article

sidaigle avatar image
sidaigle posted

Genvid SDK 1.31.0

Major Fixes and Changes


Clean Install

We fixed an issue where the error message A setup package is either missing or damaged appeared when installing on computers not already having the Microsoft Visual C++ Build Tools 2015 installed.

Minor Changes and Other Fixes

  • [documentation] Improved search results by decreasing the importance of release and upgrade notes pages.
  • [documentation] Corrected an invalid path in the UT4 cloud documentation which prevented the project from building.
  • [toolbox] Improved the genvid-azure-image tool’s create-images command by having it delete intermediate Image resources and imported VHD blobs after it completes. Use the new option --keep-intermediate-resources to skip this new step and keep these intermediate resources.
  • [toolbox] Improved the genvid-azure-image tool’s create-images command by enabling it to work with private blob storage containers that are accessible from the currently logged-in Azure CLI subscription.
  • [toolbox] Improved genvid-bastion by loading the configuration from the content of the GENVID_BASTION_CONFIG_FOLDER.
  • [toolbox] Added GENVID_BASTION_CONFIG_FOLDER and GENVID_BASTION_JOB_TEMPLATE_FOLDER environment variables to configure where the genvid-bastion command finds its job configuration. The member genvid.toolbox.RuntimeTool.template_dir is now deprecated in favor of GENVID_BASTION_JOB_TEMPLATE_FOLDER.
  • [toolbox] Added the ability to load the configuration from a folder to genvid-sdk load-config-sdk. We created a new folder {ROOTDIR}/local-services and moved the local services.nomad.tmpl from {ROOTDIR}/bastion-services/templates to {ROOTDIR}/local-services/templates.
  • [toolbox] Improved genvid-sdk -c {clusterid} setup by loading the configuration from the folder ROOTDIR/cloud-services/setup.
  • [toolbox] Improved remote cluster initial setup by adding the logs to the cluster configuration.
  • [toolbox] Moved the SDK.update_config method down to the ProjectTool base class to improve availability. The signature of the method has also changed. See the Upgrade Notes for more information.
  • [toolbox] Added genvid-sdk env commands which do the same as their genvid-bastion equivalent commands.
  • [toolbox] Added genvid-sdk cluster-version to display the version of the Cluster API managing the cluster. (Formerly the function of genvid-sdk version, which now returns the version of the Toolbox.)
  • [composed] Fixed an issue where a crash can occur when loosing game connection.
  • [gvencode] Fixed an issue where the system would crash when given a bogus channel or key in ClusterUI.
  • [gvencode] Fixed an issue where an invalid value is passed to the video configuration when no codec is defined in the template.
  • [gvencode] Fixed a network issue where too many calls to Cluster API Health were blocking the connection between composeD an GVencode.
  • [studio] Fixed an issue where the dropdown menu of the video mixer’s source field goes blank when the user edits the name of a scene.
  • [cube-sample] Fixed an issue where the DirectX sample generated warnings when compiling.
  • [bastion-api] Improved Cluster’s Setup Jobs templates by passing the CLUSTER_ID in the rendering environment.
  • [nats] Corrected name of nats-window package to be nats-windows and using the NATS version for tagging it. This also affects the job, task, and log names.
  • [nats] Merged the nats-windows task into the nats job on the cloud.
  • [nats] Fixed an issue where the NATS servers were always reloading the configuration file, possibly leading to crash.
  • [nats] Improved NATS Linux by synchronizing with the NATS Windows version.
  • [cluster-ui] Fixed an issue where the services/leaf/compression field was showing an empty choice in addition to true and false.
  • [cluster-ui, bastion-ui] Fixed an issue where a Page not found error is displayed when the user accesses the documentation link in the UI.

Known Issues

Compose window isn’t available if behind a proxy.

Future versions of Bastion will have a remote server to allow access from different machines. The current version has limited support for this and is likely to encounter some problems if the Bastion host lives behind a cluster firewall (like on an Amazon EC2 machine).

As a workaround, you can create VPN connections to the Bastion machine so the local IP of the machine is accessible from the client machine.

Auto-cut checkbox state is stored locally only.

In Studio, the state of the auto-cut checkbox in Scene Widget is only stored in local storage. (Clearing the browser cache will also clear user input.) Also, if auto-cut is turned “on” before the Studio session ends, it will be set back to “off” by default when you start the Studio session again.

