I am about to tear out the rest of my hair. I am hoping someone here has seen this behavior and can point me to a solution. I simply have tried everything I can think of, an nothing is making sense.
Summary: I have a Samsung 840 SSD, good for about 700GB, installed along with other SSDs in a box running ESXi 6.7
When I try to create a new VMFS6 datastore on this SSD, the process goes without issue. I have tried this under regular ESXi 6.7 UI, under ssh using command-line tools, and under vCenter 6.7 -- regardless, creating the datastore works just fine, as expected.
However, creating a new VM using that DS, or migrating an existing VM to that new DS, fails. EVERY TIME. The errors vary, they never seem to be the same, but it is always a combination of:
- Unable to load configuration file '/vmfs/volumes/5ea9ef7c-23eb5578-3738-000af7a1d9e1/VM1/VM1.vmx'
- An error occurred while creating a temporary dictionary file: Error.
- Failed - The file is too large
No matter what though, the VM fails to be created/migrated. Google and other searches have yielded no help for this problem.
Of note: I can browse the new datastore, and upload files. I can create folders. I can ssh to the host and create files. I can create a text file using the text editor vi and read the file using cat. I can delete the files. In other words, the datastore seems to be perfectly happy. It just won't take virtual machines!
When I have tried creating the DS under command line, I have used partedUtil to create the partition (starting at sector 2048 as recommended for VMFS6) and then vmkfstools for creating the VMFS store. Currently, the SSD has a GPT format. I have tried taking the SSD into a Linux machine and clearing out all partitions there, and trying again under ESXi. Same result: the datastore is created but VMs can't be written to it. I have tried leaving a little space at the front of the partition and some space at the end of the partition. I did this because examining the other SSD's partitions, they seemed to have some free space before and after. I figured I would try that. No dice.
I truly am baffled. I don't know what could possibly be wrong with this datastore in terms on creating VMs or VMs being migrated to it, but being totally fine for writing files and directories directly or using the Browsing UI.
Here is a quick movie I made which shows the failed process, and you can also see I am displaying the info for the device in ESXi under ssh.
I am also attaching an image with the same information being shown in the ssh console.
If you would like any more details, I can provide them. If anyone can help, I will be forever grateful.