zfsgreg-hydrogen commented on pull request zfsonlinux/zfs#4329 - ZFS Encryption by tcaputi <https://github.com/zfsonlinux/zfs/pull/4329#issuecomment-232253880>
zfstcaputi commented on pull request zfsonlinux/zfs#4329 - ZFS Encryption by tcaputi <https://github.com/zfsonlinux/zfs/pull/4329#issuecomment-232254307>
perfinionspeaking of 10G and such, why the $^#@$^@ do 802.11ac routers come with only a gigabit ethernet port?
perfinion5GHz 11ac is easily >1G why dont htey put a 10G port on the wired side?
rlaagerBecause 10G is expensive.
perfinionits not like an 802.11ac wave2 router is cheap either tho
damexperfinion, i dare you to throughput 1gbit (full duplex) through 802.11ac
damexin real world
perfinionwell sure
perfinionbut it comes close
perfinionyou can do one directional close enough to 1gig for streaming, full duplex isnt often required in a house
raxxy10g getting more important as time goes on
lblume2.5g too, it seems.
bundergbit sure, 10ge not so much, but that's just my opinion
bunderwe won't need 10ge until we get 100mbit+ ftth
lblumeAnd also replace all those pesky old cables by newer, stiffer ones.
jzaws/stiffer/stiffer & more expensive
Lalufumany AC APs have two 1GE ports
lblumejzaw: That comes without saying :) And I'm sure there must be some specially expensive audiophile versions!
jzawdirectional too to switch -->
lblumeAwww, yes, that is SO IMPORTANT.
zfskimhc72 commented on issue zfsonlinux/zfs#3075 - BUG: Bad page state in process flush-zfs-1 <https://github.com/zfsonlinux/zfs/issues/3075#issuecomment-232294656>
Lalufuperfinion: also many APs use PoE, and I'm not sure 10G PoE does exist
perfinionhmmm thats a fair point
perfinionbut not even like a linksys or whatever has wired 10G yet, but theyve been doing 11ac for ages
bunderour AP's at work are PoE, and have 2 ports, but they're cisco and the other port is a console port :P
perfinionyeah 9600 baud isnt fast enough for 10G either :P
bunderi wouldn't be surprised if it could be used as a real port with a firmware replacement, but i doubt there is a dd-wrt equivalent for commercial cisco stuff heh
bunderoh wait, i keep forgetting belkin owns linksys now
perfinionwait what? did cisco sell them off?
lblumeI'd be quite surprised that a firmware update can turn an RS232 interface into an Ethernet one.
bunder2013 perfinion
perfiniondamn how'd i miss that
bunderi know, i keep forgetting too
bundermy ~10 year old WRT is still kicking
bunderhaven't had to buy a new one
bunderlblume: is it rs232 though on the inside? i can't see them wiring a rj45 into rs232
lblumeCisco has been doing just that forver.
bunderalthough that might explain why i needed to use their cable to configure them
lblumeNot the only ones, either.
perfinionrs232 is already on basically every chip ever so wasting an ethernet PHY to "make" rs232 would be weird
perfinionpretty sure it just takes two of the wires and wires to the rs232 pins
bunderdoesn't it need 4 pins?
bundergod serial is so old i don't remember
bunderi had to fish through a box here at work to find a cable to connect my dumb terminal to my switch
lblumeThere are enough in an RJ45 plug to make it happy, at least :)
zfsironMann commented on issue zfsonlinux/zfs#4844 - l2arc_feed kernel thread keeps waking up the cpu on idle system <https://github.com/zfsonlinux/zfs/issues/4844#issuecomment-232301649>
jaakkosis it expected that zfs set atime=on will have no effect until dataset is remounted?
jaakkosand even zfs get will return "off"
jaakkosotoh, turning it off from on will take effect immediately
dasjoeYou may be the first person to test setting atime=on ;)
jaakkosthere are bugs like #351 and #1125 about it, but those are old
zfs[zfs] #351 - Changing the atime option doesn&#39;t update mounted dataset <https://github.com/zfsonlinux/zfs/issues/351>
perfinionbunder: true serial needs RX+- and TX+- yeah, stuff like atmegas now just use single with 5v/0 instead of +-15V
perfinionbunder: you dont really *need* 4 pins, there might be stuff inside the console cable dunno
dasjoejaakkos: behlendorf commented 22 days ago "Resolved by 0df9673."
