writeup fixes, additional resources, fix alt-texts

yousefi-msft · yousefi-msft · commit 033f01a8393a · 2024-10-04T15:41:02.000-04:00
diff --git a/articles/modeling-simulation-workbench/tutorial-install-slurm.md b/articles/modeling-simulation-workbench/tutorial-install-slurm.md
@@ -1,6 +1,6 @@
 ---
 title: "Tutorial: Install the Slurm Workload Manager on Azure Modeling and Simulation Workbench"
-description: "Tutorial on installing the Slurm Workload Manager in the Azure Modeling and Simulation Workbench"
+description: "Learn how to install the Slurm Workload Manager in the Azure Modeling and Simulation Workbench"
 author: yousefi-msft
 ms.author: yousefi
 ms.service: modeling-simulation-workbench
@@ -10,19 +10,19 @@ ms.date: 10/02/2024
 #CustomerIntent: As an administrator, I want to learn how to install, setup, and configure the Slurm workload manager in the Azure Modeling and Simulation Workbench.
 ---
 
-# Tutorial: Set up the Slurm scheduler in Modeling and Simulation Workbench
+# Tutorial: Install the Slurm workload manager in the Azure Modeling and Simulation Workbench
 
-The [Slurm](https://slurm.schedmd.com/overview.html) Workload Manager is a scheduler used in microelectronics design and other high-performance computing scenarios to manage jobs across clusters. The Modeling and Simulation Workbench can be deployed with a range of high-performance virtual machines (VM). Slurm is to manage scheduling of complex workloads across a set of compute nodes.
+The [Slurm](https://slurm.schedmd.com/overview.html) Workload Manager is a scheduler used in microelectronics design and other high-performance computing scenarios to manage jobs across compute clusters. The Modeling and Simulation Workbench can be deployed with a range of high-performance virtual machines (VM) ideal for large, compute-intensive workloads. Slurm clusters consist of a *controller node*, where the administrator manages, stages, and schedules jobs bound for the *compute nodes*, where the actual workloads are performed. A *node* is simply a part of the cluster, in this case a VM.
 
-The Slurm installation package is already available on all VMs. This tutorial shows you how to install Slurm across your workbench.
+The Slurm installation package is already available on all Modeling and Simulation Workbench Chamber VMs. This tutorial shows you how to create VMs for your Slurm cluster and install Slurm.
 
 In this tutorial, you learn how to:
 
 > [!div class="checklist"]
 >
 > * Create a cluster for Slurm
 > * Create an inventory of VMs
-> * Designate a controller and compute nodes and install Slurm on each
+> * Designate controller and compute nodes and install Slurm on each
 
 If you don’t have a Azure subscription, [create a free account](https://azure.microsoft.com/free/?WT.mc_id=A261C142F).
 
@@ -34,56 +34,60 @@ If you don’t have a Azure subscription, [create a free account](https://azure.
 
 ## Sign in to the Azure portal and navigate to your workbench
 
+If you are not already signed into the Azure portal, go to [https://portal.azure.com](https://portal.azure.com). Navigate to your workbench, then the chamber where you will create your Slurm cluster.
+
 ## Create a cluster for Slurm
 
-Slurm requires one node to serve as the controller and a set of worker nodes. The controller is traditionally a modestly sized VM as it isn't used for workloads and is left deployed between jobs, while the worker nodes are sized for the workload.
+Slurm requires one node to serve as the controller and a set of compute nodes where workloads will execute. The controller is traditionally a modestly sized VM as it isn't used for computational tasks and is left deployed between jobs, while the compute nodes are sized for the workload. Learn about the different VMs available in Modeling and Simulation Workbench on the [VM Offerings page](./concept-vm-offerings.md).
 
