Quan Nguyen

Bug Fixes​

• Workflow configuration page crashes intermittently
• pw workflows ls returning 404
• Workflows fallback to GitHub CDN if API is throttled
• Deleted users can cause marketplace not to load

Documentation Changes​

• Update restart workspace modal explaining workflows are canceled

Enhancements and Features​

Clarified workflow interactions with user workspace restarts​

When the user workspace is restarted, all running workflows are immediately canceled. This has always been the case, since workflows run inside the workspace, but the UI did not previously reflect this correctly. Workflows could show a canceled status while individual steps appeared to still be running.

All steps are now correctly marked as skipped, and a clear cancellation reason is available by hovering over the workflow status. In addition, sessions created by a workflow will no longer be automatically re-established after a workspace restart, as this was not intended behavior.

CLI transfers now show progress bar​

Transfers to cloud buckets using the PW CLI will now show a progress bar for each file.

User workspace status indicators​

When the user workspace is starting, there will now be a dismissible banner present on the top of every page. Additionally, you will not be able to navigate to tunnel sessions or click execute on workflows until your workspace is up.

Bug Fixes​

• Prevent executing workflows while workspace is down
• Home page widgets don't drop exactly where they are dropped
• Groups should not be required for mass import
• Cost page doesn't auto select a group
• Buckets not creating
• Snapshotting a cluster's root disk fails but shows provisioned
• Cloud node metrics still show even after node is deleted
• Instance type list not using cached list
• Workflows don't have scheduler type for cloud clusters
• Workflow logs can grow too large
• Updating marketplace items not working
• Node metrics chart tooltip overflows if there's a lot of nodes

Documentation Changes​

• Update windows CLI install instructions

Bug Fixes​

• Recreating a shared bucket with the same name does not show the newly created bucket to shared users
• Unable to set images to latest in admin panel
• Form slider styling
• Saving cluster config while running wipe out some cluster variables
• Favorite marketplace workflows have out of order inputs
• Can't update cluster partition config on private network deployment
• Kubernetes secret/configmap widget not showing results
• Session slug is not populated properly
• Session text unreadable on dark mode
• Unknown Error when opening see all instances page while node box is empty
• Cannot attach or detach workspace mount while cluster is running
• Kubernetes user workspaces created with wrong configmap name
• Snapshots not finding group is causing page to reload

Enhancements and Features​

Markdown table support​

You can now use extended markdown such as tables anywhere you could previously use markdown.

Google Assured Workloads support for Cloud Networks​

Cloud account networks now work with Google Cloud Assured Workloads

Refreshed workflow pages design​

We've refreshed the workflow pages, making them more similar to the design of the compute resource and storage pages.

Workflows can access scheduler type on input form​

Workflows can now access scheduler type via expressions:

jobs:
  echo_on_remote:
    steps:
      - name: Print scheduler type
        run: |
          echo ${{ inputs.cluster.schedulerType }}
'on':
  execute:
    inputs:
      cluster:
        type: compute-clusters
      slurmPartition:
        type: slurm-partitions
        hidden: ${{ inputs.cluster.schedulerType !== 'slurm' }}
        optional: true

Install nvidia-container-toolkit in cluster images​

New base images for clusters will have nvidia-container-toolkit pre-installed.

Support storage type on workflow form​

The storage type on the workflow builder now exists.

Bug Fixes​

• Os-login not disabled on GCP instances
• User login times not localized
• Add org fields wrong color
• Sessions widget doesn't auto refresh
• Clicking reset to default theme not working
• Unable to select image on new partition
• Storage permission "View & Edit" incorrectly named
• Errors on snapshot page not showing error message
• Spaces in remote workflow properties break the workflow
• Sparse checkout not saving
• Kubernetes home page widget always shows error
• Instance dropdown not showing "all sizes" link
• Stop button should be disabled if cluster is provisioning
• Placeholder not allowed on dropdown type
• Restore snapshot on disk page not working
• Localize helm updated at times
• Sessions page has unthemed text
• Workflow slider type always says it must be false
• Tooltips go under table headers
• Blank message pops up when deleting cluster
• Existing clusters without sftp disconnect during healthcheck
• Kubernetes config data overlaps parent panel

Other Changes​

• Localize run canceled message

Enhancements and Features​

Add partition level slurm suspend timeout setting​

Cluster partition now includes Suspend Time. Setting the value to -1 will disable node suspension.

Kubernetes execute shell now allows selecting container​

If a pod has more than a single container, a modal will pop up when clicking on execute shell allowing you to choose which container to exec into.

Bug Fixes​

• Built-in configurations table has wrong number of columns
• Tooltip in wrong place for sparse checkout list
• Explorer sidebar theme blends into app sidebar on dark mode
• Explorer view empty if only a single bucket is shared
• Unable to restart workspace if its not already running
• Workflow inputs with defaults shows required and fields without explicit label have no label

Documentation Changes​

• Examples unreadable on dark mode

Enhancements and Features​

Organization admins can revoke user login sessions​

Admins and organization administrators can now revoke user login sessions from the organization user page. When revoking sessions, there is an option to also immediately revoke login sessions on cloud resources. This will log the user out of all cloud systems managed by the platform, and kill any processes that they are running..

