2022-06-23 - Additional resource quotas on APPUiO Cloud #49
simu
announced in
Announcements
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Overview
Today, we've deployed additional default resource quotas in all projects which should improve the overall stability of the RWX storage cluster which is provided by APPUiO Cloud.
In order to avoid degradation of the RWX storage cluster (which is backed by Rook Ceph), we introduce default resource quotas for all project which restrict the total amount of RWX storage that can be requested in a single project to 25Gi. This will empower us to scale the cluster proactively in response to larger RWX storage requests.
What does this mean for users?
As a user of APPUiO Cloud, you can create RWX PVCs with a total storage request of up to 25Gi per project. If you need more RWX storage in a single project, please contact us using one of the support channels available to you (the available channels vary based on your current support package). This allows us to ensure that the storage cluster has sufficient capacity to accomodoate your larger storage requests ahead of time.
Please note that we've already configured exceptions for some existing projects which use large amounts of RWX storage.
Zone-specific restrictions
Because we provide both RWX and RWO storage using the storage cluster provided by APPUiO CLoud on the Exoscale CH-GVA-2 0 zone, we additionally configure a resource quota which restricts the amount of RWO storage which can be requested in a single project to 25Gi.
Background
Ceph has a safety mechanism which makes the whole storage cluster read-only if a certain utilization threshold is reached. For APPUiO Cloud, we currently configure that threshold at 85% of the storage cluster's capacity.
We get first alerted first about the storage cluster's utilization when it reaches 75%. Given the current cluster size, we expect that allowing users to request up to 25Gi of RWX storage per project will grow storage cluster utilization gradually enough that we're able to react to the automated alerts for cluster utilization before the cluster goes into read-only mode.
Beta Was this translation helpful? Give feedback.
All reactions