richlowe: s2 overlapping is a VTOC relic, which I don't think we do for GPT?

"s2 overlapping" sounds suspiciously like the thing some BSDs do with their disklabels where partition c (the second partition) covers (and specifies) the entire disk

Molnija: Which is why zpool create does all the partitioning for you! No need to partition or fiddle with ESP. It is done for you. We have user friendly tools!

toasterson: Which don't work.

Huh?

I have to install the partition table myself, otherwise I'm at too high of a risk of hitting the swap deadman.

(this is serious, by the way. I'm not just slandering for slander's sake.)

Why is that?

How are you partitioning differently that makes you not hit that deadman?

well, wouldn't using a separate partition for swap obviate the need to go through the ZFS layer, at least?

anyway, this isa conversation I'm done having.

goodnight.

ok. It's a bit odd why one would partition this differently. Since ZFS has it's own deadman. it's more of a you will hit the next deadman if your disks are failing.

if you are in risk of being hit by deadman, it means you are very low on resources (memory). So you would need to either give it more resources or reduce the load.

I am now reading the README in usr/pkg and found an interesting part... (full message at <libera.ems.host/_matrix/media/v3/do…a4bed587a7505198cb91836fc46a3faef0d>)

and?

which is the interesting part?

jclulow: yeah, like I said, I expected it but I didn't expect complaints about it, because it's normal.

toasterson: I wouldn't focus on the partitioning thing, it's a red herring. The real bug is pageout getting stuck when trying to swap to ZFS in low memory systems

toasterson - oh, my client didn't show the link to the rest of the message. That's just telling you what the .mog files are that the gate build leaves around on purpose.

andyf: yeah I just noted the inconsistency with ooce-extra.