GitHub input fields now load repo/branch options​

Bug Fixes​

• Sessioned google bucket deletion failing
• Missing fields not being highlighted on workflow execute
• Kubernetes workload creation times not localized
• Log viewer uses light color scheme on dark mode
• Text on IPs table hard to read on dark mode
• Customized sidebar lost when changing languages
• Home page table headers overlap tooltips
• High levels of zoom cause cluster widget to scroll
• Lustre and github icons dark mode theming
• Sessions widget requires scrolling at high zoom
• Org admin role cannot see allocations
• Tooltip in wrong place for sparse checkout list

Bug Fixes​

• Resize the terminal via the maximize/restore button doesn't properly redraw text
• Can't click "here" to config workload widget if not in customizing mode
• Session open in new tab doesnt use slug

Bug Fixes​

• Theme can enable "use org logo" even if organization does not have logo
• Cost estimate shows NaN with unknown costs
• Marketplace edit page has duplicate item overview
• Azure controller-nat unable to start compute nodes if cluster was started without partition
• Inconsistent theming on password inputs
• Publish and share icons are not the same everywhere
• Link based sessions do not open

Enhancements and Features​

Marketplace item page enhancements​

Marketplace item pages now include a details pane that provides key information before you fork or favorite an item.

• Clusters and storage resources: Display provisioning details.
• Remote workflows: Show links to files in the connected repository.

Additionally, workflows in the marketplace now include new tabs for YAML preview and run graph visualization.

Bug Fixes​

• Pw CLI permissions on cluster nodes too open
• Temporary workflow directory permissions too open
• Incorrect validation error if session field is empty
• Unable to use ai chats shared by other users
• Forking remote workflow from marketplace has wrong type
• Workflow configs not deleting
• Able to provision aws base infras without selecting a networking mode
• Fsx logs never get past collecting variables
• Workflow runs that fail validation create invalid runs

Documentation Changes​

• Document workflow named steps
• Document writing outputs

Enhancements and Features​

Allow list of public buckets​

The Parallel Works CLI can now be used to list public buckets on AWS, Azure, and Google Cloud Platform.

PW CLI cluster filtering​

When using the Parallel Works CLI, you can now filter clusters returned with the new --owned and --status flags.

Examples

# List clusters owned by the user "Jake.Thayne" 
pw cluster ls pw://Jake.Thayne

# List only clusters you own
pw cluster ls --owned

# List only provisioned clusters
pw cluster ls --status=on

Workspace storage usage shown on settings page​

You can now see the storage for your user workspace from your account page.

Bug Fixes​

• LDAP edit page never loads
• Open in new tab shows up on breadcrumbs incorrectly
• Pw ssh cannot connect to cloud cluster with private ips

Documentation Changes​

• Add pw ssh proxy command examples for port forward

Enhancements and Features​

Pw ssh proxy command​

Now can use pw ssh --proxy-command as the ProxyCommand option for both SSH and SCP Examples: Run as a ProxyCommand for SSH

ssh -o ProxyCommand="pw ssh --proxy-command %h" user@pw://Jake.Thayne/my-resource
ssh -o ProxyCommand="pw ssh --proxy-command %h" user@workspace

Run as a ProxyCommand for SCP

scp -O -o ProxyCommand="pw ssh --proxy-command %h" file.txt user@pw://Jake.Thayne/my-resource:/home/user/file.txt
scp -O -o ProxyCommand="pw ssh --proxy-command %h" file.txt user@workspace:/home/user/file.txt

Note: only scp -O works at the moment (forcing scp to use sftp protocol)

Allow pw CLI to pull from public buckets​

The Parallel Works CLI can now copy files from public buckets even if those buckets are not managed by the ACTIVATE platform. This works on AWS, Azure, and Google Cloud Platform.

Bug Fixes​

• Cloud account group permissions manager always shows own organization's groups
• Allow organization theme to override legacy theme
• Show error when app target not defined in remote apps
• Cannot create instances on cloud networks shared with organization only
• Empty google bucket on deletion

Enhancements and Features​

Pw ssh command​

• Users can now initiate ssh connection to resources using pw cli

Examples

• pw ssh workspace to connect to user workspace
• pw ssh pw://namespace/cluster to connect to cluster in a namespace (username)
• pw ssh pw://namespace/cluster echo hello to execute a remote command over ssh

Bug Fixes​

• Can not update or delete workflow configurations
• Sessions widget shows an error if there are any OpenAI sessions
• Should not show copyable public ip and snapshot button when instance is not ready
• Deleted buckets aren't automatically removed from attached storages configuration

Documentation Changes​

• Pw ssh cli docs

Bug Fixes​

• Can't tell which resource type is selected
• Breadcrumbs on editor page do not show

Bug Fixes​

• Alphabetize tags on docs
• Sessions list doesn't refresh automatically
• Organization new user defaults not saving
• Show all buckets on buckets list page
• AI Chat links open in a new window
• Should not show disk size (GB) field when attaching a disk to AWS cluster

Enhancements and Features​

Auto summarize and archive cost data​

Billing now auto summarizes and archives old cost data based on platform or organization policy Archive Cost Data

Recursive support for pw bucket commands​

pw bucket commands now support recursive copies with the -r flag.

Sessionless AWS and Azure buckets​

New AWS and Azure buckets are now sessionless, meaning after provisioning a bucket, destroying it will also delete the configuration of it. AWS bucket provision statuses are also improved.

