I'm doing a POC with a hyper-converged setup on a ESXi 6 host (currently just one host).
So I've got this VM running that presents an iSCSI LUN back to the host. More or less the way any other (non-vSAN) HCI solution works.
Everything is great until I try to reboot the host.
After the bootup the iSCSI target is unavailable (well yeah, the storage VM is not running yet) - this is totally expected.
Well, it should be, but for the ESXi it seemingly isn't - the hostd status is "not running", all management interfaces return error 503, and if I try to start hostd manually, it seemingly just crashes with a coredump in /var/log/vmware.
The only way I've found to resolve the situation is to disable all iSCSI with esxcfg-swiscsi and reboot again (the hostd starts fine after that), and then go through the process of configuring iSCSI again.
I understand that a single host is not a production config, but a full HCI cluster shutdown would basically be the same scenario.
Any ideas what could be wrong with the setup?
The ESXi has latest 6.0 June 2017 patch installed, but the behaviour is the same on the previous patch too.
My ESXi hosts can't properly boot if the configured iSCSI targets are unavailable. This can't be right.
This seems to be a basic enough scenario...