD) to store their system disks. When a server writes to the disk, the VHD writes each difference between the original, base hard drive and the current system. Because of this, deleting files from a server's hard disk actually causes the underlying VHD to grow slightly. In the Rackspace Cloud, VHDs can only get larger, and they are generally slightly larger than the underlying file system at its largest point. For example:
Build | current 'df –h' size | VHD size |
---|---|---|
Base cloud server Build | 1.8 GB | 2.5 GB |
Cloud server, with 100 GB | 100 GB | 108 GB |
Cloud server, which used to have 100 GB | 5 GB | 115 GB |
Image Creation Process
The following steps outline high level process of creating a cloud server image. For a more in-depth guide, see
Understanding the Cloud Imaging Process.
-
The image is queued for creation or is preparing to start. During this time, the coalesce is preformed to eliminate duplicate data between the back-end VHDs. The API reports 25% progress.
-
The image is being created. This step saves, compresses, and uploads the back-end VHD file to Cloud Files. This step generally takes the longest to complete, because the data is transferred to Cloud Files. The API reports 50% progress.
-
The image creation and upload is complete, and the image process is cleaned up on the back-end. The API reports 100% progress.
Image Creation Duration
A rule of thumb is once the image creation or upload starts (Image creation steps 2 and 3), this generally takes 2 minutes per GB of data on the underlying VHD. This is based on how much data the servers file system contained when it was at its largest. Several factors affect imaging time:
-
Daily and Weekly Imaging: During image creation, a coalesce is the backend operation performed during step 1 that attempts to merge duplicated data in the VHD chain. If a large amount of differential data exists between the VHDs, this process can take some time to complete, and the image creation starts after this process is done.
-
File System Size: The system size is how much disk space is used on your Cloud Server, at its largest.
-
File System Activity: Reads are not highly impactful, but writes can make the image coalesce and creation take much longer.
-
Cloud Server Age: Older servers tend to have larger VHDs,which take longer to image.
-
Cloud Server Flavor: Because the public cloud is a shared environment, resources such as disk and network are shared. Resource constraints on older flavors can slow the imaging process. Newer Rackspace Cloud flavors, such as General Purpose, Performance, and I/O, have greater I/O and network resources, improving performance.
Reasons that Image Creation Fails
If the image creation fails, one of the following issues might be the reason.
Age of the Server - Older flavor classes, such as Standard and Classic, exist on older hardware that has limited network bandwidth for imaging operations. This limit causes imaging to take longer to complete. To work around this issue, you can migrate to a newer flavor type such as General Purpose, or use a file-level backup tool such as Cloud Backup.
Coalesce Timeout - If you do not image your cloud server frequently, or if a large amount of data is changed, added, or modified, the completion time of the coalesce process can increase. If a timeout is reached, you can wait and then send the image creation request again. If you see this issue repeatedly, you might need to limit file system activity while imaging. If image creation continues to fail, Rackspace Support can trigger the coalesce process outside of a Cloud Server Image.
Updated 4 days ago