Slurm job monitor dashboard generally available​

The multi-cluster job monitor is now generally available, and can be found on the sidebar.

Bug Fixes​

• Global/job workflow env variables not being set correctly
• Workflow cancellations marked as error if cancelled during file sync
• Notifications in inbox have different heights
• Cost dashboard crashes when filters are reset
• ML workspaces does not show correct groups
• Existing clusters agent stop reporting after token expired
• Existing clusters intermittently fail to copy pw cli
• Compute nodes can't mount nfs exported disk when using root snapshot as image
• Unable to remove tags from storage configuration
• Theme issues in HPC Connect
• Notifications menu not themed correctly
• Remove edit page from buckets

Documentation Changes​

• Improve workflow docs
• Add workspace example

Enhancements and Features​

Customize language settings​

By default, we use the language you have set in your browser to display the Parallel Works ACTIVATE application. Now, you can override the browser settings in your account to a specific language. We will then prefer this setting in all cases when displaying localized interfaces.

Localized cluster page​

The cluster details page will now be available in supported languages.

Bug Fixes​

• No hover color on add resource page
• Cannot destroy cluster when starting using power button

Enhancements and Features​

Support google compute node disk attachment​

Users can now attach inline disks to google cluster partitions; each node within the partition will receive the same set of disks

Home page localization​

We recently added localization into four languages:

• Chinese
• Japanese
• Korean
• Spanish

Initially, only the navigation components, the sidebar and the command palette, were localized. Now we've additionally added localization for the home page. We will be continuing these efforts, focusing on the highest touch areas of the ACTIVATE platform first.

Bug Fixes​

• Incorrect values shown for region on storage pages
• Organization variables should not be required to run workflows
• Cannot save clusters while running
• Versioning shows on bucket details page for providers that do not support versioning
• Inline config appearing when attaching disk to running cluster
• Style of cost estimator on storage pages is inconsistent with clusters

Enhancements and Features​

Archive cost data policy​

• Added new policy to set how long high-resolution cost data will be stored before being summarized

Bug Fixes​

• Slurm jobs table not themed correctly
• Localize times on instances monitor page
• Nodes table on clusters page is missing corners
• Filters are not the same size on resource list
• Fork icon on market item page incorrect size
• Dark theme styling on org users page
• Workflow variables not manageable by org admins
• Workflow outputs not working between jobs
• Slurm job notifications are not sent

Enhancements and Features​

Workflow ssh block will now use ssh config file if available​

Enhancements and Features​

Internationalized and themed terminal​

The terminal interface is now internationalized based on your language preferences. Additionally font size and theme is customizable. The themes "light", "dark", "rose pine", and "dracula" are available.

Allow changing workspace mounts while cluster is on​

You can now update workspace mounts while a cluster is running.

Manage helm charts​

Users can now manage helm charts in kubernetes namespaces they have access to.

Control/command s support for saving yaml in workflow builder​

You can now use cmd+s or ctrl+s to save changes on the build tab when building workflows.

Users can choose application theme​

Users can set now set their own application theme from the account settings page.

Bug Fixes​

• Cannot type into terminal font size box
• Homepage widget loaders not centered
• Cluster slider does not use element text color
• Some sessions don't open in a new tab correctly
• Tunnel does not redirect to the correct platform domain
• Unable to snapshot azure cluster when zone is not specified
• Can create OpenAI sessions from update-session workflow step

Documentation Changes​

• Fix broken images on bucket pages

Enhancements and Features​

Support aws inline disk definition for cluster partitions​

Users can now attach inline disks to cluster partitions; each node within the partition will receive the same set of disks

Bug Fixes​

• Improve error message for misconfigured marketplace remote workflows
• Sliders are unstyled when theme element is unset
• Cannot list machine learning workspaces that use base infrastructure
• Terminal tabs panel overlaps SSH connection menu

Enhancements and Features​

Use platform bucket name as actual csp bucket name​

Users can now customize the bucket name when provisioning a new bucket on AWS and Google Cloud Platform. This was already available on Azure.

Add customize page​

This adds a customize page to the organization settings. On this page, you can find overviews of new user defaults, sidebar settings, and the organization theme. Clicking on one of the cards on the page will let you customize the settings.

Bug Fixes​

• Titlecase is not applied when username is returned from OIDC
• Double clicking on the terminal actions maximizes terminal

Enhancements and Features​

Prevent account creation option and TitleCase options for OIDC​

Organizations with OIDC configured can now disable automatic account creation. We also added an option to automatically convert usernames to TitleCase.

Cluster agent push update​

There is now a button on clusters allow you to initiate an agent update. This will update the PW Agent and CLI to the latest version.

Bug Fixes​

• Idle agent connections are never closed
• Unable to delete machine learning workspaces that failed to provision
• Resources creation fail with unknown errors
• Header breaks with expression in text field
• AZML storage capacities not showing up after save
• Editing and discarding multiple times breaks form
• Marketplace remote workflows now fetch description from remote markdown
• Snapshots become unassociated when users are deleted

Bug Fixes​

• Cannot use API key directly in Postman or similar tools due to trailing colon

Bug Fixes​

• Unknown error on cluster access save

Bug Fixes​

• Placement group visibility was based on the wrong input condition
• Maximizing the terminal doesn't cover breadcrumbs

Bug Fixes​

• Snapshot option being shown on cluster form when attaching existing disk
• Terminal text sometimes gets squished together

Enhancements and Features​

Add storage details page​

There is now a details page on all storages which gives information about the configuration of the storage.

