Which ESXi 6.* builds create VMFS 6 volumes with 128 sector offsets ?

[continuum]

There is at least one ESXi 6.* build that creates VMFS-volumes that use 128 sectors offset.
See Why does ESXi 6.5 GUI creates VMFS-volumes that are not alligned to 1MB blocks ?
You can check via GUI or
partedUtil getptbl device
or find lines in the vmkernel.log

2018-12-24T12:06:26.924Z INFO     b'gpt\n243201 255 63 3907029168\n1 128 3907029134 AA31E02A400F11DB9590000C2911D1B8 vmfs 0\n'

2018-12-24T12:06:26.924Z DEBUG    Got partGeom: '['1', '128', '3907029134', 'AA31E02A400F11DB9590000C2911D1B8', 'vmfs', '0']'.
Please help to find those build numbers.
I believe we should warn our users not to use those builds.
Or rather to rebuild the affected datastores.
I would also like to see some vmkernel.logs of affected builds.
3 cases with weird behaviour in less than a week is not a large number - it just strikes me as odd and I got interested.
Ulli

[daphnissov]

continuum​

Can you be more specific as to what you're asking here? Is it a call to action for users to perform a test creation VMFS-6 datastore on their build and report results? I'm glad to help here and could test in my lab. Do you know what major builds are affected? 6.0? 6.5? 6.7? All of the above? What would the test procedure look like and what logs/artifacts would you need to confirm your suspicions?

[a_p_]

From what I understood, it was an issue with how the Embedded Host Client created the VMFS partitions.

I can't tell you for sure whether the vCenter Web Client did something similar, but I don't think so.

Looking at my old notes, it seems that the issue has been solved with the Embedded Host Client v1.21.

With that in mind, all datastores which were created using the Embedded Host Client with a version prior to v1.21 are misaligned.

Such versions were included in ESXi 5.5 , ESXi6.0, and ESXi 6.5.

For an overview of versions included in the different builds, see e.g. https://esxi-patches.v-front.de/

As a side note: The start sector has been aligned to 2,048 blocks, but I'm not sure whether the end block has already been aligned with what partedUtil getUsableSectors reports.

André