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.
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.
- 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.
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 | | |
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.
The following table lists the minimum requirements setting up a High Availability (HA) cluster:
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 |
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.
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.
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.
The Operating System used will depend completely on the flavour of Kubernetes deployed. Examples are RancherOS, Ubuntu, CoreOS etc.
For Bare Metal deployment: Ubuntu 18+
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 |
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.
Internet Connectivity Requirement During updates and initial installation, internet connectivity is needed. For usual operations, internet connectivity isn’t required.
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.
Last modified 1yr ago