What really is a Snapshot? Let's break into the more details. A Virtual Machine Snapshot is a technology that is executed to save a specific state of that VM by purpose of preserving VM's data, power state of VM and also its virtual memory. You can generate many snapshots to keep different states of your VM and is required for VM Backup procedure and is a great ability in test / pilot scenarios. So you can revert to any snapshot state if you need by Snapshot Manager. Then should be remembered very change is this duration (from snapshot to recent moment) will be discarded.
But what have been affected whenever we create a new snapshot and what are the Pros and Cons of this feature? In this post I want to describe more detail about virtual machine snapshot feature inside the vSphere environments…
As a detail view snapshot exactly is a replica copy of VMDK in a specific moment. So it can be used for recovering a system from a failure state. All the Backup solutions work with snapshot every time they start a VM backup task to provide a copy of that. So as I said before snapshot generation provide a copy from these component’s contents:
1. VM Settings (Hardware settings or any changes to the VM itself)
2. VMDK state (Data has been write inside the VM Guest OS)
3. VMEM Content (Virtual Memory like clipboard or swap contents)
But what have been affected whenever we create a new snapshot and what are the Pros and Cons of this feature? In this post I want to describe more detail about virtual machine snapshot feature inside the vSphere environments…
As a detail view snapshot exactly is a replica copy of VMDK in a specific moment. So it can be used for recovering a system from a failure state. All the Backup solutions work with snapshot every time they start a VM backup task to provide a copy of that. So as I said before snapshot generation provide a copy from these component’s contents:
1. VM Settings (Hardware settings or any changes to the VM itself)
2. VMDK state (Data has been write inside the VM Guest OS)
3. VMEM Content (Virtual Memory like clipboard or swap contents)
So must be careful of using revert action because it will return all of this objects to the snapshot state. During snapshot generation it will create a delta file with .vmdk extension (also called redo logs or delta links) that acts like a child for its parent .vmdk (main VMDK before snapshot creation). So guess OS cannot do write operation on parent vmdk anymore and after that any disk writes action will be happened into the delta/child disks. First of all, child disk will be created from its parent and then the other successive children snapshots will be created from latest delta.vmdk in this chain. As its name shows, delta means difference between current state of the VM disk and last snapshot creation moment. Now any change in VM (and its Guest OS) will be writing to this new VMDK (delta file) from this moment. So we can say they are important as their parent.
But what is the exact content of snapshot files? And where data has been written after taking snapshots? Consider A as the primary VMDK of virtual machine when there is no snapshots. B series (B1, B2) are children of A and also C files (C1, C2) are descendant of B. If you are after C2 snapshot, content data after reverting to C1 state is included base VMDK (or flat file) and previous delta files: A+B1+B2+C1. Flat.vmdk is raw data structure of base file but it’s not a separate file when you check it into the datastore.
Into the Virtual Machine File System (VMFS), delta disk act as a sparse disk and it’s required to know about how data will store in virtual disks. There is a mechanism called COW (copy-on-write) for optimization of storage spaces. It means there is nothing into the VMDK until data copy occurs. I will explain more about COW mechanism and sparse disks more deeply in another post.
Now when you create many snapshots and cause complexity in the parent/child relations between snapshots, you may need to execute consolidation to reduce this confusing situation. It will merge these redo logs/delta vmdk inside a single vmdk to avoid complex status of snapshot managing. If the child disks are large in size, the consolidation operation may take more time.
There is also some another files related to the snapshot operation:
VMSN: It is a container for memory contents of VM. As VMware said if the snapshot includes the memory option, the ESXi host writes the memory of the virtual machine to disk. The VM is stunned during memory is being written but sadly you cannot pre-calculate time duration of that because it’s dependent on many factors such as disk performance and size of memory.
Remember VMSN always will be generated even if you don’t select memory option in snapshot creation. But its size is much lesser in non-memory state. So VMSN size is an overhead for total space calculation of snapshot generation in the datastore.
VMSD: It is the snapshot database and is the primary source for snapshot manager usage and its contents are relation tree of snapshots. So snapshot.vmsd consists of current config and active state of virtual machine.
But what is the exact content of snapshot files? And where data has been written after taking snapshots? Consider A as the primary VMDK of virtual machine when there is no snapshots. B series (B1, B2) are children of A and also C files (C1, C2) are descendant of B. If you are after C2 snapshot, content data after reverting to C1 state is included base VMDK (or flat file) and previous delta files: A+B1+B2+C1. Flat.vmdk is raw data structure of base file but it’s not a separate file when you check it into the datastore.Now when you create many snapshots and cause complexity in the parent/child relations between snapshots, you may need to execute consolidation to reduce this confusing situation. It will merge these redo logs/delta vmdk inside a single vmdk to avoid complex status of snapshot managing. If the child disks are large in size, the consolidation operation may take more time.
There is also some another files related to the snapshot operation:
VMSN: It is a container for memory contents of VM. As VMware said if the snapshot includes the memory option, the ESXi host writes the memory of the virtual machine to disk. The VM is stunned during memory is being written but sadly you cannot pre-calculate time duration of that because it’s dependent on many factors such as disk performance and size of memory.
Remember VMSN always will be generated even if you don’t select memory option in snapshot creation. But its size is much lesser in non-memory state. So VMSN size is an overhead for total space calculation of snapshot generation in the datastore.
VMSD: It is the snapshot database and is the primary source for snapshot manager usage and its contents are relation tree of snapshots. So snapshot.vmsd consists of current config and active state of virtual machine.
This is a sample of VMSD file:
" .encoding = "UTF-8"
snapshot.lastUID = "1"
snapshot.current = "1"
snapshot0.uid = "1"
snapshot0.filename = "vm-Snapshot1.vmsn"
snapshot0.displayName = "Clean"
snapshot0.createTimeHigh = "360682"
snapshot0.createTimeLow = "2032299104"
snapshot0.numDisks = "1"
snapshot0.disk0.fileName = "vm.vmdk"
snapshot0.disk0.node = "scsi0:0"
snapshot.numSnapshots = "1"a "
I will talk about snapshot quiescing option and some more detail in the next part as soon as possible. For more information you can refer to following links:
" .encoding = "UTF-8"
snapshot.lastUID = "1"
snapshot.current = "1"
snapshot0.uid = "1"
snapshot0.filename = "vm-Snapshot1.vmsn"
snapshot0.displayName = "Clean"
snapshot0.createTimeHigh = "360682"
snapshot0.createTimeLow = "2032299104"
snapshot0.numDisks = "1"
snapshot0.disk0.fileName = "vm.vmdk"
snapshot0.disk0.node = "scsi0:0"
snapshot.numSnapshots = "1"a "
I will talk about snapshot quiescing option and some more detail in the next part as soon as possible. For more information you can refer to following links:
No comments:
Post a Comment