Allocation Request Consistency Warnings

You probably reached this page by clicking on a "More information" link in the Nectar dashboard. The purpose of this article is to explain warnings you may receive, and how to resolve them. 

If you need help or advice on quotas and quota consistency, you can raise a Support Ticket. When the requested quotas are modest and sensible, your chances of a quick approval are increased.

The Nectar Research Cloud allocations system allows you to either select a pre-defined Resource Bundle, or request a variety of custom resources. Unfortunately, some of these resources are difficult for a novice user to understand. In particular, some combinations of resources do not make sense. In the past we have had many cases where the inconsistencies were not initially noticed, and the user had to request a further amendment to fix the problem. Since these apparently inconsistent resource combinations sometimes do make sense, we do not forbid them outright. Instead, the allocation request workflows will warn you of the inconsistency. You then have two options:

• You can adjust the custom resources you are requesting to resolve the inconsistency (or select a pre-defined Resource Bundle), and Submit the request again. This will re-check the form, and complete the submission if the consistencies are now resolved.
• You can "Submit ignoring warnings". In this case, the approver will see the same warnings and may query you about them before approving your request.

TABLE OF CONTENTS

• User warnings
  • Requested Service Unit Budget is zero
  • Requested Service Unit Budget seems too small
  • Requested instances > requested VCPUs
  • No VCPUs requested
  • No instances requested
  • Volume storage requested without any instances
  • Database servers requested without any database storage.
  • Database storage requested without any database servers.
  • Database storage requested without object storage for backup.
  • Use of advanced networking requires at least 1 router
  • Use of advanced networking requires at least 1 network
  • Floating IPs require at least 1 network and 1 router
  • Load balancers require at least 1 network and 1 router
  • Openstack clusters require instance quota
  • Openstack clusters require network quota
  • Openstack clusters require floating IP quota
  • Openstack clusters require load balancer quota
  • Openstack clusters require router quota
  • Requests to use the Reservation service require a duration
  • Requests to use the Reservation service require a reservation count
  • Requests for access to enhanced flavors must be explained in the "Justification ..." field
• Approver-only warnings
  • There are no current competitive grants for this request.
  • Non-default RAM > 4GB per core ratio
  • Non-default RAM < 4GB per core ratio
  • -local allocation requests volume storage in
  • -local allocation requests shares in
  • Quota should be authorized by the other site before approving storage quota
  • Problem with your Approver registration: contact Core Services

User warnings

Requested Service Unit Budget is zero

Explanation: A Service Unit (SU) Budget is required to use the Custom Resource Bundle. Please refer to What are Service Units for information on SUs and also to How to estimate a Service Unit Budget. Alternatively, the Allocation Request form itself has guidance on how to do the estimation: look for "Unsure how to estimate a usage budget? Checkout this guide."

Impact: If an Allocation is approved with a zero SU Budget, any running instances that consume SUs will put the allocation "over budget". This will cause the Allocation Expiry procedure to commence.

Remedy: Unless this is a storage only allocation, estimate how many SUs you will need for the requested project period (up to 12 months), and set the Service Unit Budget to that value.

Requested Service Unit Budget seems too small

Explanation: A Service Unit (SU) Budget is required to use the Custom Resource Bundle. Based on the number of VCPUs and project time period requested, we predict that your project will run out of SUs before the end of the time period.

(Our prediction is based on certain assumptions about your future usage patterns.  These assumptions may not be correct, and your actual usage could be more ... or less ... than we predict.)

Impact: If your project runs out of SUs early, project expiry will commence early.

Remedy: Revisit your estimate of how many SUs you will need for the requested project period, and adjust the Service Unit Budget if necessary. Please refer to What are Service Units for information on SUs and also to How to estimate a Service Unit Budget. 

Requested instances > requested VCPUs

Explanation: This warning relates to the number of instances and VCPUs you have requested. When you launch a Nectar instance at least 1 VCPU will be allocated, depending on the flavor that you select. When you request quotas for more instances than VCPUs, some of the instance quota will be unusable.

Impact: This is probably harmless since excess instance quotas don't consume a scarce resource. However, if your intention is to actually run the number of instances implied by the instance quota, then you will need more VCPU quota to achieve this.

Remedy: Either decrease the instance quota or increase the VCPU quota.

No VCPUs requested

Explanation: You have requested quota for 1 or more instances, but no quota for VCPUs.

