Release 22.02 is a major version extended by three service packs (SP1, SP2 and SP3), that includes scalability, security, usability improvements, new features and bug fixes.
The following section will breakdown the main additions and updates to the Zadara Compute platform (zCompute).
Load Balancer API¶
Added compatibility with AWS fixed-response and redirect action types.
Control over VPC subnet MTU¶
Users can now view and change MTU values per virtual network.
- The relevant networks are:
Edge networks (for MSP administrators only)
The minimum allowed MTU value is 1450 for standard VPC subnets or 1500 for direct subnets or edge networks.
The maximum allowed MTU value depends on the global MTU configured in the cluster (e.g. 9000 or 1500). If the global MTU configured in the cluster is 9000, the maximum allowed MTU value for all virtual networks will be limited to 8950.
If a VM experiences MTU related connectivity issues and it resides on a public network (connected via a route-table to an internet-gateway), then it is recommended to set that network MTU to no more than 1500.
LIMITATION: It is not possible to create a direct subnet in the GUI with desired MTU. However, the MTU can be updated post-creation.
This feature only changes the MTU as reported to the user VMs via DHCP, it does not change the actual MTU of the virtual network. VMs that use MTU value from DHCP will be updated only after the DHCP lease is renewed.
New Load Balancer Service Version¶
In order to support Kuberenetes integration with the managed load balancer service, the functionality and APIs were enhanced to support fixed-response and redirect action types that are compatible with their AWS equivalents.
Fixed-response: Return a custom HTTP response
Redirect: Redirect requests from one URL to another
An example terraform script that create a Kuberenetes cluster that uses zCompute APIs for persistent volumes and a load balancer service type is available at: https://github.com/Neokarm/neokarm-examples/tree/master/k8s.
NK-2921: Getting an Action Failed status when trying to create a listener in the ALB on an cluster upgraded to version 22.02-SP1, but with the previous LBAAS engine.
Workaround: Upgrade the LBAAS engine before creating a new load balancer.
NK-3247: zCompute officially supports Terraform 2.47. Terraform AWS Provider versions higher than 3.33 fail on LBAAS ARN validation.
The new action types are applicable only after a cluster is upgraded to version 22.02.1, and then the load balancer engine is upgraded.
After your zCompute account has been upgraded, all new load balancer instances will be created using the new engine image. The upgrade of already running load balancer instances is opt in. Users should upgrade their existing load balancer in order to receive the new functionality.
The new load balancer target-group types (fixed-response, redirect) in older images are not supported and require upgrade.
Creating DNS VM in any VPC network (from Version 22.02-SP2)¶
In this release, Zadara introduces the option for the user to select the subnet range to create a CoreDNS within a specified subnet range. This will allow accessing the CoreDNS VM from outside the VPC when using a VPN or VPC Peering, and forward DNS requests to it.
By default, the DNS is created with IP address 169.254.64.253, which is only routable inside the VPC.
In this release, the user will have the option to select from which of the VPC subnets to allocate the DNS IP.
Users can modify a VPC and switch the VM DNS Subnet. This operation is mutli-step operation: First disabling the DNS, then changing the VPC network, and finally re-enabling the DNS. The GUI does this sequence automatically, so it is recommended to use the GUI for this operation.
Custom policies support¶
IAM admins and tenant admins can now upload custom symp and AWS policies and then apply them to users, using the new symp APIs or AWS API.
NK-3786: Terraform is not supported at the moment, since it uses a the AWS CreatePolicyVersion API, which is not supported.
Creation of a new policy is only available via APIs and not from the GUI. Once the new policy has been created, the admin can apply it to users using the GUI.
Fixes (Versions 22.02-SP2 and SP3)¶
Versions 22.02-SP2 and SP3 each include a collection of minor fixes.
Notable fixes (Version 22.02-SP1)¶
NK-2374: Fixed sporadic slow response times on volume operations.
NK-2767: Fixed an issue with Windows VM sometimes hanging on migration.
