[illumos-gate] 17301 libdiskstatus: dprintf() cleanup -- Hans Rosenfeld <rosenfeld⊙gho>

While I work on reconditioning a Dell R640 for OmniOS ZFS duties, I'm seeing prices for Optane DIMMs have come WAY down on the used market. A couple of those as SLOG devices in my head seem ideal, stupid low latency and high bandwidth, but can Illumos SEE them?

Thank you @Woodstock !

keith is probably rolling in his alpacas

Including those on 17301's siblings (I think you have a few left still) will grease the RTI wheels for sure.

@richlowe Is that about Optane?

(ENOCONTEXT)

danmcd: got it

one of them still lacks reviews, #17308 (code.illumos.org/c/illumos-gate/+/4152)

→ CODE REVIEW 4152: 17308 zfs: dprintf() cleanup (NEW) | illumos.org/issues/17308

Kurlon: They are in the DIMM form factor or the PCIe form factor?

DIMM form factor

You would not be able to use them as an SCM slog device.

Figured as much, oh well, standard optane nvme slogs it is then.

You could enable the mode where the memory controller transparently is faulting data back and forth, but we did not end up doing the DIMM enabling work for a variety of reasons.

And that mode wouldn't be too useful as a slog. Yeah, just kind of how it shook out.

Yeah, between costs, and PCIe channels exploding, I can see why the tech never really took off. Interesting to see VMWare now going all in on NVMe memory tiering NOW...

I thought they were all in on suing people who used to like them?

No suing yet, but the threat letter volume is huge.

I'm very glad we ditched VMware for XCP-ng. So much nicer to work with a friendly vendor.

weren't there issues booting illumos on xcp-ng? did those ever get sorted out?

I haven't picked our next move yet, though for one system I'm locked to VMWare weather I like it or can afford it or not... :/

jbk, I haven't tried booting OmniOS under XCP-ng for a very long time. My need for it went away.

I don't know if the problems were ever resolved. It was an OmniOS bug but I don't remember the details.

what's the RTI status of 17391 limit man links to proto area ?

(my change for illumos.org/issues/17416 adds a bunch of links in man3c)

→ BUG 17416: Want more xlocale.h functions for C++ locale support (In Progress) | code.illumos.org/c/illumos-gate/+/4242

sommerfeld: one problem, though probably not the problem, is sqlite_mprintf() sign-extends the "posix id" in update_cache_pid2sid()

since the query is built the same way each time, I suspect it's actually fully reversible and works. much to my chagrin.

sommerfeld: an insight is that even when your test works, idmapd caching does not.

sommerfeld: the cut down find/idmap dump I showed will showed the mapped posix IDs don't stick

richlowe: does it not stick across a idmapd restart, or not stick at all?

(assuming cache.db is left alone across the restart)

at all

It's going to be something very simple when I find it, or even the sign-extension I just said was symmetrical

I am a mechanical engineer interested in Solarish systems. I do CAD design for living, and just have read this interesting article about the future if mechanucal CAD programs, which points some really painful issues with the CAD programs. Interestingly the author is a cofounder and product officer of Oxide Computer Company. This was surprising.

This is the article if you are interested: lobste.rs/s/dir3rd/new_era_for_mechanical_cad