perfinioni never looked taht hard at them
dasjoeRS232 over RJ45 seems to be the norm, although signal and pin mappings differ
jaakkosdasjoe: but yeah, thanks for pointing out
jaakkosi think i'll workaround in my own way for because it's a bit difficult to change our zfs right now
dasjoeAlso, you can (and should) connect most devices' UART via 3 pins: rx, tx, gnd
dasjoejaakkos: I assume will be here some time this week, behlendorf has been quite active lately
zfsironMann opened pull request zfsonlinux/zfs#4847 - Fix handling of errors nvlist in zfs_ioc_recv_new() by ironMann <https://github.com/zfsonlinux/zfs/pull/4847>
zfsperfinion commented on pull request zfsonlinux/zfs#4847 - Fix handling of errors nvlist in zfs_ioc_recv_new() by ironMann <https://github.com/zfsonlinux/zfs/pull/4847#discussion_r70597153>
perfinionnice i kept hitting that bug over and over
DeHackEdare we still discussing 10gig in the consumer market
perfinionDeHackEd: sure, why not
perfinionubnt has a 10G router in development supposedly out next year tho :(
bunderurgh, what kind of bozo locks the mail room :/
PMTbunder: probably the kind who values the mail
bunderdunno, its never been locked
bunderalthough we are hinging on a mail strike still
PMTperhaps they're hiding a union organizer's body in there
bunderi had security let me in, it was just locked heh
PMTwell, there went that theory.
bundermy mail strike theory doesn't hold up well either, that wouldn't affect interoffice mail
bunderbut i'm going with it because i can heh
bunderflap flap flap flap flap
zfsmailinglists35 commented on issue zfsonlinux/zfs#4844 - l2arc_feed kernel thread keeps waking up the cpu on idle system <https://github.com/zfsonlinux/zfs/issues/4844#issuecomment-232348675>
zfsironMann commented on pull request zfsonlinux/zfs#4847 - Fix handling of errors nvlist in zfs_ioc_recv_new() by ironMann <https://github.com/zfsonlinux/zfs/pull/4847#discussion_r70627354>
zfsironMann commented on pull request zfsonlinux/zfs#4815 - Add RAID-Z routines for SSE2 instruction set, in x86_64 mode. by ironMann <https://github.com/zfsonlinux/zfs/pull/4815#issuecomment-232365150>
dasjoeryao: #1828 seems a bit stale, is this still an issue?
zfs[zfs] #1828 - /boot/grub is not readable by GRUB on boot <https://github.com/zfsonlinux/zfs/issues/1828>
ryaodasjoe: Unfortunately, I don't remember this problem at all. It is probably fixed in GRUB2.
ryaoLong fixed.
ryaodasjoe: I can think of one possible issue. Maybe if the pool is newer than what GRUB2 supports. I would need to look at the code to be sure.
dasjoeI thought so
zfsryao commented on issue zfsonlinux/zfs#1828 - /boot/grub is not readable by GRUB on boot <https://github.com/zfsonlinux/zfs/issues/1828#issuecomment-232374046>
ryaodasjoe: Thanks for pinging me about it.
dasjoeryao: sure, I was just browsing the issue list, looking for interesting things. I don't keep IRC logs, but you may find something about that issue there
ryaodasjoe: Sadly, ENOTIME right now. I have this one thing at work that has been clogging my queue for a while, but it is a top priority. It is wearing me out as I am essentially reinventing compiler technology from scratch for something ZFS-related. People at work do not want me to discuss the specifics until they make them public, so I cannot go into detail, but this is sufficiently different than a compil
ryaoer that standard tools/methods can't be used and it must be implemented from first principles. The input is neither text nor binary.
ryaoI do get to implement front-end/back-end because there does exist an input and there do exist instructions as an output. That is why I am describing it to myself as a compiler. :/
ryaoActually, I'll say one other thing without going into specifics. The only thing in ZFS that has a front-end/back-end along the lines of what I just described is zfs send, but it is very simplistic and has had some rather valid public criticism by a few people. Solving that criticism is also a hard problem. ;)
prometheanfireryao: you know if we need to run any 'cleanup' on https://github.com/zfsonlinux/zfs/commit/31b6111fd92a6bc79a34cb18fd919a29c407ff67
ryaoprometheanfire: What do you mean by cleanup?
prometheanfireryao: And because of that, the xattr
prometheanfirestuff will never be marked as unlinked. So the whole unlinked stuff
prometheanfirewill stay there until shrink cache or umount.
prometheanfireactually, guess just umout works
ryaoprometheanfire: This is an annoying bug that is good to finally see die.