 ### Create the Slurm controller node
 
-1. From the chamber overview page, select **Chamber VM** from the **Settings** menu, then either the **+ Create** button at the top or the **Create chamber VM** button in center of the page.
-    :::image type="content" source="media/tutorial-slurm/create-chamber-vm.png" alt-text="Screen shot of chamber VM overview page with Chamber VM in Settings and the create options on the page highlighted by red outlines.":::
+1. From the chamber overview page, select **Chamber VM** from the **Settings** menu, then either the **+ Create** button on action menu along the top or the blue **Create chamber VM** button in center of the page.
+    :::image type="content" source="media/tutorial-slurm/create-chamber-vm.png" alt-text="Screenshot of chamber VM overview page with Chamber VM in Settings and the create options on the page highlighted by red outlines.":::
 1. On the **Create chamber VM** page:
-    * Enter a **Name** for the VM
-    * Select a VM. For the controller, you can select the smallest VM available, which is the D4s_v4.
-    * Leave the **Chamber VM image type** and **Chamber VM count** as the default of "Semiconductor" and 1.
+    * Enter a **Name** for the VM. Select something to indicate that this is the controller node.
+    * Select a VM size. For the controller, you can select the smallest VM available.
+    * Leave the **Chamber VM image type** and **Chamber VM count** as the default of "Semiconductor" and "1".
     * Select **Review + create**.
-    :::image type="content" source="media/tutorial-slurm/configure-create-chamber-vm.png" alt-text="Create chamber VM page with the name and VM size textboxes and the create button highlighted in red outline.":::
+    :::image type="content" source="media/tutorial-slurm/configure-create-chamber-vm.png" alt-text="Screenshot of create chamber VM page with the name and VM size textboxes and the create button highlighted in red outline.":::
 1. After the validation check passes, select the **Create** button.
 
-Once the VM deploys, it's available in the desktop dashboard.
+Once the VM deploys, it's available in the connector desktop dashboard.
 
-### Create the Slurm worker cluster
+### Create the Slurm compute cluster
 
-A *cluster* is a collection of VMs, referred to as *nodes* that perform the actual work, at the direction of the controller. Similar to the steps in creating the controller, return to the **Chamber VM** page and create a cluster. The Modeling and Simulation Workbench makes it easy to create identical copies of a VM.
+A *cluster* is a collection of VMs, individually referred to as *nodes* that perform the actual work. The compute nodes have their workloads dispatched and managed by the controller node. Similar to the steps taken in creating the controller, return to the **Chamber VM** page and create a cluster. The Modeling and Simulation Workbench allows you to create identical VMs.
 
 1. On the **Create chamber VM** page:
-    * Enter a **Name** for the VM cluster. Use a name that identifies these VMs as nodes. For example, include the word "node" in the name.
+    * Enter a **Name** for the VM cluster. Use a name that identifies these VMs as compute nodes. For example, include the word "node" in the name.
     * Select a VM appropriate for the workload. Refer to the [VM Offerings](concept-vm-offerings.md) page for guidance on VM capabilities and sizes.
     * Leave the **Chamber VM image type** as the default of "Semiconductor".
     * In the **Chamber VM count** box, enter the number of nodes required.
     * Select **Review + create**.
 1. After the validation check passes, select the **Create** button.
 
-VMs are deployed in parallel and appear in the dashboard. It isn't customary or even necessary access worker nodes, however you can ssh to worker nodes if needed.
+VMs are deployed in parallel and appear in the dashboard. It isn't typically necessary to access worker nodes, however you can ssh to worker nodes in the same chamber if needed.
 
-### Connect to controller node VM
+### Connect to the controller node
 
-Slurm configuration is performed on the controller node.
+Slurm installation is performed from the controller node.
 
-1. Navigate to the connector. From the **Settings** menu of the chamber, select **Connector**, then the connector that appears in the resource list.
-    :::image type="content" source="media/tutorial-slurm/connector-overview.png" alt-text="Screenshot of connector overview page with Connector in Settings and the connector highlighted with a red rectangle.":::
+1. Navigate to the connector. From the **Settings** menu of the chamber, select **Connector**, then select the sole connector that appears in the resource list.
+    :::image type="content" source="media/tutorial-slurm/connector-overview.png" alt-text="Screenshot of connector overview page with Connector in Settings and the target connector highlighted with a red rectangle.":::
 1. From the connector page, select the **Desktop dashboard** URL.
 1. The desktop dashboard opens. Select your controller VM.
 