NK-2814: Fixed an issue with no connectivity between VMs with EIPs and other VMs with EIPs on the edge-network subnets. Effectively this meant that VMs between different VPCs could not connect over elastic IPs without Zadara ops intervention.
NK-3144: Fixed issues on local routes delete in route tables, usually leading to UDP packets stop working after removing their elastic IP.
Notable fixes (Version 22.02)¶
There is limited availability of version 22.02, and it should not exist anymore in the field. All changes are listed under version 22.02-SP1.
Release 21.12 is based on the 21.10 version. The version includes various core changes in order to significantly improve the guest operating system performance (primarily Microsoft Windows).
Release 21.10 is a major release which includes scalability, security, usability improvements, new features and bug fixes.
The following section will breakdown the main additions and updates to the Zadara Compute platform (zCompute).
Support multiple subnet ranges for Elastic IPs¶
It is now possible to add multiple non-contiguous subnets to the same Elastic IP pool. Each may have its own gateway. This is aligned with the way site operators purchase public IP ranges. As part of the change Zadara added the following enhancement:
The ability to add multiple subnet ranges to the same Elastic IP pool
Each subnet may have its own default gateway
Virtual Routers will not consume Public IPs anymore, instead a single public IP will serve the entire edge network. This IP is required to access external name servers.
Implicit DNAT Functionality is deprecated and will stop working in one of the next releases. This means: a VM without an EIP will not be able to reach an external destination. If such a flow is required, customers should use the explicit NAT GW service to implement this functionality. Also note that certain OS (e.g. Ubuntu, Windows) configuration for Zeroconf will overlap and interfere with the edge routers subnet address range - VM using DNAT cannot access VM using EIP.
Can’t retire a subnet from the edge network without removing all the EIP from this subnet that is in use by instances. The Edge network DNS servers that are configured for the edge network must be accessible from all edge subnets If multiple edge subnets are configured, then all of them must be configured with the same DNS servers. configuring different DNS servers for different subnets will cause unexpected behavior
These features require additional upgrade procedures for existing VPCs as external connectivity will be temporarily disrupted.
Site operators, for more info please contact Zadara support
CentOS 8 has been added to the zCompute Machine Images Marketplace¶
Following customer requests, we added a link to download CentOS 8 to our marketplace.
Customize the system TLS certificate¶
MSPs that have cloud administration rights can now easily update the web service TLS certificate via the API or UI, without the need to contact Zadara support.
Go to configuration -> Cluster certificates. You’ll see the default self signed certificate To upload the certificate just click the + button and upload the public certificate and private key.
Certificate - The certificate file is the end-user certificate for your domain which you received from the certificate authority. Drop the certificate file in the designated area, -or- Browse to the location.
Private Key - The private key file is the file which contains the private key which matches your certificate. Drop the private key file in the designated area, -or- Browse to the location.
Private Key Passphrase - In case the private key is encrypted by password, the passphrase should be specified in this field, otherwise the field should be omitted.
On certificate creation the cluster will validate that the certificate is not expired. After creating the certificate, it can be installed by selecting it and clicking “Install”.
Known issues: The certificate can’t be updated while there are nodes in maintenance or in failed states and therefore unreachable. There is a GUI validation to prevent certificate change in these cases.
Cluster certificate operation validates nodes are active before starting
Load balancer can be created without the internal Security Group¶
In the past, the system automatically opened the ports to any source. This is very convenient for applications that are service web applications but may be insufficient for tighter security control. Now, upon creation of a Load Balancer instance the user can select whether to keep the mentioned behavior, or like in other clouds, to have full control of the security group by attaching user controlled security groups for the instances. Note: If configured without an Internal Security Group controlled by the platform the access to listener ports is blocked by default. The user must add their own security groups to enable traffic.
The granularity of the protection plan can now be set in minutes¶
Force delete VM snapshot with timer based on snapshot creation¶
The UI now contains a force delete button action to terminate stuck snapshots. Note that when creating a snapshot between different storage pools is an operation that may take a long time. To prevent premature attempts to delete snapshots that may succeed the button is grayed out until a 10 minute timer has been completed.
