Ayush T.inFeatures
GPU instances for inferenceWe run small ML models next to our API. A few GPU instance types (even older cards) in one region would be enough to start.
Open1commentRohan T.inFeatures
Off-site backups to another regionBackups are stored in the same datacenter as the server. We need an option to copy them to a second region for disaster recovery.
In progress4commentsRohipinFeatures
Hourly billing for bare metal serversBare metal is billed monthly only. Hourly billing would let us use it for short load tests and CI runners.
Open0commentsRohan T.inFeatures
Spending alerts per projectSend an email or webhook when a project passes a monthly spend limit, so a runaway autoscaling group does not surprise us.
Planned1commentAmine E.inFeatures
VPC peering between projectsOur staging and production live in different projects. Peering their private networks would let us share one database cluster safely.
Planned2commentsRohipinFeatures
Create a VM from a snapshot in another regionWe would like to launch a copy of a production VM in another region from its snapshot, for testing and migration.
Open0commentsAnatoliy G.inFeatures
API tokens with limited scopesOur CI token can delete any server. We need tokens that are read-only or limited to one project.
In progress2commentsRohan T.inFeatures
Bare metal: IPMI / serial console in the panelWhen a bare metal server does not boot after a kernel update we have no way in. Please add remote console access from the control panel.
Open2commentsvijaykinFeatures
Point-in-time recovery for managed databasesDaily backups are not enough for us. We need to restore a database to a specific minute after a bad migration.
Planned2commentsAbeer M.inFeatures
S3-compatible object storage lifecycle rulesPlease support lifecycle rules on buckets so old logs move to a cheaper class or expire after N days.
In progress2commentsCaped C.inFeatures
Terraform provider: support for firewalls and floating IPsThe Terraform provider covers VMs and volumes but not firewalls and floating IPs, so part of our infrastructure is still managed by hand.
Planned2commentsRohipinFeatures
Kubernetes: cluster autoscaler supportNode pools have a fixed size. Please support the cluster autoscaler so pools grow and shrink with pending pods.
In progress4commentsjoeinFeatures
IPv6 on all VMs by defaultSome of our customers are IPv6-only mobile networks. Please give every VM an IPv6 address without a manual request.
Open0commentsJovie O.inFeatures
Live resize of block volumesGrowing a volume needs a VM restart today. Online resize would let us expand disks without a maintenance window.
Done0commentsAmine E.inFeatures
Managed PostgreSQL 17Only 15 and 16 are available. We want to move to 17 for the new logical replication features.
Done2commentsJovie O.inFeatures
Floating IP reassignment without downtimeWhen we move a floating IP to a standby VM there is a 20-40 second gap. An atomic reassign (or a documented way to pre-warm the target) would let us do real failover.
Planned3commentsCaped C.inFeatures
Firewall rules by tagRight now firewall rules are attached to each VM. Applying rules to all servers with a tag (for example web or db) would make large setups manageable.
Done2commentsjoeinFeatures
Kubernetes 1.31 upgrade pathOur clusters are still on 1.29. A one-click minor version upgrade with a node drain would be great.
Done0comments