prometheanfiretempted to reboot my storage node just for this
ryaoprometheanfire: I'll try cherry-picking it later, although as I said, my queue is clogged.
prometheanfirethough it's mostly block io that it does
prometheanfireI run off of master, so don't feel preasure from me
prometheanfirethere are a few other xattr fixes in the log from yesterday
ryaoIt would be easier if I didn't have to do R&D in a completely new language. People at work are requiring that I write code in Go.
prometheanfireryao: sorry to hear that
ryaoprometheanfire: It could have been worse. They could have made me write this thing in Java. However, I regret to say that I would find Java easier since the tools are more mature. e.g. I can attach a debugger to it without much fuss.
prometheanfireACTION has been wanting to try rust
ryaoACTION had been very happy with C. :/
ToxicFrogI have a nostalgic affection for C, but I prefer Rust for those very rare occasions when I need a language that's C-ABI compatible.
ToxicFrogI'm sorry about Go :(
ryaoI am very tempted to rewrite our entire codebase in C for the sake of a comparison.
ryaoToxicFrog: You can access C from Go with CGo, but people prefer not to do it.
ToxicFrogACTION nods
ToxicFrogI've managed to mostly avoid Go at work despite interacting a lot with a team that uses it
ToxicFrogThat's probably going to change as we share more infrastructure at work, though.
ToxicFrogAt least it's not Java or C++.
ryaoThat rewrite will need to wait until I have a working algorithm implemented correctly. On the bright side, I know plenty of wrong ways to implement it by now.
ryaoThe right way is to go back to first principles, although writing this in a new language as one of the very first things I do is painful. Having to interrupt thinking about logic to think about language semantics when the thing is too big to fit in my head or any 1 diagram that I could resaonably create is a productivity killer.
ryaoAnyway, that was a great way to relieve some pent up frustration. Now to get back to constructing this extremely complicated transformation.
Melianryao meant: Anyway, venting was a great way to relieve some pent up frustration. Now to get back to constructing this extremely complicated transformation.
NukienI'm still a C nut. Do a lot of microcontroller work, 95% in C, the rest in assembly when absolutely necessary.
NukienNo C++ thankfully
ryaoNukien: Our CEO decided that all new projects would be done in Go.
NukienI'll shoot myself in my own foot myself thank you very much
NukienWhat did he base that decision on ? And who had input into it ?
ryaoNukien: I would love C++ in comparison to Go. You can basically write C code in it and C++-style C++ is less painful in my opinion.
ryaoNukien: I think it is because docker is written in Go and being in Go gives software a feeling of being hip or something like that.
NukienAh, so heard about it or read about it on a flight, and here you are ...
ryaoNukien: I have had more time than that to gain experience with it, but what I want to do is exercising rarely used language features and I need to think about syntax whenever trying to transform a diagram into code. I simply wouldn't need to do that in C or C++ as I could just write what I want without having to look up documentation
ryaoActually, I can make this easier on myself. I'll write these complex things in C first and then rewrite them in Go. No one at work needs to know that I used C as an intermediate language.
NukienHah sounds like you need to write a translator ...
NukienOr ... another compiler !
GrayShadeare you using Go just for prototyping or will that become the final version?
ryaoGrayShade: The code that will run our infrastructure will be Go code unless I manage to convince people C is better by demonstratng it after rewriting it on my own time. The code is meant to be separate from the main ZoL codebase, but I plan to write a C version to make it avaliable to others and I could see myself getting that into the OpenZFS codebase.
ryaoThat said, the thing that was a pain because I had to figure out both its structure and how to do it in Go is far easier to write in C. I wish I had done this sooner.
ryaoOr well, far easier for me to write in C. No guessing as I am trying to work out structure.
ryaos/work out/prototype/
Melianryao meant: Or well, far easier for me to write in C. No guessing as I am trying to prototype structure.
GrayShaderyao: it can be frustrating. I'm learning a new language myself and at times it gets pretty bad
ryaoGrayShade: I am doing the equivalent of writing a top down parser without YACC in a new language (although this doesn't look top down, although I ought to examine that in more detail).
ryaoThe tokenization is actually the complex part... I might be able to (ab)use YACC for this, even though the input isn't a real language...
GrayShadeI found top-down parsers the easiest to understand, but I'm sure what you're doing is hard
ryaoYacc in go: https://golang.org/cmd/yacc/
ryaoGrayShade: Have you ever written a complex one without the assistance of YACC?