Enhancements and Features​

Support attaching inline disks to clusters with auto zone​

You can now attach inline disks to google clusters which have an automatically-selected zone

Add email-based user search in admin panel​

The search box on the admin panel user's page now supports filtering with email addresses in addition to usernames.

Platform and organization policies​

Introduced platform (admin only) and organization policies, which are rules and configurations that are applied to all users and resources at either platform or organizations level. Platform policies take precedence over organization policies.

Bug Fixes​

• Snapshots list doesn't load if there are any unassociated snapshots
• Custom subdomain session redirects to incorrect platform
• Linked based sessions don't work without type being specified
• Access drawer intermittently saves the wrong permissions
• Organization theme not shown on first login

Enhancements and Features​

Add delete session button to sessions home page widget​

You can delete sessions from the home page now.

Vertical terminal tabs​

We've changed the terminal tabs to a vertical view. This is in preparation for upcoming support for splitting a terminal tab into multiple terminals.

Menu bar for terminal​

A menu bar was added to the terminal to open terminals in new tabs and to clear the terminal.

Bug Fixes​

• Favorite shortcut on resource list has wrong label and icon
• Disable hyper threading whenever the aws instance type supports it
• Ensure RStudio authentication works with subdomain routing
• Clear field when options list is empty
• Shared clusters can be stopped and resumed by users with login access
• Double render terminal
• App ui flashes before redirecting to login page
• Cannot upload thumbnail for existing and legacy clusters
• JSON field not removed when dependency is switched to false
• Users with certain characters in their names cannot use sagemaker
• Connecting through jumphost had incorrect ordering of ssh auth methods
• Cluster ip address editable while cluster is running
• Marketplace publish dropdown shows null as initial option
• Fails to autoreconnect to servers with banner

Documentation Changes​

• Update API documentation layout
• Add system/user-centric to API route descriptions

Bug Fixes​

• Ldap login fails in some server setting combinations
• Cannot select snapshot for inline disks attached to azure clusters
• Organization theme not showing on initial login
• No error logged when compute nodes fail due to insufficient capacity
• Kubernetes clusters do not connect

Other Changes​

• Rename "Not Applied" to "Automatic" in Azure zone selection

Enhancements and Features​

Allow setting organization themes​

Organization admins can now set themes for organizations. This will allow you to change the color of different UI elements for users in an organization.

Bug Fixes​

• Intermittent issue where users sometimes cannot get bucket credentials
• Unable to select an AZ for a disk

Enhancements and Features​

Copy azure snapshot to another region​

Users can now copy Azure snapshots to another region from within ACTIVATE.

Autofill defaults when clicking add workspace mounts​

We now autofill default values when adding a workspace mount to make it easier to understand what values to use for each input box.

Bug Fixes​

• Legacy cluster resource account dropdown not populating correctly
• Add oidc icon to login providers
• Incorrect zone shown when creating new Azure cluster
• Workflow graph reset button is black
• Forking a remote workflow does not fork the latest version
• Workflow run page intermittently crashes
• Using snapshots for azure compute nodes
• Sparse checkout not showing
• Connecting to multiple existing clusters with "autoreconnect" can take a long time
• Daily scale down not running if set to more than 1 day

Documentation Changes​

• Sorted aws s3 policies
• Add docs for multizone

Bug Fixes​

• Can't select google capacity reservation

Enhancements and Features​

Apps are now determined based on YAML definition​

Local Apps are now determined based on if the app property is found inside of the workflow YAML definition.

You can make a workflow show up on the apps list by simply adding the following to any workflow definition:

app:

Copy AWS snapshots to another region​

Users can now copy AWS snapshots to another region from within ACTIVATE.

Retroactively assign reservation cost to groups​

After capacity reservations are added or updated from the organization settings, the reservation's past cost data will also be assigned to the newly designated group.

Custom AI chat providers​

You can now connect any OpenAI-compatible API directly to the ACTIVATE interface using your own authentication tokens.

This lets you chat with models from your chosen provider while keeping full control over your usage and data. Chat history is saved securely within our platform, so you can easily continue previous conversations or reference them later.

Support includes in strings​

Workflows can now check if a value includes a string with the following syntax

 ${{ inputs.name includes 'foo' }}

Bug Fixes​

• Connecting to an existing cluster immediately after saving uses previous configuration
• Potential deletion of incorrect route53 records
• Shortcuts not centered
• Some text fields don't have borders

Bug Fixes​

• Cannot renew proxy certificates
• Skip job if all of its steps are skipped
• Clusters could be named with invalid characters in some cases

Enhancements and Features​

Kubernetes cost management plugin provisioning status UI​

You can now track the deployment progress of the kubernetes cost management plugin via the UI.

Bug Fixes​

• AWS compute nodes mounting EFS filesystems were not mounting with same-zone mount points
• Footer overlaps login panel on some screen sizes
• Cli bucket ls command not working

Enhancements and Features​

Savable app configurations​

Now you can select a configuration from the app launcher to launch an app with a predefined set of inputs.

Target tags are a new concept which correspond to the tags on computing resources, marking a compute resource as compatible with a given app configuration.

