Google Compute Engine Flashcards
What is a Virtual Machine considered in GCE?
Another word for VM is Compute Engine Instance
What type of service is GCE?
GCE is an Infrastructure as a Service offering which means that you have manage and scale your Virtual Machines but they are flexible for lifting and shifting your current infrastructure into the cloud.
Region
Geographical locations that make up one or more zones, usually chosen based off of what area has the most network traffic
Zone
An area within a region chosen to provide the least latency and proximity to the point of service
Boot Disk
Predefined images will have default standard persistent disk (block) storage but can be changed to SSD. Size is also predefined but can be changed.
Application Images
Can be used if you require Microsoft SQL per-installed on your Window’s Server. Custom images are available per project where an image has been created from a boot disk.
Snapshots
You can snapshot a persistent disk and use this as a part of a new instance creation even if they’re part of a running instance. Global resources that are available to any instance in the same project. Snapshot schedules can prevent data loss.
Existing Disks
Disks that can be attached to a new instance.
Labels
The management section has label options that work in key-value pairs to differentiate between projects, enviornments and more.
Deletion Protection
Can be enabled by setting the deletionProtection property on the instance resource. Only nessasary on critical VM resources that need to stay running indefinately. Can only be run by users who have the compute instance create permissions.
Metadata
Allows for future queries to pull information about the instance. It’s stored on the Metadata server and can be queried from the Compute Engine API or the instance itself pulling info such as hostname, instance ID, or any custom metadata we apply. Also can apply script URLs to buckets to ensure execution.
Start-up Scripts
These are scripts that run everytime your instance boots up or restarts. Usually to ensure certain software is installed or services are started. Procesed through the same metadata server and can also be located in Google Cloud Storage.
Preemptibility
Brings the cost of an instance down, but only lasts a maximum of 24 hours. Stopping and starting an instance resets the counter. No SLA is applied to these and they can be removed with as little as 30 seconds of notice. Can ease workload of existing instances.
Availability Policy
When Google maintenance is performed your VM will be moved to another host. GCE will live migrate if this setting isn’t set. You applications will suffer performance wise but will have no downtime. If your app needs high availability, you may want to terminate and start another host. Google sends a report either way.
Automatic Restart
If underlying hardware causes your instances to crash, then GCE offers a setting to automatically restart the instance by setting the automatic restart field to “true”. Google will report any restart.
Shielded VM
Offers ability to harden your VM with security controls using vTPM to provide a virtual root-of-trust to verify the identity of the VM and ensure they’re part of a specified region or project. Does not add any cost to your VM.
Deletion Rule
Allows us to either delete or keep the boot disk when the instance is deleted, not enabled by default. Make sure the checkbox is unchecked if you need to keep the boot disk.
Node Affinity Labels
Sole tenancy, a physical compute engine server designed for your dedicated use (the underlying host hardware and hypervisor handles only your GCE Instances), since some companies require phsyical resources to be seperate from others. VMs will still have the same live migration functionality and shared instances but there are cost implications.
Graphics Processing Units
Used with Graphics intensive workloads such as 3D rendering or Virtual Applications. It can only be attached to predefined or custom machine types and is only available in certain zones.
Tensor Processing Units
Custom Designed using Google’s experience in Machine Learning (ML) and can maximize performance and flexibility in building TensorFlow clusters and other ML workloads. TPU should be in the same zone as compute engine instance. Preemptibility can be enabled on TPU nodes.
Instance Group Rules
Each machine in an instance group comes from the same instance template which defines the machine types, boot disk images, and other instance properties. Global resources but zonal resources in the template will restrict the template to zones the resource resides in.
Managed Instance Group
Allows your workload to be scalable and highly available via automated services in the groups such as autoscaling or autohealing. Always recommended unless unavailable because of preexisting configurations.
Unmanged Instance Group
Allows your workloads to be balanced across a fleet of VMs that are not identical. They are self-managed so autohealing and autoscaling are not supported.
Port-Name Mapping
Used in conjunction with a load balancer and can be specified by a port name and associated port number
AutoScaling Policy
Enables many settings based on CPU usage, HTTP(S) Load Balancing usage, or Stackdriver metrics. Metrics can be customized, the simplist form is CPU Usage. Always acts conservatively and rounds up statistics.
AutoHealing Policy
Validates whether each VM instance in our group is running and ready to accept requests aka a healthcheck. Typically basedon HTTP(S), TCP, or SSL protocols. Parameters should be configured to help determine if a VM is unhealthy and a new instance needs to be added.
Default Limits
24 CPUs, 4096GBs of persistent disk storage, 15000 instances per network