Impact: Without this quota, you will not be able to launch Nectar instances.

Remedy: Either set the instance quota to zero or increase the VCPU quota.

No instances requested

Explanation: You have requested quota for 1 or more VCPUs, but no quota for instance.

Impact: Without instance quota you will not be able to launch Nectar instances.

Remedy: Either set the VCPU quota to zero or increase the instance quota.

Volume storage requested without any instances

Explanation: Volume storage resources can only be accessed when they are attached to an instance in the same project and the same availability zone. You have not requested instance quota, and therefore won't be able to use the volume storage as intended.

Impact: Without instance quota you will not be able to create an instance to attach volumes to.

Remedy: Either remove the volume storage quota (set it to zero), or add instance and VCPU quotas so that you can create an instance to attach the volume(s) to.

Database servers requested without any database storage.

Explanation: The Nectar database service requires both a server and disk storage to hold the database.

Impact: Without database storage quota you will not be able to create a database server.

Remedy: Either remove the database server quota (set it to zero), or add at least 1 GB of database storage. Note that the maximum supported size for a Nectar database is 1 TB.

Database storage requested without any database servers.

Explanation: The Nectar database service requires both a server and disk storage to hold the database.

Impact: Without database server quota you will not be able to create a database server.

Remedy: Either remove the database storage quota (set it to zero), or add at least 1 database server.

Database storage requested without object storage for backup.

Explanation: The Nectar database service uses object storage for backup, and uses your object storage quota for this.

Impact: If you do not have sufficient object storage quota, you will not be able to use the database backup facility. If your database is going to contain valuable information, it ought to be backed up regularly. Note however that there are other ways to backup a database over the network.

Remedy: Request sufficient object storage quota to hold your anticipated database backups.

Use of advanced networking requires at least 1 router

Explanation: All Nectar-supported use-cases for Neutron advanced networking require at least 1 router and 1 private network. You have not requested router quota, and therefore won't be able to use advanced networking.

Impact: Without router quota you will not be able to use advanced networking.

Remedy: Either remove all advanced networking quotas, or add quotas for at least 1 router and 1 private network.

Use of advanced networking requires at least 1 network

Explanation: All Nectar-supported use-cases for Neutron advanced networking require at least 1 router and 1 private network. You have not requested network quota, and therefore won't be able to use advanced networking.

Impact: Without router quota you will not be able to use advanced networking.

Remedy: Either remove all advanced networking quotas, or add quotas for at least 1 router and 1 private network.

Floating IPs require at least 1 network and 1 router

Explanation: All Nectar-supported use-cases for floating IPs require at least 1 router and 1 private network. You have requested floating IP quota without network and/or router quota, and therefore won't be able to use the floating IPs.

Impact: Without router and private network quota you will not be able to use floating IPs.

Remedy: Either remove all advanced networking quotas, or add quotas for at least 1 router and 1 private network.

Load balancers require at least 1 network and 1 router

Explanation: All Nectar-supported use-cases for load balancers require at least 1 router and 1 private network. You have requested load balancer quota without network and/or router quota, and therefore won't be able to use the load balancer.

Impact: Without router and private network quota you will not be able to use load balancers.

Remedy: Either remove all advanced networking quotas, or add quotas for at least 1 router and 1 private network.

Openstack clusters require instance quota

Explanation: Kubernetes clusters running under the Openstack Cluster Orchestration Engine service use instances to provide the nodes in the cluster. Your project will require sufficient Openstack compute quota (instance and VCPUs) to allow the nodes to be launched. We recommend quota for at least 2 nodes in each cluster to provide redundancy and to facilitate smooth upgrades.

Impact: Without sufficient instance quota, you may not be able to set up a robust / maintainable Kubernetes cluster.

Remedy: We recommend that you increase the instance quota to at least 2 times the number of clusters that you intend to run, and adjust the VCPU quotas accordingly.

Openstack clusters require network quota

Explanation: Each Kubernetes cluster running under the Openstack Cluster Orchestration Engine service makes use of a private network.

Impact: Without sufficient network quota, you will not be able to launch a Kubernetes cluster.

Remedy: Increase your network quota so that you have one network for each cluster.

Openstack clusters require floating IP quota

Explanation: Each Kubernetes cluster running under the Openstack Cluster Orchestration Engine service needs one floating IP address for the control infrastructure, and (probably) at least one more for your services.

