Client Namespace

[Visnusaran Murugan]

Hello All,

I have a 2 Node 4 Target minimal setup to try out BeeGFS in our environment. Configured with Connection Auth File and everything seems to work as expected. Requirement is to horizontally scale based on load. Data is going to be short lived and wouldn't need replication to start with. 

Question:

1. clients also have identical conn auth file to communicate with server. Not sure if client identification is possible in BeeGFS so as to restrict what it sees or mounts. We have a single directory per user in our current NFS setup and I'd like to expose just a subdirectory for each client with BeeGFS too. But not able to get relevant docs on that regard. 

It would be great if can get a pointer to docs if available or a suggestion on an alternative approach. 

[web...@netapp.com]

Hello Visnusaran,

It's not possible to do exactly what you're asking (mount a single directory of a BeeGFS file system instead of the whole file system). However, bind mounts can be used to achieve something similar, depending on your exact requirements. My suggestion is to:

• Create a directory on the BeeGFS file system for each node as you planned.
• Mount the entire BeeGFS file system to each node in a location only the root or other privileged user can access. 
• Create a bind mount on each node from the appropriate directory in that location to whatever location you were originally planning to expose the file system on. (For example, bind mount /mnt/private/beegfs/node1 -> /mnt/beegfs.)

I have seen bind mounts used in combination for a number of purposes (including something like yours). In fact, we use a similar approach in the BeeGFS CSI Driver to isolate individual BeeGFS directories to individual containers. There is at least one location in the BeeGFS documentation that mentions bind mounting you can use as an example, though it is for a very specific use case. It might be easier to just consult more general Linux bind mounting documentation online.

Regards,

Eric Weber

Software Engineer

E/EF Series Solutions

NetApp

[Visnusaran Murugan]

Thanks Eric,

We use a mix of virtual Machines and Pods based on the cloud provider we deploy work loads to. We do not have access to the host machine from there the K8Pods are created. 

I'll try bind mount. Since workloads run under an unprivileged user, it should be good enough.

[Allan Thrower]

As your workloads are cloud based. Have you considered using BeeOND to provide a filesystem for each workload?

Allan

--
You received this message because you are subscribed to the Google Groups "beegfs-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fhgfs-user+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fhgfs-user/fbbd4805-9888-4fb2-8fd0-64d2110caa52n%40googlegroups.com.