Ceph Image Import Location?

[jamgwert81]

I'm new to trying out HPE VME.  I've recently tested deployment using the new 8.0.8 unified installer with 8.0.9 VMEM on three HPE DL380 G10 that were recently decommissioned from our VMware VSAN environment.  These hosts each contain 21 qty 1.5TB NVME drives and 2 qty 256GB Sata drives.  I've created a RAID1 logical drive for the 256GB drives to load the OS on.  I've configured an LACP lag to be shared by MGMT and Compute.  I have a single 10GB NIC running as an access port for the Storage network.  I've setup this HCI cluster about 6 times over trying to learn this product.  I'm able to deploy this pretty quick now after wiping it and starting over from scratch.  The current setup is using 8 of the 1.5TB drives in each host to create the Ceph shared storage.  I'm able to manage and migrate the VMEM vm after deployment and migrate the disks to mvm-volumes shared storage.  I've been able to import a Microsoft Server 2025 ISO into Virtual images and go thru initial Provisioning deployment of a new VM and prep it with VirtIO drivers and tools.  I've syspreped it and am trying to "capture" (Import an Image) from this VM, but I keep running into an issue doing this.  Everytime I trigger the "Import an Image" from the syspreped VM, the progress takes forever and eventually crashes the VMEM VM.  To me while the image capture is running it appears to eat all the storage on the VMEM VM and eventually runs out of space. 

I've tried unchecking "local" as "Active", and set the flag for "Image Target" on "morpheus-images" Datastore, but it seems everytime I go to "Import an Image" it doesn't seem to be placing it on the "morpehus-images" datastore.

Is this a bug? or am I not understanding what is happening?

Is there are proper way to add Ceph shared storage to "File Shares"?  If so how can I add it?  The path for //var/morpheus/kvm/images is not valid when trying to add to "File Shares".  Is there a way to make the "morpheus-images" Datastore be a target for imported images?  To clarify, adding the ISO or an image from my desktop via the UI appears to be placing them in the correct shared "morpheus-images" Datastore, it's only if I trigger the "Import as Image" from any existing VM already present in VMEM.

I'm new to this product coming over from VMware VSAN so just trying to understand what else I can/should be doing to make basic image capture work so I can continue trying out VM deployments with this.

Thanks for any assistance,

John

[PeterTzvetanov]

Hello jamgwert81,

The images that are getting uploaded and later used for deployments are always send to the VME itself. In the VME it is important 2 things. Either you add over "Infrastructure" -> "Storage" in the sections "Buckets" an S3 storage or in the section "File Shares" an NFSv3/CIFS target and declare this as default. You could also resize the VME main space and then you need to expand it via the cli of the VME as every linux based filesystem with the proper commands. In the second choice with the resize you just need to keep in eye on what is already been uploaded against the how much is left.

With regards  

[jamgwert81]

Peter I appreciate the response.  I understand the S3 Bucket and File Share (NFS/CIFS) can be setup.  I was hoping I could leverage my Ceph storage instead of setting up another external image location. 

I was able to extend the /dev/vda2 default 100GB drive to 500GB.  This did allow me to do the "Import as Image" without crashing the VMEM VM.  The process took FOREVER but finally finished.

I think I'm starting to understand what is happening now. "Import as Image" is not importing it to where I thought ISOs are uploaded "morpheus-images -  /var/morpheus/kvm/images", which is a Ceph location.  It appears that the Ceph storage locations are only available to the 3 hosts that make up the Ceph cluster, and that storage is not attached or available for the VMEM VM. It appears that after the image is captured, once I go to deploy that newly imported image the first time, it then takes FOREVER to push the image from the VMEM vm out to the "morpheus-images -  /var/morpheus/kvm/images" Directory Pool. Finally at that point the image becomes available to the cluster. I'm also confused that if I have a thin provisioned VM that I capture as an image, when go to deploy it to the HVM cluste, it appears to be converted to a think image as it's file size now matches the maximum disk size I initially deployed. My test was a Windows Server 2025 vm with a 40GB disk (thin provisioned). That would be about a 20GB file. After capture and push back to "morpheus-images" on the cluster, the size of the file becomes 40GB. If I'm having to move a 20GB file across the network to capture the first image, and then push a 40GB file back to the "morpheus-images" Directory Pool, I can see why this process is taking a long time to complete. In my setup I have two 10GB NICs in a bonded LACP lag for Compute & Management, and a single 10GB NIC on the storage\Ceph side so I'd think the throughput would be fine for moving large files, but I guess I don't know enough about the VME Manager vm and what speed the virtual NIC is configured for? If that is 1GB then that could explain slow transfer speeds maybe. IDK?

My assumption is the "Import an Image" is a higher level so that it could potentially be deployed to other clusters of any type, which is why it's done this way. Just seems like it takes longer to deploy a basic template that takes seconds to create in VMware. I'm used to VMware where you have an golden VM that can be easily converted to a Template, and then back to a VM for servicing. There is no copying of a file to anything when you convert it from a VM to Template or from a Template to a VM, which is why it's fast. In this software it seems to me you make a copy of the image to a mystery location on the appliance, you then make another copy of that image as you push it back out to the "morpheus-images" shared storage for the cluster.

Seems odd to me you can't leverage the Ceph storage location for this purpose so that is seen by the cluster as well as the VMEM vm. I have over 10TB of storage available on Ceph, why force me to have another location external to that for staging images for deployment on the cluster?

[jamgwert81]

I'm also noticing that when you Delete anything from Virtual Images, it must only be deleting it from the mystery location on the appliance, as the images and ISO files are not getting removed from the "morpheus-images - /var/morpheus/kvm/images" Directory Pool.

I added the ESXi 8 ISO file to Virtual Images, and deployed an Instance with this.  I then deleted the instance and then deleted the ISO file, but the file still remains in the "morpheus-images /var/morpheus/kvm/images" location.  In fact I can see two others I've deleted that still remain, the 80GB file is an attempt at a Windows Server image that I deleted, as is the other 40GB file at the bottom.

Is this a bug?

Deleting the ESXi 8 ISO:

Still present in "morpheus-images - /var/morpheus/kvm/images":