Audio captures microphone or infinitely loops.

When using the WASAPI audio auto-capture mode, every sound going through the machine’s default audio-device gets captured. This includes system alerts and microphone input, as well as your web browser. When test-viewing your game stream using the same local machine running the game, the game will also capture its own audio stream with ~10s of latency. This yields some echo with progressively worse sound quality since it is compressed every time it is streamed.

You can disable audio in 2 ways:

  • Set settings.encode.input.silent to true in your configuration. (The default setting.)
  • Use our raw audio-capture, which is now the default setting in Unity.

Only one application with the Genvid SDK integrated may run at a time if not launched using Nomad.

If an application using the Genvid SDK is launched without Nomad, a default session name is assigned to it. This is useful when testing your application in the Unity engine, Unreal engine, or booting your application manually.

Currently, the Genvid SDK is unable to support multiple applications launched this way unless you change the environment variable GENVID_SESSION_NAME to a unique name. You can also launch your applications via Nomad to avoid this issue.

We will fix this in a future release. .. Story SDK-3770

Twitch and YouTube Synchronization Advisory

During testing, we’ve found some synchronization issues when broadcasting via both Twitch and YouTube. See the Twitch and YouTube sections for details and workarounds for each.

Azure reports requested instance size not available.

On Azure, we are using a compute instance type Standard_NV6 for the game machine. For some accounts, only the promo version of this instance type is available (Standard_NV6_Promo).

When doing the Terraform setup on the Azure module, you may encounter the error:

The requested size for resource 'XXXX-game-1' is currently not available in location 'XXXX'

As a workaround, go to your Terraform Settings page and change the instance_game_type to Standard_NV6_Promo, then go to the Commands page and try Plan apply -> Apply again.

iPhone issues with Twitch.

Since the Twitch API doesn’t return any values for the attributes of the getPlaybackStats() call on iOS devices, you will encounter imprecise or inconsistent synchronization performance.

Streaming to an iOS device with the Standalone player is currently not functional.

Streaming to an iOS device (iPhones and iPads) with the Standalone player is currently not functional. This issue will be fixed in a future release.

Internet connection slowdown can cause the game data to desynchronize when Ingest is running.

When running Studio with Ingest, the game data may desynchronize from the video stream when the Internet connection experiences a slowdown. Once the slowdown clears up, the game data will resynchronize with the video stream.

Links for a local cluster are disabled in RemoteGUI interface.

When using RemoteGUI for a local cluster, opening a tab by clicking a link from the Composition, Encoding, or SDK GUI buttons disables the link (indicated by it turning from blue to red). The link will be re-enabled when you close the corresponding tab.

When changing Availability Zones in Azure, resources need to be destroyed and rebuilt.

When changing AZs in Azure, deleting the public IP might fail due to it not being detached from the virtual machine. This is a bug in the Terraform provider. See the issue on the Hashicorp Github for more information.

As a workaround, manually detach the offending public IPs from the virtual machines in the Azure portal.

Toolbox is not fully compatible with Python 3.9.x.

The Genvid Toolbox is not fully compatible with Python 3.9.x at this time. To avoid problems using our toolbox, we recommend you downgrade Python to version 3.8.x.

genvid-sdk upload-images-sdk command fails to upload the Docker image on Azure Cloud.

Some Docker images can’t be uploaded to Azure Cloud using the genvid-sdk upload-images-sdk command due to large sizes. (You will get an SSL error and a Max retries exceeded message.)

As a workaround:

  1. Opoen the Azure UI.
  2. Upload your image manually.
  3. Run genvid-sdk upload-images-sdk again.

Performance issue when using the Chroma Key Filter.

When using the Chroma Key Filter, we recommend using a c5.4xlarge or better for the encoding machine. Less powerful instances may force the system to accumulate some frames in memory, leading to out-of-memory conditions.

The amount of frames accumulating depends on many factors–the encoding parameters, the source FPS, the amount of variation between the source images content, the richness of their content, etc. For more information on monitoring the memory usage of GVEncode, see the Hashicorp documentation.

Required to manually remove the subdomain NS record when updating the subdomain on an existing AWS cluster.

When creating an AWS cluster using basic_cluster_alb_ssl or minimal_cluster_alb_ssl, the domain and subdomain are provided. When updating an existing cluster’s subdomain, the route53 NS record for the previous subdomain needs to be removed manually from the main domain. This is due to how the NS record is handled within our Terraform.