Configurations can be saved by filling out the workflow form and saving it as a configuration, or can be provided by the workflow developer as built-in configuration. This is done in the "Build" tab of a workflow by editing the yaml. These built-in configurations can be selected from dropdown and are also shown in the configurations tab, just like manually saved configurations.

Allow org member to get group allocation used​

Organization members can now user teams endpoint to get their group's allocation used

Display helm charts​

Can now see the helm charts deployed into the selected namespace for the Kubernetes cluster

Helm chart details and rollback​

You can now see the details for helm charts installed in your kubernetes clusters. Additionally, you can rollback helm revisions directly from the ACTIVATE interface.

Bug Fixes​

• Disconnected workspace message on terminal disappears too quickly
• Attach session slug to subdomain
• Underscores in subdomain cause problems for some workflows
• Prevent trying to start existing cluster if user workspace is down
• UseTLS not working in sessions with custom domains
• Gray section is showed when clicking follow on log viewer
• Unable to start cluster is user workspace is down
• Sessions table has link to workflow on shared sessions
• Pressing back after logging in brings you back to the login page
• Missing zones on AWS eu-north-1 region
• User workspace terminal intermittently disconnects

Documentation Changes​

• Update base infra docs for AWS explaining networking modes

Enhancements and Features​

AI Chat conversation history​

Conversations with AI Chat providers are now saved, so you can continue previous chats at a later time.

AWS public IP networking mode​

This introduces a new networking mode when provisioning organizational base infrastructure. This provides compute nodes with their own public IP addresses. This is useful in cases where you want egress charges to be billed to the same group as a cluster instead of being absorbed as an organizational cost. The security group that is attached to compute nodes will block inbound access from internet but will allow traffic from other compute nodes.

Bug Fixes​

• No groups on kube or nfs commands, add alias for kube (k8s, kubernetes)
• Cannot edit kubernetes cluster
• Scrollbar missing on logs viewer, line numbers sometimes cut off

Bug Fixes​

• Aws efs and fsx provision fails when there are no private subnets in base infrastructure

Enhancements and Features​

Add gpu filter to instance selector​

Users can now filter instances by whether or not they have a GPU

Add billing alerts​

Admin user can now subscribe to billing alerts

Support nvidia.com/gpu in kubernetes resource quotas​

Organization administrators can now define resource quotas for GPU resources labeled with nvidia.com/gpu

List pods on workload details page​

On the details page for a workload, there will be a list of all the pods associated to that workload with links to the individual pod details page.

Allow resources to be provisioned in eu-central-1 region on aws​

Organization admins can now provision base infrastructure in the eu-central-1 region on AWS.

Ability to see the logs for pods​

Users can now monitor and download logs for actively running Kubernetes pods

Bug Fixes​

• Tooltips are hidden under text box for kubernetes cluster creation page
• Ephemeral storage has off status in attached storage panel
• Extend Azure Managed Lustre timeout to 60 minutes
• Cannot provision azure managed lustre before moving the size slider

Documentation Changes​

• Add instructions to create a manual session
• Add blog post for Q1 of 2025

Enhancements and Features​

Add architecture and networking feature filter to instance selector​

You can now filter instances by architecture and networking feature on instances list when selecting an instance for a cluster.

Ability to share resource with any group within organization​

We've found that the sharing feature had limited usefulness because previously you had to be a member of both groups to share a resource. Now, you will be able to share with any group inside of the same organization, regardless of your personal group membership.

Bug Fixes​

• Cannot create netapp ontap storages due to name validation failure

Bug Fixes​

• Can't scroll quick connect when you have many clusters
• Removed escape sequences showing up on terminals after using interactive program like vim
• User notifications settings by default are misleading
• Logs from older sessions of clusters that share the same name can appear in newer cluster sessions
• Compute resource job monitor shows jobs no longer in the queue

Bug Fixes​

• Log viewer expand does not work on provision status modal
• Bootstrap can fail if link command fails
• Remove inconsistent session number padding
• Instance selector has double scrollbar
• Marketplace descriptions can overflow their container
• Existing cluster save while connected needs confirmation
• Tooltips sometimes under textboxes and improve kubernetes error messages

Enhancements and Features​

Node detail page​

A new Nodes tab has been added on a clusters information page. This tab will list all current pending, active, and failed nodes. Clicking on a node from the list will show details for that node.

Add windows support to CLI​

You can now install the PW CLI on Windows.

Bug Fixes​

• Rerunning a workflow run would not populate inputs
• Workflow display name is not used in running notification

Documentation Changes​

• Add a page for Creating OS Snapshots

Enhancements and Features​

Add display name to storage creation page​

User can now input display name in storage creation page

Add ability for popover drawers to be resized​

• Added resize handle to JSON tab for v3 clusters

Bug Fixes​

• Cannot add/remove ssh private keys for existing clusters

Documentation Changes​

• Merge Buckets pages into one page

Bug Fixes​

• Typo on admin/groups page
• Storage deletion time is missing
• Kube/config file is not updated after making changes to the configuration on the platform
• Clicking workflow runs ID does not navigate to the correct workflow run page

Enhancements and Features​

Integrate sessions with Kubernetes provider​

Users can now create sessions with the Kubernetes provider, supporting pods, services, and deployments

View provision status from cluster list page​

Cluster provision status can now be opened from cluster list page

Add autoselect option to workflow input forms​

You can now add the autoselect property to a dropdown field when building a workflow. This will make it so the first option is automatically selected.

