Choose the right OpenShift size

---

---
template: inverse
class: center, middle

# Choose the right OpenShift size
Jan Šafránek, jsafrane@redhat.com <br/>
<img src="GitHub-Mark-120px-plus.png" width="25em"> jsafrane

---

# Choose the right OpenShift size

* OpenShift form factors
  * Highly Available
  * Single Node
* Hypershift
* Microshift

---

# OpenShift

--
* Distribution of Kubernetes.

--
* Managed by operators.
--

* API.
--
* Form factors:
  * Highly available.
  * Single-node.
  * Selected during installation.

---
# Highly Avialable OpenShift

* 3 control plane nodes.
    * Not schedulable by default.
* Tolerates loss of 1 control plane node.
* Upgrades / reconfigurations without disruption.

---
# Highly Avialable OpenShift
<img class="plain inline" width="800px" src="ha.svg" />

---

# Highly Available OpenShift
## Control plane nodes

* [Bare minimum](https://docs.openshift.com/container-platform/4.15/scalability_and_performance/recommended-performance-scale-practices/recommended-control-plane-practices.html#master-node-sizing_recommended-control-plane-practices):
    * 4 vCPU cores.
    * 16 GB of RAM.
* [Tested maximum](https://docs.openshift.com/container-platform/4.15/scalability_and_performance/recommended-performance-scale-practices/recommended-control-plane-practices.html#master-node-sizing_recommended-control-plane-practices):
    * 16 vCPU cores.
    * 96 GB of RAM.
    * Fast storage for etcd.
    * Maximum:
        * 2,000 Nodes.
        * 10,000 Namespaces.
        * 150,000 Pods.
        * 10,000 Services.

---

# Highly Available OpenShift
* Control plane node
```
sh-5.1# free -h
           total   used   free  shared  buff/cache available
Mem:    15Gi  6.0Gi  658Mi    73Mi       9.1Gi     9.3Gi
```

* Worker (without Prometheus :-))
```
           total  used   free  shared  buff/cache available
Mem:    30Gi 1.9Gi   24Gi    61Mi       5.4Gi      29Gi
```

---

# Single-node OpenShift (SNO)

* 1 node.
* *If it breaks, it breaks.*
* Upgrades / reconfiguration can have downtime.
* More workers can be added.
* Use case:
    * Edge devices with full OpenShift API.
    * Small development cluster.

---

# Single-node OpenShift (SNO)

* Official requirements:
    * 8 vCPU cores.
    * 16 GB of RAM.
    * 120 GB.

```
sh-5.1# free -h
       total  used  free   shared  buff/cache  available
Mem:   30Gi   10Gi  813Mi  100Mi   19Gi        20Gi

sh-5.1# uptime
[...] load average: 0.61, 0.52, 0.34
```

---

# Single-node OpenShift (SNO)

---

# Single-node OpenShift (SNO)

Helpful features:

* Real-time kernel.
* CPU isolation.
* NUMA-aware.
* local-storage-operator.
* LVMS operator.

---

# OpenShift installation tools

* [`openshift-install`](https://github.com/openshift/installer/).
* Paywall: [`assisted-installer`](https://github.com/openshift/assisted-installer).
* Paywall: Advanced Cluster Management.
* *HyperShift*.

---

---
# [HyperShift](https://hypershift-docs.netlify.app/)

* "OpenShift within OpenShift".

---

# [HyperShift](https://hypershift-docs.netlify.app/)

---

# 	[HyperShift](https://hypershift-docs.netlify.app/)

---
# [HyperShift](https://hypershift-docs.netlify.app/)

---

# [HyperShift](https://hypershift-docs.netlify.app/)

* *Management cluster*: a cluster that hosts other clusters.
* *Hosted / guest cluster*.
    * With *hosted control plane* in the *management cluster*.
    * Highly Availabile on request.
* Doesn't have all OpenShift APIs!

---

---

# MicroShift

* Minimal distribution of Kubernetes.
* No OpenShift APIs!
* RPM & yum & systemd services ♥.
* Use case: very small edge devices.
    * 2 GB memory.
    * 2 vCPUs.

```
$ free -h
       total  used  free   shared  buff/cache  available
Mem:   2.7Gi 1.3Gi 130Mi     21Mi       1.5Gi      1.3Gi
```

---

# Summary

* Use Highly Available OpenShift if you can.
    * Consider running smaller clusters with HyperShift instead of large clusters.
* On single node:
    * If you need OpenShift APIs & operators & console: Single Node OpenShift.
    * If you need as small as possible: MicroShift.

---