Hello
What is the best way to install the linstor gateway?
I see many examples with linux package managers.
Is it possible to do that on either piraeus or the linstor operator for a Kubernetes environment?
No, not really. Those two things (LINSTOR/Piraeus Operator and LINSTOR Gateway) are not intended to be used together in that way.
For Kubernetes, using the LINSTOR Operator to deploy a hyper-converged LINSTOR cluster, or using the LINSTOR Operator to configure the LINSTOR CSI driver to attach to an external LINSTOR cluster, are the intended ways to interface Kubernetes with LINSTOR directly.
However, you could build an HA NFS cluster using LINSTOR Gateway, and then use Kubernetes’ NFS support to “interface” the two systems like that.
This blog describes all three of those approaches at a high level.
So it seems like it would make more sense to be disaggregated with NFS exports available to external customers and K8s applications.
It’s still not clear to me if we can mix Linstor Gateway with an existing Linstor deployment that runs on kube.
Is it possible to let’s say use Pireaus Operator to deploy this specific storage cluster in Kubernetes, to ease management, and configure NFS exports in a HA manner?
We’re trying to get rid of Ganesha essentially.
It’s still not clear to me if we can mix Linstor Gateway with an existing Linstor deployment that runs on kube.
If you’re running LINSTOR Gateway outside of Kubernetes for NFS exported volumes, and using LINSTOR running on Kubernetes for directly attached physical volumes, yes you can mix and match in that way.
Is it possible to let’s say use Pireaus Operator to deploy this specific storage cluster in Kubernetes, to ease management, and configure NFS exports in a HA manner?
You could use the Piraeus Operator to deploy the storage cluster in Kubernetes and then use NFS server pods to export LINSTOR provisioned Kubernetes PVs over NFS as described in this KB article (in case that’s relevant).
But, LINSTOR Gateway relies on DRBD, DRBD Utils, resource-agents, has in-kernel and userland NFS/iSCSI/NVMe-oF requirements, and uses DRBD Reactor as the underlying cluster resource manager (moving NFS exports and VIPs around), which at this time isn’t something we’ve tried to support in Kubernetes. Doesn’t sound impossible, but does sound complex.
1 Like