GrayShadeBy the way, what was that thing in the news about parsing CPU manuals to generate compiler backends?
ryaoI didn't hear about it.
GrayShadeNothing more complex than let's say pascal
GrayShade(I took top-down to mean recursive descent)
GrayShadeCheck out @kennwhite's Tweet: https://twitter.com/kennwhite/status/752925134730715136?s=15
GrayShadeErm sorry for the check out part. I copied the whole line thinking it's just the link
GrayShadeI don't know how to use a phone
ryaoThinking about this some more, I don't think I can apply YACC to it. I can't say why right now. I can say that I know not to say "I am doing something patentable in the US" when I hear that the investors want to see more IP" in the future though.
djsryao: if I haven't already sent you this link... http://bford.info/pub/lang/peg.pdf
djsbeing as you are writing parsers... ;)
djsI haven't used these thing extensively but my limited experience was pretty good
GrayShadedjs: is go expressive enough for those? I don't know the language
djs*far* nicer than lex/yacc/bison
ryaoThanks, but upon thinking about how to make a yacc grammar, it occurs to me that no grammar will work because I need a non-deterministic state machine while all parsers have deterministic state machines.
djsGrayShade: no idea... first thing I would do is google for a go peg parsing library ;)
GrayShaderyao: they patented abs() afaik
ryaoWell, no grammar for off the shelf parser generators will work.
djsryao: the precedence stuff that peg parsers have might do what you need
ryaoGrayShade: There is a bright side to this. Having a patent on something implemented in CDDL (C) and Apache 2.0 (Go) licensed code is really only a defensive thing.
ryaoAt least, the plan at work is that all of our new OSS projects' code will be Apache 2.0. Getting it into OpenZFS would mean it gets to be CDDL though.
ryaodjs: I don't see how to do this with precedence. The parser was designed on the whiteboard to be non-deterministic. I don't think the problem of parsing a machine language requires that.
djsryao: well, not knowing the details it's hard to say, can you give me an example of how it's non-deterministic?
ryaoBy non-deterministic, I mean the second token could very well put the parser into two simultaneous states. The total number of simultaneous states will be huge.
djssounds like you need a quantum computer ;0
ryaodjs: I know that you want to help, but I am afraid that I don't see any way that I could respect my manager's wishes and explain in more detail.
GrayShaderyao: peg works well with that
djsyeah, no worries
ryaoGrayShade: I'll read up on it. Thanks.
GrayShaderyao: i mean they can do backtracking, but there are some optimizations (like memoization / caching) you can do so you don't end up with an exponential-time parser
GrayShadefun trivia: there's a language with a COME FROM instruction, like GOTO in reverse. there was some disagreement about what happens when there are multiple COME FROM statements with the same source
GrayShadethough i'm pretty sure i forgot most of what i knew about pegs, so i might be wrong. i read about them in another life
GrayShadethey're basically BNF embedded in a programming language
djsthe thing I really liked most about pegs was that the lexer and parser are not two separate things so what forms a token can vary by parsing context
ryaoGrayShade: PEG doesn't look like what I need. The more I think about this. The more alien it is in comparison to anything that I know previously. After I can talk about this publicly, I think that I am going to go ask people at my university's CS department if they recognize this kind of parser. If it is a new class, then maybe I could get that PhD my younger self wanted after all.
djsoh man, the arcane world of grammars and language classes isn't somewhere I'd want to venture
ryaodjs: Well, work required it for this thing that I can't elaborate on, but I think there are sufficient hints to get the gist. :)
GrayShaderyao: haha, well, i didn't expect it to be that easy.
ryaoThis might be appropriate: https://xkcd.com/664/
ryaoAlthough definitely not 48 hours.
ryaoThe part about a paragraph probably applies. There is a page on wikipedia describing the problem, but it lacks a section about this variation. It appears to not exist in literature.
ryaoThe problem, not the solution. I can't find that either.
ChibaPetryao: Do you know of any recent guides for getting EL7 on a ZFS root? I'm hoping to not have to write one, although I'm willing if I must.
ryaoChibaPet: I am afraid that I am not up to date on that, although the process should be analogous to that for other distributions.
GrayShadeEL7 uses systemd, fwiw
ryaoGrayShade: I am aware. I do not need to revise what I said.
ryaoGrayShade: People have done / on ZFS with systemd on Gentoo and Ubuntu.
ryaoThey have even usd dracut. I think prometheanfire is one of them.