Clear indication added for incompatible instance sizes¶
Users are now allowed to create an instance type of any size, even when the size is larger than the node’s size (RAM or CPU). However, both the UI and API indicate the instance type is incompatible with the hardware size.
lbaas-manager and Core DNS image updates¶
A new load balancer and core DNS engine images have been created to account for bug fixes that were carried out. Once the first phase of the upgrade is done, the admin needs to go to the upgrade panel and click upgrade all engines (no screenshot attached). Once image upgrade is done and the engines are updated all the users that have LBaaS and core DNS instances will see that updates are ready
DHCP to Pass user configured DNS when default VPC is disabled¶
By default the system created a VPC DNS service which is a forwarding DNS for the user’s configured DNS in the DHCP options while at the same time it’s the authoritative DNS for the VPC entities. Hence the instances are configured to point to the VPC DNS. Before the change, if the user chose not to use VPC DNS (example, when using an Active Directory) then the system did not pass through the user configured name server directly to the VMs as the user expected. Now when the user disables the VPC DNS the system will provide the configured DNS all the way to the VM via DHCP.
Direct-Subnet allocation pools can be updated while attached¶
It is now possible to update allocation pools while the direct network is attached.
Fixed an issue where trying to restore from previous version snapshot failed¶
Due to fields that were updated in later releases, VM snapshots restored from previous versions that didn’t include the os_type field failed. This has now been addressed.
After refresh on the account page, the Enforce MFA appeared unenforced¶
What was fixed
The status now always reflects the actual setting
After a long network disconnect, DNS and LBaaS go into an error state¶
What was fixed Fixed issue where after a long disconnect, managed services fail to refresh state and lose the state. Relevant for CoreDNS (VPC) and LBaaS
Workaround for pre-R8 In order to fix it in the field, identify the VMs supporting the service and restart the service VMs from the symp CLI.
VPSA based VM and Volume DR¶
zCompute can now orchestrate the protection and restore of individual Instances and Volumes between two zCompute systems connected to different VPSAs. The data transfer is done using VPSA volume mirroring between sites, and zCompute has enhanced the protection plans capability to support setting of mirroring on the volumes and to place the required metadata for restoration on an NFS share on the destination VPSA.
Self service password reset¶
Users can reset their passwords directly from the login page using the “Forgot your password?”.
Added support for virtual network appliances within a VPC¶
Some virtual network appliances are either a virtual firewall that requires full control of the traffic coming in/out of the interface, or may be a VPN appliance that has extra IPs representing the client endpoints.
To support the above, there are now two new settings:
Enable/Disable Source/Dest Check - which only controls the L3 IP check. Item 3 above. (This operation can be done by any tenant member)
Enable/Disable Port security. This one can only be set by MSP or Cloud admins.
The options were added to the security groups settings of the interface
Improve VNC console security¶
Starting of R7 version, users must explicitly allow Zadara administrators or MSP administrators to connect to the VM VNC, otherwise only users from the same project can access the VM VNC.
Auto-populate instance name as a DNS record¶
When creating a new instance from the Management user interface, the DNS name field in the VM creation wizard will auto populate the VM name as the DNS name.
VPC deletion wizard¶
A VPC is a construct with many sub resources and/or attached resources. A VPC deletion wizard was added to easily delete a VPC and it’s related resources as long as there are no running workloads in the VPC.
Allow user to set VPC as default VPC¶
The user is now able to change the default VPC by setting a new VPC as the default one.
