Private Cloud Deployment

GMetri can run a dedicated cluster in the AWS/GCP/Azure region of your choice to make sure you internal compliance needs are met. Please note that you need to be subscribed to an enterprise tier license to request this chargeable feature. Contact us to know more.

Does GMetri support on-premise deployments/Are there additional costs for the same?​

There are cases where security policies demand an on-premise deployment of cloud services (firewalled zone).

For such cases, at an additional cost, GMetri supports on-premise deployments.

Additional Costs​

• There will be a one-time cost that can be determined if you contact us with the following details:

  • The hardware configuration of the servers on which this service is to be deployed.

  • Whether you use a cloud service provider like Google Cloud/AWS/Azure/Others or if you manage, provision your own servers and server racks (bare-metal).

  • The general practice followed to (remotely) access the servers.

  • The protocol to open certain network ports for the servers.

• There may be additional SLA charges as needed.

These costs are in addition to the standard Enterprise Tier annual/monthly license cost you would incur even without the on-Prem deployment.

The option to deploy on-premise is available with the Enterprise Tier License only.

Which network ports needs to open for a private cloud deployment?​

Ideally, within the cluster, communication needs to be open - i.e. all cluster nodes should be able to contact all ports of all other cluster nodes.

However in case a list of open ports is required, you can use the following list:

Description	Within Cluster (In AND Out)	External Load Balancer (In)	Internet (Out)
Kubernetes Specific	22 TCP 80 TCP 443 TCP 2376 TCP 2379 TCP 2380 TCP 6443 TCP 6783 TCP 6783-6784 UDP 8472 UDP 9099 TCP 10250 TCP 10254 TCP	80 TCP 443 TCP	git.rancher.io: 35.160.43.145:32 35.167.242.46:32 52.33.59.17:32 .gmetri.com .gmetri.io .vrgmetri.com .docker.io
Workload Specific	30000-32767 TCP & UDP
Experience Specific	3000-3100 TCP 3306 TCP 5432 TCP 6739 TCP

Deployment Types and Hardware Requirements​

Private Cloud Deployment (firewalled zone)​

The GMETRI platform deploys as a Kubernetes cluster.

Below are the hardware and network requirements to set up a Kubernetes cluster in a private cloud that supports the GMetri platform.

Server Requirements​

The following table lists the minimum requirements setting up a High Availability (HA) cluster:

On Cloud (AWS/GCP/Azure)​

Spec	Production Values
Recommended Node Type	AWS: m4.xlarge/m5.xlarge GC: e2-standard-4 Azure: D4d v4
Nodes (Virtual Machines) with vCPUs >2Ghz	3 worker nodes Min requirement for each node: 4 core vCPU + 16 GB RAM
RAM	16 GB × number of nodes
Storage (SSD Based or better)	Min 256 GB per node supporting >500 sequential IOPS
Managed DB instance*	Managed DB instances from AWS/GCP/Azure with the minimum config of 4core vCPU + 16 GB RAM

Bare Metal (HA Setup)​

Spec	Production Values
Servers with every core >2Ghz	3 (etcd + control plane) + 3 worker nodes Minimum 4 servers on at-least 3 different physical machines (in case the servers are VMs) Min requirement for each node: 4 core vCPU + 16 GB RAM
RAM	16 GB × number of nodes
Storage (SSD Based or better)	Min 256 GB per node supporting >500 sequential IOPS
Managed DB instance	3 x 4 core vCPU + 16 GB RAM

• Full network connectivity between all systems in the cluster

• Ability to configure open ports on these systems

• Failure resilience for production servers: 1 for master nodes and 1 for worker nodes separately. The resilience is of (n-1)/2 nodes per node-type where n is the number of nodes.

Single Rack-server/Desktop Deployment​

We ideally recommend that GMetri is deployed on a cluster setup in a private cloud consisting of a minimum of 3 servers. However, in cases where that is not possible, GMetri can also be deployed on a single rack-server or a desktop.

Hardware Requirements​

The following are the recommended hardware requirements for such a deployment:

Hardware	Specification	Example
CPU	A 4-core/8-thread processor	Intel i7-7700K
RAM	32 GB DDR-4	Corsair LPX 32GB (2x16GB) 3200MHz
HDD	256 GB SSD or SAS storage	Samsung 860 EVO 250GB SSD
LAN Ports	A single LAN port is enough	(Part of the server/desktop)

Keep in mind that this is not a HA (High Availability) setup. This means that any hardware failure could lead to permanent loss of data. In cases where data resiliency is critical, we recommend a cluster setup.

Other Requirements​

Operating System Requirements​

The Operating System used will depend completely on the flavour of Kubernetes deployed. Examples are RancherOS, Ubuntu, CoreOS etc.

Preferred OS​

For Bare Metal deployment: Ubuntu 18+

Partitioning​

Ensure that there is no swap partition.

Partition	Space Allocation	Partition Type
NO swap	none
/ (root)	20 GB	ext4
/home	10 GB	ext4
/var	Remaining space	ext4

Network Requirements​

Network Response Time​

Spec	Web-browser to Datacenter	Within Datacenter
Bandwidth	5 Mbps minimum	100 Mbps minimum
Latency	250-300 milliseconds	1 millisecond on LAN

Within the cluster, ideally communication needs to be open - i.e. all cluster nodes should be able to contact all ports of all other cluster nodes.

In case a list of specific ports is needed, refer to this here.

Internet Connectivity Requirement During updates and initial installation, internet connectivity is needed. For usual operations, internet connectivity isn’t required.

Example Rack Server/Desktop Tower configuration​

An example of a single rack server or a desktop tower to support the above deployment would be the following:

An example of a single rack server or a desktop tower to support the above deployment would be the following:

Hardware	Recommended Spec	Example
CPU	A CPU with 4 threads	Intel i5-7400 or better
RAM	16GB DDR4 or better	Corsair LPX 16GB 3200MHz
HDD	256 GB SSD or SAS storage	Samsung 860 EVO 250GB SSD

4 or 6 such servers/desktop towers are required to form a cluster as mentioned above.

For any other queries, write to [e-mail]