[Docs] Add section on node affinity

docs/conf.py

@@ -128,6 +128,7 @@ def current_version():
     "replacements",
     "linkify",
     "substitution",
+    "myst_nb"
 ]
 myst_url_schemes = ("http", "https", "mailto")
 panels_add_bootstrap_css = False

docs/index.rst

@@ -44,8 +44,8 @@ MLRun provides the following key benefits:
 - **Feature management** – ingestion, preparation, and monitoring
 - **Works anywhere** – your local IDE, multi-cloud, or on-prem
 
-Table Of Content
-----------------
+Table Of Contents
+-------------------
 
 .. toctree::
    :maxdepth: 1
@@ -67,6 +67,7 @@ Table Of Content
    ci-pipeline
    load-from-marketplace
    secrets
+   runtimes/node-affinity
 
 .. toctree::
    :maxdepth: 1

docs/model_monitoring/index.rst

@@ -1,7 +1,7 @@
 .. _model_monitoring:
 
-Model Monitoring Overview
-========================
+Model Monitoring Overview (Beta)
+==================================
 
 MLRun provides a model monitoring service that tracks the performance of models in production.
 See the following sections for details and examples.

docs/model_monitoring/initial-setup-configuration.ipynb

@@ -9,7 +9,7 @@
     }
    },
    "source": [
-    "# Enable Model Monitoring\n",
+    "# Enable Model Monitoring (Beta)\n",
     "To see tracking results, Model Monitoring needs to be enabled in each model.\n",
     "\n",
     "To enable Model Monitoring, include `serving_fn.set_tracking()` in the Model Server.\n",

docs/model_monitoring/model-monitoring-deployment.ipynb

@@ -8,7 +8,7 @@
     }
    },
    "source": [
-    "# Model Monitoring Overview"
+    "# Model Monitoring Overview (Beta)"
    ]
   },
   {

docs/runtimes/node-affinity.md

@@ -0,0 +1,86 @@
+# Node affinity for MLRun jobs and Nuclio functions
+Node affinity can be applied to MLRun and Nuclio functions to determine on which nodes 
+they can be placed. The rules are defined using custom labels on nodes and label selectors. 
+Node affinity allows towards Spot or On Demand groups of nodes.
+
+## On Demand vs Spot 
+
+Amazon Elastic Compute Cloud (Amazon EC2) provides scalable computing capacity in the Amazon Web Services (AWS) Cloud. 
+Using Amazon EC2 eliminates your need to invest in hardware up front, so you can develop and deploy applications faster. 
+
+Using the Iguazio platform you can deploy two different kinds of EC2 instances, on-demand and spot. 
+On-Demand Instances provide full control over the EC2 instance lifecycle. You decide when to launch, stop, hibernate, start, 
+reboot, or terminate it. With spot instances you request EC2 capacity from specific availability zones and is 
+susceptible to spot capacity availability. This is a good choice if you can be flexible about when your applications run 
+and if your applications can be interrupted.
+
+## Stateless and Stateful Applications 
+When deploying your MLRun jobs and Nuclio functions to specific nodes, please take into consideration that on demand 
+nodes are best designed to run stateful applications while spot nodes are best designed to stateless applications. 
+MLRun jobs and Nuclio functions which are stateful, and are assigned to run on spot nodes, may be subject to interruption 
+and will to be designed so that the job/function state will be saved when scaling to zero.
+
+## Node Selector
+Using the **Node Selector** you can assign MLRun jobs and Nuclio functions to specific nodes within the cluster. 
+**Node Selector** is available for all modes of deployment in the platform including the platform UI, 
+command line, and programmable interfaces.
+
+To assign MLRun jobs and Nuclio functions to specific nodes you use the Kubernetes node label 
+`app.iguazio.com/lifecycle` with the values of:
+
+* preemptible – assign to EC2 Spot instances
+
+* non-preemptible – assign to EC2 On Demand instances
+
+```{admonition} Note
+By default Iguazio uses the key:value pair 
+<br>
+```app.iguazio.com/lifecycle = preemptible```
+<br>
+or
+<br>
+```app.iguazio.com/lifecycle = non-preemptible```
+<br>
+to determine spot or on demand nodes.
+```
+
+You can use multiple labels to assign MLRun jobs or Nuclio functions to specific nodes. 
+However, when you use multiple labels a logical `and` is performed on the labels.
+
+```{admonition} Note
+* Do not use node specific labels as this may result in eliminating all possible nodes.
+* When assigning MLRun jobs to Spot instances it is the user’s responsibility 
+  to deal with preempting issues within the running application/function.
+``` 
+
+**To assign an MLRun job to a node:**
+1. From the platform dashboard, press projects in the left menu pane.
+2. Press on a project, and then press Jobs and Workflows.
+3. Press New Job, or select a job from the list of running jobs.
+4. Scroll to and open the Resources pane.
+5. In the **Node Selector** section, press **+**.
+   <br>
+   <br/>
+   <img src="../_static/images/ml_run-job_resources_node_selector.png" width="800"/>
+<br>
+<br/>
+6. Enter a **key:value** pair. For example:
+   <br>
+   <br/>
+   <img src="../_static/images/mlrun_jobs_key_non-preemtible.png" width="800"/>
+   <br>
+   <br/>
+   or
+   <br>
+   <br/>
+   <img src="../_static/images/mlrun_jobs_key_preemtible.png" width="800"/>
+   <br>
+   <br/>
+   When complete press **Run now** or **Schedule for later**.  
+
+**Assign an MLRun job to a node using the SDK:**
+You can use node selection using the SDK by adding the key:value pairs in your Jupyter notebook. 
+For On demand use the following function:
+```func.with_node_selection(node_selector={'app.iguazio.com/lifecycle': ' non-preemptible '})```
+For Spot instances use the following function:
+```func.with_node_selection(node_selector={'app.iguazio.com/lifecycle': ' preemptible '})```