 ## Create an inventory of VMs
 
+Slurm installation requires that you have a technical inventory of the compute nodes, as well as their host names.
+
 ### Get a list of available VMs
 
 Configuring Slurm requires an inventory of nodes. From the controller node:
 
 1. Open a terminal in your desktop.
     :::image type="content" source="media/tutorial-slurm/open-terminal.png" alt-text="Screenshot of desktop with terminal button highlighted in red.":::
-1. Execute the following bash script to print a list of all VMs in the chamber. In this example, we have one controller and five nodes. The command prints the IP address in the first column and the hostname in the second.
+1. Execute the following bash script to print a list of all VMs in the chamber. In this example, we have one controller and five nodes. The command prints the IP addresses in the first column and the hostnames in the second.
 
     ```bash
     $ ip=$(hostname -i | cut -d'.' -f1-3)
@@ -96,11 +100,11 @@ Configuring Slurm requires an inventory of nodes. From the controller node:
     10.163.4.9 wrkldvmslurm-nod034b970    
     ```
 
-1. Create a file with just the worker nodes, one host per line and call it *slurm_worker.txt*. For the remaining steps of the tutorial, you'll use this list to configure the compute nodes from your controller. In some steps, the nodes need to be in comma-delimited format. For those instances, we use a Linux shortcut to format without having to create a new file. Create a file of your Slurm worker nodes as your hostnames are different than what's listed in this tutorial even if you created identically named VMs.  
+1. Create a file with just the worker nodes, one host per line and call it *slurm_worker.txt*. For the remaining steps of this tutorial, you'll use this list to configure the compute nodes from your controller. In some steps, the nodes need to be in a comma-delimited format. In those instances, we use a command-line shortcut to format without having to create a new file. Create the file of your Slurm worker nodes as your hostnames will not match the name you entered in the portal.
 
-### Get technical specifications about compute nodes
+### Gather technical specifications about the compute nodes
 
-Assuming that you created all the worker nodes in your cluster using the same VM, choose any of the nodes to retrieve technical information about the worker cluster. Select one of the nodes from the list you created earlier and execute the following commands:
+Assuming that you created all the worker nodes in your cluster using the same VM, choose any node to retrieve technical information about the platform. In this example, we use *head* to grab the first host name in the compute node list:
 
 ```bash
 $ ssh `head -1 ./slurm_worker.txt` lscpu
@@ -135,14 +139,14 @@ NUMA node0 CPU(s):   0-3
 Flags:               fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology cpuid pni pclmulqdq vmx ssse3 fma cx16 pcid sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single tpr_shadow vnmi ept vpid ept_ad fsgsbase bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx512_vnni arch_capabilities
 ```
 
-You'll be asked by the ssh client to verify the ECDSA key fingerprint of the remote machines. Executing the following command returns the technical inventory of the node. Take note of the following parameters:
+You'll be asked by the ssh client to verify the ECDSA key fingerprint of the remote machines. Take note of the following parameters:
 
 * **CPU(s)**
 * **Socket(s)**
 * **Core(s) per socket**
 * **Thread(s) per core**
 
-To discover the **available** memory of a worker node, execute *free* command for any similar compute worker from your controller and note the **available** output from the *free* command:
+Slurm also requires an estimate of available memory on the compute nodes. To obtain the available memory of a worker node, execute the *free* command on any of the compute nodes from your controller and note the **available** memory reported in the output. Again, we use the first worker node in our list using the *head* command.
 
 ```bash
 $ ssh `head -1 ./slurm_worker.txt` free
@@ -153,11 +157,11 @@ Swap:             0           0           0
 
 Note the available memory listed in the "available" column.
 
-## Installation
+## Install Slurm on your cluster
 
-### Install MariaDB
+### Pre-requisite: Install MariaDB
 
-Slurm requires the MySql fork of MariaDB to be installed from the Red Hat repository before it can be installed. Azure maintains a private Red Hat repository mirror and chamber VMs have access to this repository. Install and configure MariaDB. You're asked to create a database password. Create a secure
+Slurm requires the MySql fork of MariaDB to be installed from the Red Hat repository before it can be installed. Azure maintains a private Red Hat repository mirror and chamber VMs have access to this repository. Install and configure MariaDB with the following:
 