GrayShadei know, i was thinking the instructions on the arch linux wiki might work
GrayShadeat least as a starting point
ChibaPetI'll work something out I guess. :)
GrayShadethough there might be better sources available
ChibaPetI need to learn more Dracut anyway.
zfstcaputi commented on pull request zfsonlinux/zfs#4329 - ZFS Encryption by tcaputi <https://github.com/zfsonlinux/zfs/pull/4329#issuecomment-232409496>
GrayShaderyao: i think you mentioned having a blog. was it the one from your workplace or do you have another one?
prometheanfireusing it now
ChibaPetprometheanfire: Is it relatively straightforward using Dracut to make an initramfs you can stuff into the UEFI boot partition? I like the model where I have the UEFI boot partition and everything else stuffed into an encrypted pool. (Flavour for RAID.)
ryaoGrayShade: I did make one, but I haven't kept up with it. I did do a few posts at my employer's blog and I ought to do more. I will probably have plenty to discuss about this once I can publicize the full details.
zfstonyhutter commented on pull request zfsonlinux/zfs#4760 - OpenZFS 4185 - add new cryptographic checksums to ZFS: SHA-512, Skein, Edon-R by tonyhutter <https://github.com/zfsonlinux/zfs/pull/4760#issuecomment-232411216>
zfsbehlendorf closed issue zfsonlinux/zfs#1828 - /boot/grub is not readable by GRUB on boot <https://github.com/zfsonlinux/zfs/issues/1828>
GrayShadeyeah, i wanted to add it to my feed reader so i wouldn't miss the exciting conclusion of this saga :D
ChibaPetIt was fairly straightforward doing it with Slackware's mkinitrd, and my impression is that Dracut is more flexible.
ryaoGrayShade: It could just be that I am completely off track, but given the pain that this has given me and the lack of literature on the topic, I suspect that I am on to something. Hopefully, I really am.
zfstonyhutter commented on pull request zfsonlinux/zfs#4329 - ZFS Encryption by tcaputi <https://github.com/zfsonlinux/zfs/pull/4329#issuecomment-232411772>
cjhowehi guys, i'm on my like 5th try of installing zfs on linux from this https://github.com/zfsonlinux/zfs/wiki/Ubuntu-16.04-Root-on-ZFS#step-6-first-boot
cjhowei can get to the reboot step, but then my computer won't boot up to ubuntu, and the machine supports UEFI and 512B sectors
GrayShadecjhowe: does grub install with no complaints?
cjhoweyeah, it even said it succeeded
GrayShadewhat happens after reboot?
ryaocjhowe: It sounds like you have a BIOS that wants to UEFI boot the system if it sees GPT partitioning. Is this a Lenovo system? My laptop has that problem.
zfsbehlendorf commented on pull request zfsonlinux/zfs#4329 - ZFS Encryption by tcaputi <https://github.com/zfsonlinux/zfs/pull/4329#issuecomment-232415543>
cjhoweit's a toshiba
cjhowebut also, I am using GPT and trying to UEFI boot
ryaocjhowe: Assuming that your drives are not larger than 2TB, the workaround is to switch to DOS partitioning. You can destroy the GPT and make a DOS MBR with a partition where ZFS is located. Then install GRUB2 to it. It will owrk.
ryaocjhowe: Oh, that is different then.
ryaoI haven't tried that configuration yet, so I can't give any tips aside from telling us what you see during boot.
GrayShadecjhowe: the sledgehammer alternative is to boot from an usb drive
ryaocjhowe: Also, if you don't see GRUB2, then it is a GRUB2 install issue for which people in #grub2 should be willing to help troubleshoot.
Melianryao meant: cjhowe: Also, if you don't see GRUB2, then it is a GRUB2 install issue for which people in #grub should be willing to help troubleshoot.
zfstcaputi commented on pull request zfsonlinux/zfs#4329 - ZFS Encryption by tcaputi <https://github.com/zfsonlinux/zfs/pull/4329#issuecomment-232416379>
cjhowecan't i use bootctl or something? isn't that included with systemd?