Unable to delete subnets in cluster basic_cluster_alb_ssl using AWS.

When changing subnet_cidr_block_size in basic_cluster_alb_ssl, decreasing the number of AZs, or reordering existing AZs, the cluster will get stuck trying to delete the subnets.

Reapplying the configuration won’t resolve the issue.

Workaround: delete and recreate the cluster.

Bastion cluster Settings page modal window not closing.

In the Bastion Settings page for a cluster, the Properties modal window for custom tags doesn’t close when you click outside of it.

As a workaround, click on Properties again to close the modal window.

Terraform upgrade leads to ‘Error: Failed to query available provider packages’.

When upgrading an existing cluster to Terraform 0.14, it’s possible to receive Error: Failed to query available provider packages. You can be resolved by deleting the Terraform lock file from %USERPROFILE%.genvidterraformworkdir{cluster name}tf and then reinitializing the cluster.

Limitations

Loading configurations in the incorrect order can overwrite values.

Whether you’re using SSL on a cloud or a local cluster, if you load youtube.sample.hcl before loading the web sample, the web sample will overwrite the value associated with the embed_ssl from the web.hcl file.

Large NTP time offsets will prevent the stream from playing.

If you encounter large NTP time offsets, your stream may stop playing. As a workaround, make sure that the game machine is configured to update its clock automatically. For local configurations, replace the AWS IP address with pool.ntp.org.

See the AWS configuration section for more information.

Studio can’t take inputs of the same control from multiple windows.

If Studio is opened in more than one browser window and inputs are received from different sources to one control, the final value will then bounce between the different inputs instead of updating it to the latest one. For example, in Audio Mixer Widget, if a user changes the slider of Master Gain to 20 db and another user then changes it to 0 db, the slider handle will bounce between 20 db and 0 db instead of being set to 0 db.

Sending input from multiple Studio windows to one control is not recommended until we resolve the issue.

The application doesn’t work properly after Windows comes back from sleep.

Nomad has some difficulties with sleeping jobs, including itself. This is unlikely to get fixed in Nomad, but shouldn’t happen in a production environment. The best workaround at present is to not allow Windows to go to sleep when you are running the service, or simply to restart the service with a genvid-bastion reinstall. (Unfortunately, genvid-bastion restart is not sufficient.)

The SDK clamps framerates under 1.0.

To prevent issues, we raised the lower limit for the framerate to 1.0 for any stream. The SDK will clamp the value to 1.0 FPS if it’s less than 1.0.

Default value for an Azure resource group name may be too long.

If no values are provided for the variable resource_group_name, the value is set to a combination of {bastionid}-{clustername}. This can yield names longer than 64 characters, which is the maximum number of characters allowed for an Azure resource group name.

To avoid this problem, always provide a value of 64 characters or less for the variable resource_group_name.

See the Azure Github for more information on Azure naming conventions.

AWS clusters subnet collision error.

If you reduce the subnet_cidr_block_size and increase the number of AZs for an AWS cluster at the same time, it causes a subnet collision error.

If you encounter this issue, reapply the configuration to resolve it.

No new tags on Azure disk when updating a cluster.

When applying new tags to Azure clusters, tags will not be applied to disk unless the virtual machine instance that owns the disk gets recreated during the apply process.

As a workaround, delete and recreate the cluster to resolve the issue.

Cluster names are limited to 64 characters.

Cluster names are limited to 64 characters due to limitations with the file system in Windows. Things like the system’s username and cluster name factor into the max number of characters of the file directory.

While it is technically possible to go above 64 characters, the results can be unpredictable.

Multiple services restarting in a short period of time can lead to an unresponsive local cluster.

Local bastion/cluster deployments utilize only one Consul agent, which only allows 200 concurrent connections per client. Due to its blocking queries, multiple restarts of all services in a short period of time can lead to reaching the maximum concurrent connections to Consul.

To avoid this issue, we recommend waiting at least 30 seconds between service restarts.

releaserelease notessdk 1.31
10 |600 characters needed characters left characters exceeded

Up to 8 attachments (including images) can be used with a maximum of 1.0 MiB each and 10.0 MiB total.

Article

Contributors

sidaigle contributed to this article

Related Articles