 ```bash
 sudo yum install -y mariadb-server
@@ -166,22 +170,20 @@ sudo systemctl enable mariadb
 mysql_secure_installation
 ```
 
-The secure installation script requires further configuration.
+The *mysql_secure_installation* script requires more configuration.
 
 * The default, new installation password isn't set. Hit **Enter** when asked for current password.
-* Enter 'Y' or hit Enter when asked to set root password. Create a new, secure root password for MariaDB, then reenter to confirm. You'll need this password later to set up the Slurm controller in the following step.
+* Enter 'Y' when asked to set root password. Create a new, secure root password for MariaDB, then reenter to confirm. You'll need this password later when you set up the Slurm controller in the following step.
 * Enter 'Y' for the remaining questions for:
   * Reloading privileged tables
   * Removing anonymous users
   * Disabling remote root login
   * Removing tests databases
-  * Reloading privilege tables.
-
-### Install Slurm on the cluster
+  * Reloading privilege tables
 
 ### Install Slurm on the controller
 
-The Modeling and Simulation Workbench provides a setup script to aid installation. It requires the parameters you collected earlier. Replace the placeholder in these example commands with the parameters you collected. Execute these commands on the controller node. For the *\<clusternodes>*, substitute the comma-separated list of hostnames. The examples include a shortcut to do so, reformatting your compute node list in the proper comma-delimited format.
+The Modeling and Simulation Workbench provides a setup script to speed installation. It requires the parameters you collected earlier in this tutorial. Replace the placeholders in these example commands with the parameters you collected. Execute these commands on the controller node. The *\<clusternodes>* placeholder is a comma-separated, no space list of hostnames. The examples include a shortcut to do so, reformatting your compute node list into the proper comma-delimited format to prevent having to create another file. The format of the *sdwChamberSlurm* script is as follows:
 
 ```bash
 sudo /usr/sdw/slurm/sdwChamberSlurm.sh CONTROLLER <databaseSecret> <clusterNodes> <numberOfCpus> <numberOfSockets> <coresPerSocket> <threadsPerCore> <availableMemory>
@@ -208,7 +210,7 @@ Complete!
 ```
 
 > [!TIP]
-> If your installation shows an [ERROR] message, you may have mistyped a parameter. The installation wasn't successful. Review your information and repeat the step.
+> If your installation shows an [ERROR] message, check that you haven't mistyped any parameter. Review your information and repeat the step.
 
 ### Install Slurm on compute nodes
 
@@ -234,7 +236,7 @@ chown -R munge:munge /etc/munge/munge.key
 END
 ```
 
-Using the same file of the node hostnames that you used in the setup, execute the bash script you created. From the controller, execute the following command:
+Using the same file of the node hostnames that you previously used, execute the bash script you just created on the node.
 
 ```bash
 $ for host in `cat ./slurm_nodes.txt`; do ssh $host sudo sh ~/node-munge.sh; done
@@ -272,7 +274,7 @@ rm munge.key
 
 ## Validate installation
 
-To validate that Slurm installed, a Chamber Admin can execute the *sinfo* command on any Slurm node, either the controller or compute node. Run *sinfo* to verify the status.
+To validate that Slurm installed, a Chamber Admin can execute the *sinfo* command on any Slurm node, either the controller or a compute node.
 
 ```bash
 $ sinfo
@@ -342,3 +344,6 @@ chamberSlurmPartition1*    up   infinite      3   idle nodename[1-3]
 * [Slurm Workload Manager Quick Start Administrator Guide](https://slurm.schedmd.com/quickstart_admin.html)
 * [Slurm Workload Manager configuration](https://slurm.schedmd.com/slurm.conf.html)
 * [Slurm Accounting Storage configuration](https://slurm.schedmd.com/slurmdbd.conf.html)
+* [VM Offerings on Modeling and Simulation Workbench](./concept-vm-offerings.md)
+* [Create chamber storage](./how-to-guide-manage-chamber-storage.md)
+* [Create shared storage](./how-to-guide-manage-shared-storage.md)
