ahh nothing like causing kernel panics on a friday evening :)

sjorge: it looks like we used to (up to 2015Q3) build the version in wip, there's also one in pkgsrc-joyent of unknown state

uugh

so i'm fairly sure VMware has an extremely unique (*cough*wrong*cough*) interpretation on how various standards deal with deprovisioning space

like SCSI

NVMe

...

which is rather unfortunate

i.e. they advertise features that the standards pretty clearly say 'is set, it means XXX works'

err if set

yet they don't

(side note: DO NOT attempt TRIM on a VMware NVMe device, and absolutely do not enable autotrim unless you like panic loops :P)

(we probably shouldn't panic if a TRIM command errors out on NVMe, but it shouldn't advertise support if it doesn't)

jbk have you checked if it has been reported to vmware?

for those following along: with "rem_drv + add_drv ahci" I was able to load the driver after booting with disable-ahci=true

dmesg looks like this: gist.githubusercontent.com/wiedi/a1…544530063e4fe41a47966f5b1/dmesg.txt

gist.github.com/wiedi/15b7977627857cc8351438cecbdfe4c1 on the two problematic controllers sd only attached on one of the 4 drives each

jperkin: that lines up with the timeframe from my memory, any chance either the older python or the go-carbon/go-carbonapi could be revived?

tsoome_: there's all sorts of things out there about it.. somehow i gather they don't care