Bug Fixes​

• On-prem clusters intermittently fail to disconnect
• Workflow with subdomain does not work for marketplace workflow
• Storage deletion logs are missing for some types
• Connecting to an existing cluster which shares home directory with another cluster causes disconnect
• Workflows do not populate default values when running as app
• Workspace mounts shown incorrectly on cluster details page
• Existing cluster reconnect notification did not show correct error

Documentation Changes​

• Update Starting & Stopping Clusters

Enhancements and Features​

Add reason for cluster automated delete request​

Cluster automated delete request now has reason displayed

Add impersonated by to provision status record​

Admin users now can see Impersonated By on resource provision and destroy records

Bug Fixes​

• Add session button disappears
• Responsive search bar on smaller resolutions
• Attached storages panel doesn't include filesystems

Enhancements and Features​

Subdomains for sessions​

Users can now use a custom subdomain when creating a session

Command to list kubernetes clusters available​

The command pw kube ls is now available to list available kubernetes clusters. In addition to basic listing, it also supports output format flags -o json and -o table.

Manually create link type sessions​

You can now manually create sessions of the "link" type. Link sessions are simply links to other applications or services from within the platform. In the past, a workflow would've created these in case a session is managed outside of the ACTIVATE platform.

Provision status enhancement for clusters​

Added more details to cluster provision status records. You can now click on individual provision status lines to see more details.

Can now stop cluster controllers​

When attempting to shut down a cluster you will now be given the option to stop the controller. While stopped, costs will be reduced to only the cluster's storage. When you try to start a stopped cluster, you will be prompted to restart the cluster.

Allow existing clusters to reconnect​

Auto Reconnect option has been added to existing cluster options. If enabled we will try to automatically connect to the remote vm. You must first connect manually by pressing the power button.

Bug Fixes​

• Snapshot name not showing if name is too long
• Unknown error when sharing manual sessions
• Cost estimate table showing empty values
• Cluster image dropdown doesn't filter out failed snapshots
• Unable to update cluster definition when editing via json view
• Can't update azure cluster with controller in NAT mode
• Extra recipients in allocation thresholds are optional
• Delete configuration tooltip says turn off instead of destroy

Documentation Changes​

• Additional updates for new cluster UI
• Remove Dashboard page from the user guide
• Aws ec2 permission is missing register image policy

Enhancements and Features​

Add password to workflow schema​

Workflows now have a password field which can be used by workflow developers to hide text as its being entered.

Cluster provisioning failures will show a banner​

If a cluster fails to provision, a red banner will be shown at the top of the page indicating that the cluster should be destroyed.

Workflow run page improvements​

You can now cancel a workflow run and view running sessions from a workflow's Run page

Bug Fixes​

• Orphan alerts should not be sent for cross platforms
• Auto close toasts, limit to 3 toasts on screen at once
• Cannot select old sessions in provision status viewer
• Tooltip hide under the toggle after recursive renders

Enhancements and Features​

Image dropdown now shows required disk size​

We now show the required disk size you will need to use an image in the image selection dropdown.

Snapshot disks that were configured inline​

You can now create snapshots from disks that were configured inline in the cluster configuration page. You may do so from the clusters "Sessions" page in the attached storages table.

Show workload metrics on kubernetes workload pages​

You can now see CPU usage, Memory Usage, and Storage usage for kubernetes workloads (statefulsets, deployments, etc). These are viewable from the workload pages. These metrics show combined usage for all pods managed by that resource. We will have individual pod metrics in an upcoming release.

Platform administrators receive deletion failure alerts​

Platform administrators will now receive an alert whenever the platform detects that a resource deletion might not have been processed correctly.

Ability to edit YAML directly from kubernetes definition​

You can now edit Kubernetes resources by modifying them directly on the YAML editor.

Azure AI chat in public preview​

You can now provision and chat with OpenAI models on Azure. To enable this feature preview, use the feature preview menu.

Cloud cluster observability​

Cloud clusters will now have CPU, Memory, and Root Disk charts at the top of the cluster details page.

Bug Fixes​

• Load from market place clears out some fields
• Disable snapshot button while request is pending
• Snapshot names not showing on cluster details page
• Publish from action bar does not work
• Cannot snapshot disks provisioned on clusters using auto zone selection
• Connect via ssh button not disabled while cluster starting
• Not found if clicking edit from jobs page
• Cost dashboard print not working
• Cannot start shared cluster with admin/write access
• Azure managed lustre is not marked as ready
• Some secrets don't show eye icon
• Search bar crashes with some special characters
• Disabled menu items have incorrect styling
• Broken link to docs on new workflow
• Can't start google nodes of a3 family
• Traffic light does not turn off when there is no requested nodes
• Unable to make snapshot of snapshot
• Intermittent google key creation failures
• Use attempts for gaia delete jobs instead of expireTime to avoid context deadline exceed
• Creation page now just says name instead of resource name
• Provision status stucks at collecting variables and delete requested when rabbitmq connection fails
• Various session bug fixes

Documentation Changes​

• Update user guide with new cluster configuration layout
• Add information about user defaults
• Add new Managing Orgs pages

Enhancements and Features​

Existing provider now has password support​

In addition to keyboard-interactive, the existing cluster provider can now support older password ssh connections

Improved information display for kubernetes​

Kubernetes workload pages now use a more readable format to show relevant information.