Impact: Without sufficient floating IP quota, you will not be able to launch a Kubernetes cluster.

Remedy: Increase your floating IP quota so that Kubernetes can use at least two floating IPs for each cluster.

Openstack clusters require load balancer quota

Explanation: Each Kubernetes cluster running under the Openstack Cluster Orchestration Engine service makes use of a two load balancers internally. If you have externally facing services, they will need at least one additional load balancer.

Impact: Without sufficient load balancer quota, you will not be able to launch a Kubernetes cluster.

Remedy: Increase your load balancer quota so that you have three load balancers for each cluster.

Openstack clusters require router quota

Explanation: Each Kubernetes cluster running under the Openstack Cluster Orchestration Engine service makes use of a router for its private network.

Impact: Without router quota, you will not be able to launch a Kubernetes cluster.

Remedy: Increase your router quota so that you have one router for each cluster.

Requests to use the Reservation service require a duration

Explanation: The Reservation service allows you to create a "booking" to use a resource such as a GPU or big memory node. To prevent problems with overbooking, we limit the total length of each project's active bookings to a given number of days.

Impact: If you don't specify how long you want to be able to create bookings for, you will not be able to create bookings.

Remedy: Set the "days" value to a number of days greater than zero, or clear all of the other fields in the Reservation service section of the form.

Requests to use the Reservation service require a reservation count

Explanation: The Reservation system allows you to create a "booking" to use a resource such as a GPU or big memory node. To prevent problems with overbooking, we limit each project to a given number of active bookings.

Impact: If you don't specify how many simultaneous bookings you want to be able to make, you will not be able to create bookings.

Remedy: Set the "reservations" value to a number greater than zero, or clear all of the other fields in the Reservation service section of the form.

Requests for access to enhanced flavors must be explained in the "Justification ..." field

Explanation: Nectar's enhanced flavors ("CPU Optimised", "RAM Optimised" and "Huge RAM") use considerably more hardware resources (per VCPU) than the standard flavors. When you request them, we need you to provide a justification so that we can assess whether or not your request is technically warranted.

Impact: If you request enhanced flavors without explaining why you need them, your request is liable to be declined.

Remedy: Provide a technical justification for your allocation request in the "Justification and details of your Proposed Cloud Usage" text box. Include an explanation of why you need the enhanced flavors you are requesting.

Approver-only warnings

These warnings should only be shown to Nectar allocation approvers.

There are no current competitive grants for this request.

The normal criteria for approving an allocation as National is that it has at least one grant that is both "national competitive" and current; i.e. hasn't ended by the end of this calendar year. This allocation doesn't satisfy that. You should either uncheck the "National funding" checkbox, OR add an explanation in the "Special RC-NAS approval reasons" field, OR go back to the user and ask them to update the Grant details with a current grant.

The RC-NAS policy allows exceptions to be made, but the reasons for exceptions need to be documented by the approver.

Non-default RAM > 4GB per core ratio

A non-default RAM quota has been set for this allocation. Since this may have happened accidentally, it is advisable to check that this is (still) intended.

Note that setting a memory quota greater than the 4GB per VCPU ratio will only work if the allocation has access to flavors with a higher ratio. Currently none of the public flavors support this.

Non-default RAM < 4GB per core ratio

A non-default RAM quota has been set for this allocation. Since this may have happened accidentally, it is advisable to check that this is (still) intended.

<site>-local allocation requests volume storage in <other-site>

You have classified this allocation as 'local' to your node, but the allocation includes quota for Volume Storage at another Nectar node. The node-local restrictions implemented in July 2019 means that the project will only be able to launch instances at your node, and will be unable to use the Volume Storage quota.

<site>-local allocation requests shares in <other-site>

You have classified this allocation as 'local' to your node, but the allocation includes quota for File Shares at another Nectar node. The node-local restrictions implemented in July 2019 means that the project will only be able to launch instances at your node, and will be unable to use the File Share quota.

Quota should be authorized by the other site before approving <site> storage quota

This just is a reminder that you should always get permission from the site where the storage (volume, fileshare) lives before approving AZ specific storage quota. If you have that permission, this warning can be ignored.

Note that applies to 'national' allocations. (For 'local' allocations, zone-specific storage quotas in another AZ won't be usable, and should not be approved; see above.)

Problem with your Approver registration: contact Core Services

Please raise a Support Ticket about this with Core Services.