GrayShadecjhowe: afaik, systemd-boot can't boot directly from zfs, so you'd need another partition or drive
ryaocjhowe: If you are not using GRUB2 as the UEFI payload, then you need to setup Linux to be booted by UEFI with its initramfs.
ryaoThe kernel must support it and the initramfs must have ZFS support.
cjhoweif i run efibootmgr my entry doesn't show up at all...
prometheanfireChibaPet: yep
ryaocjhowe: That is likely the thing breaking boot.
cjhowethere's no way it should be able to run without setting the EFI variables
cjhowebut how do i tell grub to make the EFI variables?
ryaocjhowe: People in #grub or your distribution's support channel would know. Others here might know, but I do not know. Unfortunately, UEFI boot has been on my backburner for so long that I probably won't get to it any time soon.
ryaocjhowe: Also, don't keep reinstalling on ZFS like you are doing. You just want to be able to chroot to get back into the existing install as needed (and maybe update /etc/zfs/zpool.cache after exiting it). It saves time. This might be helpful to you: https://gist.github.com/ryao/3c345f206b19c9795109
cjhoweyeah, i've been trying to get the chroot mounted again
cjhowethe biggest problem is i have to unmount the stuff zpool tries to mount, and then remount rpool/ROOT/ubuntu to /mnt, then remount everything else
cjhowebecause rpool/ROOT/ubuntu doesn't mount automatically
ryaocjhowe: Use -N
ryaocjhowe: More specifically, use this: zpool import -N -o cachefile=/tmp/zpool.cache rpool
ryaocjhowe: You will likely want to cp /tmp/zpool.cache into /path/to/chroot/etc/zfs/zpool.cache before enter a chroot environment so that it can be cached in the initramfs.
cjhoweoh, i see
ryaocjhowe: You could also do mountpoint=/ on rpool/ROOT/ubuntu. It is rather harmless as long as the initramfs passes zfsutil as a mount option.
ryaoThat assumes it uses mount.zfs.
cjhoweah, i did it
cjhoweyeah, i guess you've got it mounting on / with some chroot alternative?
cjhowei know if you do -R /mnt in the zpool call it should put it in the right place though
ryaocjhowe: Yes.
prometheanfireChibaPet: add_dracutmodules+="zfs crypt crypt-gpg"
prometheanfireChibaPet: that's in my dracut conf
ryaocjhowe: I forgot to mention -R. I am glad that you caught it quickly.
prometheanfireChibaPet: this is my init initrd=\initramfs-4.4.14-gentoo.img rd.luks.uuid=eeec1057-7445-4510-a737-da8fc0085a6c rd.luks.allow-discards=eeec1057-7445-4510-a737-da8fc0085a6c rd.luks.crypttab=0 root=zfs:mthode-work/OS zfs.zfs_arc_max=3221225472 init=/usr/lib/systemd/systemd ro
ryaocjhowe: I am saying this off the top of my head.
zfstonyhutter commented on pull request zfsonlinux/zfs#4329 - ZFS Encryption by tcaputi <https://github.com/zfsonlinux/zfs/pull/4329#discussion_r70665875>
cjhoweso i'm chroot'd in, there's definitely data on the EFI partition
prometheanfireis fransurbo in this chan?
prometheanfirewanted to work with him with the zfs/cinder stuff
cjhoweryao, thanks for the help, i think it's fixed, efibootmgr showed the entry after i ran grub-install
cjhoweor.... my computer is deleting my efibootmgr entries
ryaoprometheanfire: I am not sure if he has ever been in IRC, but if he has, it has been years.
zfstcaputi commented on pull request zfsonlinux/zfs#4329 - ZFS Encryption by tcaputi <https://github.com/zfsonlinux/zfs/pull/4329#discussion_r70669207>
ChibaPetprometheanfire: ty! :) I'm laggy (meeting) but I appreciate the pointers. Might get a chance to try implementing this weekend.
prometheanfireChibaPet: cool
zfsbehlendorf pushed to master at zfsonlinux/zfs - Comparing 81edd3e834...ae25d22235 <https://github.com/zfsonlinux/zfs/compare/81edd3e834...ae25d22235>
zfsbehlendorf closed pull request zfsonlinux/zfs#4815 - Add RAID-Z routines for SSE2 instruction set, in x86_64 mode. by ironMann <https://github.com/zfsonlinux/zfs/pull/4815>
zfsbehlendorf commented on pull request zfsonlinux/zfs#4847 - Fix handling of errors nvlist in zfs_ioc_recv_new() by ironMann <https://github.com/zfsonlinux/zfs/pull/4847#issuecomment-232427590>
HeebieHello. I have some filesystems in a pool.. which are supposed to be active, but if I query and get the "readonly" attribute for these, I get this: "<TANK>/<FS1>/<FS> readonly on temporary"
HeebieLoad is building on this server, as well.
Heebieand there is currently a snapshot sending elsewhere (via ZREP)
Heebiezpool iostat -v shows plenty of writes to the pool, and to every vdev, and zpool status -v shows no issues.