Bug Fixes​

• Capacity reservation table moves around on refresh
• Various marketplace/workflow bugs
• Allow clicking outside the HPC Connect modal to close it after connection
• Node runtime is incorrect when it has been more than 1 day

Bug Fixes​

• Openstack doesn't have "Latest" image option
• Cannot create azure ip address
• Traffic light can crash compute module when debugging nodes provisioning
• Recreating a cluster with the same name as a previously deleted one retained the configuration of the deleted cluster.
• PW CLI not authenticated inside of workflows

Documentation Changes​

• Add video walkthrough to Navigating the Platform

Enhancements and Features​

Storage runtime alerts​

You can now receive runtime alerts for storage. You may choose to turn on/off runtime alert in storage's properties tab. This will send you a notification after your storage has been running longer than the interval specified. You can configure if you receive the notifications via the platform inbox or via email through your account settings.

Bug Fixes​

• Prepend parallelworks: to groups prefix in kubernetes

Enhancements and Features​

Allow setting latest image​

Admins can now set the latest image directly from the admin panel

Update bootstrap scripts while cluster is running​

Users can now update cluster bootstrap script and its toggles (controller/compute) while cluster is running

Kubernetes available in feature preview​

Organizations can now add kubernetes clusters and share them with specific groups. Users in those groups will be able to use the pw CLI to add a context to their local kubeconfig, allowing users to run kubectl commands. Kubernetes resources are now visible in the platform for users that have access to the cluster.

Bug Fixes​

• Fields not pre-populated when editing capacity reservations
• /etc/profile.d/pw.sh variables are not quoted
• Azure ML workspace fails to provision due to incorrect domain
• Traffic light doesnt display correctly for next gen clusters

Enhancements and Features​

Capacity Reservation description field​

You can now set descriptions on Capacity Reservations

Ability to manually create sessions​

You can now manually create sessions without having to start a workflow. Try this out by clicking the Add session button on the Sessions page.

Bug Fixes​

• Netapp ontap cannot be deleted
• Move optional session options to advanced
• Start ssh connection button is not full width
• Workflow executing message shows even if theres a validation error
• Azure ML studio is only soft deleted
• No confirmation on marketplace item deletion
• Space in reservation id causes node provision failure
• Remote workflows cannot be forked from marketplace

Enhancements and Features​

Azure on-demand capacity reservations billing​

Azure on-demand capacity reservations reservations are now billed to the group added in the organization settings page

Bug Fixes​

• Remove session tags for sessionless cloud infrastructure
• Path being removed from shells

Enhancements and Features​

Add custom nfs configuration to cluster attached storages​

You can now specify custom NFS mounts via the cluster configuration form. This applies to cloud clusters and OpenStack clusters.

Bug Fixes​

• Ai chat does not skip searching buckets when no bucket attached
• Provisioned snapshot never marked ready
• Detach custom nfs fails on validation
• Cluster definition shows inline disk options when selecting persistent disk
• Validate ontap name is provided and flatten API structure
• Ebs and efs cost estimation does not reflect throughput and provisioned iops
• Group shutdown fails

Bug Fixes​

• Marketplace fork from add page redirects to incorrect page

Enhancements and Features​

Search bar​

You can now search the platform and documentation via a new search bar at the top of the UI. Open this with cmd + k on Mac OS or ctrl + k on Windows.

Allow using AWS and Google cluster controllers as nat instances​

Use Controllers as NAT Gateways option has been added on AWS/Google infrastructure configurations. Enabling this option will make it so all traffic from compute nodes are proxied through the controller. All compute nodes will share the same IP as the controller node. This will enable tracking egress costs from compute nodes.

Provision public IP addresses on Azure​

You can now provision public IP addresses on Azure independently of other compute resources. After provisioning a public IP it can then be assigned to a compute cluster controller node.

Ai chat provision status​

Improved provisioning status logs shown when creating an AI Chat resource

Bug Fixes​

• Crashes when extra email is not found as platform user
• Aws snapshots show provisioned before actually ready
• Platform image with space and parenthesis could not be deleted
• Better error messages on aws disk provisioning error
• Page crashes if feature previews not loaded in time
• Able to save google and azure lustre without selecting instance type

Enhancements and Features​

Organization specified default sidebars​

Organizations can now define a default sidebar for new users. Users that have not customized their sidebar will automatically use the organization's default settings. If the organization does not specify a customized sidebar then all sidebar items will be visible by default.

Optional dns server​

Adds an optional DNS service to platform deployments which can resolve slurm cluster hostnames if needed

Add cli list for cluster, app, workflow, and session​

Add Email and Notify Allocation Alerts options to organization settings​

Added Cost Alert notification type and user preferences.

OpenStack clusters can assign floating ip to compute nodes​

Bug Fixes​

• Snapshots show provisioned earlier than expected

Documentation Changes​

• Merge apps and workflows pages

Enhancements and Features​

Update openstack cluster definition page​

Added cluster cost estimation for openstack

Admin-only debug mode for cloud clusters​

Platform administrators can now toggle Debug Mode in cluster Advanced Settings section.

