richlowe: I don't check IRC as often as I used to. Tailscale seems generally status quo. I do rebases as they release new stuff, but I haven't done any real work in it otherwise.

Also I often forget to post an announcements and since I just use the binaries I've been bad about maintaining the packages, but my fork gets pretty regular tags and corresponding binary releases.

github.com/nshalman/tailscale/releases is the place to check.

fwiw I have been pulling those releases and everything seems to function properly

hm... sata disks... 642M repaired, 56,97% done, 0 days 02:22:27 to go

sata or ata?

sata

our ATA driver is painfully slow (which is why anyone using gen 1 hyperv VMs might have thought they were running off tape :P)

WDC-WD4004FZWX and HGST-HUS726T4TALA6L4

wonder if it's maybe the model

(drive model)

my HGST drives i don't think are that slow resilvering

and *knock on wood* have been pretty error free

which is more than i can say of a certain model of SAS drives

this supermicro X10SAE is pretty old

though speaking of that, i'll hopefully have some sata improvements up for review soon..

trying to test those, hit a different bug that made it harder to test :)

:D

resilver/scrub speed depends a lot on the "shape" of data in the pool. Lots of tiny files vs. big sequentially written files.

thats true, of course.

an omnios used filed #16590 (fenix) which seems intriguing, if you're networky

BUG 16590: speed difference between IPv4 & IPv6 on 10Gbit connections (New)

↳ illumos.org/issues/16590

uh. ouch.

I'm happy to provide additional data/tests as directed.

sommerfeld: I feel like at least with the sequential resilvering stuff that's less of an issue these days