Additional changes and bug fixes¶
NK-1103 - Clone image operation to allow creation of an image with IDE devices or extra data disks
NK-1082 - Add the option to export filtered event list to CSV from the management user interface
NK-1108 - Delete project now deletes all the default resources created with the project
NK-797 - Fixed an issue where Compute Overview’s “Top VMs” sort worked only with CPU
Version R5 (V21.1.0)¶
NK-227 - Managed resources (like: ENI, EIP, Security Groups,etc…) are now hidden by default and need to explicitly toggle to see them in all views
NK-294 - Added a capability to the GUI to change Direct Network IP addresses allocation pool
NK-470 - An issue if a wrong error when extending volume on a live instance was resolved
NK-502 - VM snapshots are now restored to the requested pool
NK-594 - MFA is now supported using authenticator application on a mobile device (e.g. Google Authenticator), for both Console and CLI login
NK-771, NK-777 - The system now allows admins and tenant_admins to create instance types and also specify the CPU topology over sockets
NK-795 - A network mini-graph was added on the nodes list view
NK-809 - Added clear indication and system alarm to the cloud admin when running out of EIP per IP pool
NK-831 - Prevent deletion of ENI from the GUI
NK-865 - Ability to export filtered lists as CSV file was added to all GUI tables
NK-897 - GUI and Sym/AWS web shell now work even if firewall is forwarding different public port to 443
NK-904 - Snapshots and volumes are now deleted when deleting image from the marketplace
NK-906 - Tenant Admin gets all the permissions of a member
NK-913 - Images and Marketplace Images now have an Operating system field The Operating system is inherited by the Instances created out of these images
NK-955 - The reboot button was removed from the instance view
NK-967 - A new inspector command was added that can be used to validate the MTU of all network interfaces after changing the MTU in the cluster
Version R4 (V5.5.4)¶
Improvements to the VM Creation dialogue¶
NK-733 - Instance creation wizard Steps are according to functionality
Step1 - Compute
Step2 - Storage
Step3 - Network
Step4 - Config
Show the VPC virtual router IP on the direct subnet¶
NK-668 - It is, now, easier to set the static routes in the external routers in case that VMs on non direct VPC subnets need to be accessed via the direct subnets or need to access entities on the direct subnet that are outside of the direct subnet allocation pool
Additional changes & bug fixes¶
NK-234 - An issue, where direct-network was not listed as route in route-table, was fixed
NK-444 - Improved Resource Consumption tile to make it easier to understand
NK-546 - create default VPC operation was added to the GUI
NK-621 - When attaching Elastic IP, if there is only one network available, it’s now selected automatically
NK-648 - The issue of Load Balancer creation with elastic IP that belong to an instance was resolved
NK-661 - Basic event filtering was fixed, and now working as expected
NK-663 - Project delete is now possible even if a NAT gateway still exists in the project but in a deleted state
NK-681 - AWS compatible API now support import from snapshot
NK-703 - CoreDNS VMs removed from the network topology diagram. Clicking the VPC shows if DNS is enabled
NK-729 - When creating key pair for a VM the keys are automatically download, to prevent inaccessible instance
NK-730 - Neokarm Policies were renamed Symp Policies
NK-739 - Cleanup of aging images from the marketplace
NK-800 - The “default” security group is properly displayed even if a load balancer is configured in the same VPC
Version R3 (V5.5.3)¶
New VPC creation wizard¶
A wizard was added to assist you with creating the 3 common topologies of VPCs with private or public subnets
The Wizard can activated from the Quick Start and from the VPCs pane
Elastic Network Interfaces¶
A new screen was added to manage Network Interfaces. Network interfaces may be attached and detached from an instance and may even move between instances or be a route target. Until now, you could only create it implicitly when connecting an instance to a subnet. Now it has its own screen that exposes the full functionality.
A user can now Create/attach/detach/add security groups and update the name of any network interface.
A VM with client tools available on the marketplaces Toolbox v5.5.0. The images contain the Symp CLI, AWS CLI, Python 2.7, Boto3 and Terraform. The instance should be created with an SSH key pair.
Additional changes & bug fixes¶
NK-386 - Added Storage Pool update mechanism
NK-405 - The GUI was fixed to allow subnet name change
NK-422 - The GUI was fixed to enable detach security groups from a network interface
NK-467 - Account members can now generate Access keys
NK-468 - All the networking related tabs are now merged into the networks tab Under instance view. EIP, Security groups and Interfaces operations are done from the same place.