When turned on, users can ssh into the node without waiting for it to be ready using ssh root@<nodeip>. Can be done from the user workspace or user's personal computer (if public key is added in the platform). Node clean up will not happen if there is any error in provisioning.

Bug Fixes​

• Sidebar not always highlighting active item
• Apps shared to marketplace were not labeled as apps

Documentation Changes​

• Customizing the platform
• Improved API docs

Enhancements and Features​

Support capacity block type for aws reservation​

Supports aws capacity block reservation type on next gen clusters

Enhancements and Features​

Add flavor pricing table to openstack infra page​

• Add flavor pricing table to openstack infra page. This page will display the synced flavors from openstack project.

Add openstack to realtime​

Realtime now processes openstack data.

Bug Fixes​

• Able to to start a workflow without selecting required field while using app.target as the default value
• Environment variables not available on workflow cleanup steps
• New ssh key not showing after creation
• Starting a session on existing clusters would fail under certain cases

Documentation Changes​

• Add page for managing static IP addresses
• IP page corrections/improvements
• SSH key update

Bug Fixes​

• Updated workflow yaml not reflected with page navigation
• Alphabetize IP address options
• Sessions started on shared clusters not working
• Icon not showing for sessions started from github workflow
• Root snapshots not in image list

Enhancements and Features​

Nfs list, attach-storage commands​

• Adds pw nfs ls command to the cli
  • This lists all of the nfs storages on a user's account
• Adds pw cluster attach-storage command to the cli
  • Required Flags: --cluster-uri, --storage-uri, --mount-point
  • This will attach the given storage to the given cluster at the given mountpoint
  • Does the same action as attaching the storage on the cluster definition page
  • The user will be able to see the attached storage on the cluster definition page after running the command sucessfully

nfs ls and cluster attach-storage commands​

You can now use the new pw nfs ls and pw cluster attach-storage commands from an authenticated cluster session.

Sessions home page module​

Sessions are now available as a home page module which can be added using the customize button on the home page.

Bug Fixes​

• Justify sidebar settings to the right

Enhancements and Features​

Add customizable sidebar​

Users can now customize the items that appear on their sidebar. In a future update, organizations will also be able to customize the default sidebars for users in their organization.

Azure OpenAI service available in feature preview​

Can now provision Azure OpenAI resources from within the platform and directly chat via the UI.

Bug Fixes​

• Custom thumbnails not showing
• Cannot expand logs
• Text overlaps on add resource page

Bug Fixes​

• Elastic image dropdown is empty for legacy google clusters

Enhancements and Features​

Show running workflow steps while session is pending​

On a session's page, current running workflow steps will be shown while the session is pending.

Change default layout for home page​

We've updated the default layout of the home page. If you have not customized your home page, you will automatically be moved to the new default layout which shows:

• favorite workflows
• recent workflow runs
• favorite compute resources

To customize your layout, use the customize button at the top right of the home page. Note that users with a custom layout are not affected by this change.

Bug Fixes​

• Marketplace workflows with dots in name can now create sessions
• External auth UI doesn't fit compactly
• Saving cluster configuration gives an error if storage is attached
• Controller ip substituted incorrectly in workflow inputs
• Clusters can share one static ip address
• Whitespace not showing correctly in changelogs
• Ephemeral storage provision status does not update when persistent storage also attached
• Api key and token generation has slightly different UI

Documentation Changes​

• Document unique naming system

Enhancements and Features​

Update controller configuration for Amazon Linux support​

Support bootstrapping clusters that use the Amazon Linux 2 image

Bug Fixes​

• Unify border style
• Button is invisible on cloud platform

Enhancements and Features​

Allow adding images through platform​

Platform level slurm cluster images can be managed through the admin panel.

Allow setting default session names​

You can now set default session names on workflows:

session:
  mySession:
    prompt-for-name:
      default: "defaultname"

Allow job environment variables on workflows​

Allows setting environment variables on a job level. Environment variables set at a step level will override the environment variables set at a job level.

jobs:
  main:
    env:
      foo: "bar"
    steps:
      - run: echo ${foo} # bar
      - run: echo ${foo} # baz
         env: 
          foo: "baz" # will override env variable set on job level

Stop workflow run on session stop​

Stopping a session will now also stop the workflow run that started the session and any sessions associated with the workflow run.

Improve workflow update-session logs​

Logs and input validation for updating session has been improved.

Bug Fixes​

• Move health check below controller node configuration in provision status
• Cannot delete some old storage sessions
• Sessions not deleting after cancelling workflow run
• Cleanup step running when step is skipped due to if being false
• Show message about user workspace starting on workflows page

Enhancements and Features​

Support client_secret_basic token endpoint auth method​

There is now a dropdown on the OIDC auth method to allow choosing either client_secret_basic or client_secret_post. These auth methods are used for the platform to authenticate to the token endpoint after a successful OIDC user authentication.

Organization admins now can specify scopes for oidc​

Scopes can now be specified when adding or updating an OIDC auth method in organization settings.

Bug Fixes​

• Azure NetApp Files does not work on newly provisioned infrastructures
• Root disk snapshots times out before finishing
• Escape oidc redirect uris
• Validate group names meet constraints
• Random crash on home page
• Workflow inputs.sh has wrong format when using objects
• Workflow cleanup steps not running after canceling
• Workflow sessions not found if contains url params