View previous topic :: View next topic |
Author |
Message |
tnt Veteran
Joined: 27 Feb 2004 Posts: 1222
|
Posted: Fri Jul 31, 2015 8:17 am Post subject: |
|
|
mrbassie wrote: | More recently I went from 0.6.2-r4 to r5 to 0.6.3 to 0.6.4 to .1 then .2. Everything went fine. |
should I go from 0.6.3 to 0.6.4 and then to 0.6.4.2 ?
I have to skip 0.6.3-r2 as I had serious problems running it and had to revert to 0.6.3:
https://groups.google.com/a/zfsonlinux.org/forum/#!topic/zfs-discuss/DdgjKitaxAk _________________ gentoo user |
|
Back to top |
|
|
mrbassie l33t
Joined: 31 May 2013 Posts: 772 Location: over here
|
Posted: Fri Jul 31, 2015 3:56 pm Post subject: |
|
|
tnt wrote: | mrbassie wrote: | More recently I went from 0.6.2-r4 to r5 to 0.6.3 to 0.6.4 to .1 then .2. Everything went fine. |
should I go from 0.6.3 to 0.6.4 and then to 0.6.4.2 ?
I have to skip 0.6.3-r2 as I had serious problems running it and had to revert to 0.6.3:
https://groups.google.com/a/zfsonlinux.org/forum/#!topic/zfs-discuss/DdgjKitaxAk |
I honestly don't know, but it's worth a try.
The best person to ask would probably be ryao but I don't think he's been on here for a while. |
|
Back to top |
|
|
tnt Veteran
Joined: 27 Feb 2004 Posts: 1222
|
Posted: Mon Aug 03, 2015 3:38 pm Post subject: |
|
|
the same issue with 0.6.4 packages and gentoo-sources 3.14.47:
Code: | zfs 1823616 0
zunicode 315324 1 zfs
zavl 3753 1 zfs
zcommon 30124 1 zfs
znvpair 37112 2 zfs,zcommon
spl 46898 3 zfs,zcommon,znvpair |
also with both 0.6.4 and 0.6.4.2 packages and gentoo-sources 3.18.19:
Code: | zfs 1827728 0
zunicode 315356 1 zfs
zavl 3753 1 zfs
zcommon 30188 1 zfs
znvpair 37656 2 zfs,zcommon
spl 47218 3 zfs,zcommon,znvpair |
and then I did the following:
- I stopped zfs service
- I manually rmmod-ed all the related modules (zfs, zunicode, zcommon, znvpair, spl)
- I started zfs service again
and modules loaded without problem, zpool was assembled without a problem, and all filesystems were mounted without a problem.
so, I did couple of stops and starts of zfs service and everything went fine.
next thing, I rebooted the system hopping that I broke through the ice and that now everything will be fine.
unfortunately, I was wrong. after the reboot, again no pools.
manual unload and load of zfs modules fixed the issue once again.
so, my zfs service was added to the 'boot' level. I removed it from there and added it to the 'default' level.
reboot.
everything is fine.
conclusion: there is something missing (probably some 'after' or 'depend') in zfs service init script which lets it run before it should. maybe defices are still not ready or something similar.
Code: | rc-update show boot
binfmt | boot
bootmisc | boot
consolefont | boot
fsck | boot
hostname | boot
hwclock | boot
keymaps | boot
localmount | boot
loopback | boot
mdraid | boot
modules | boot
mtab | boot
net.lo | boot
procfs | boot
root | boot
swap | boot
swapfiles | boot
sysctl | boot
termencoding | boot
tmpfiles.setup | boot
urandom | boot |
any idea what that could be?
personally, I have added the following line to zfs init.d script:
Code: | after binfmt consolefont fsck hostname hwclock keymaps localmount loopback mdraid modules mtab net.lo procfs root swap swapfiles sysctl termencoding tmpfiles.setup urandom |
and now it runs fine in 'boot' level also, but I guess it would be much better if we find out exact service it depends on and the script starts shipping with it included. _________________ gentoo user |
|
Back to top |
|
|
mrbassie l33t
Joined: 31 May 2013 Posts: 772 Location: over here
|
Posted: Tue Aug 04, 2015 10:32 am Post subject: |
|
|
That's interesting. That did not effect me
Code: | binfmt | boot
bootmisc | boot
fsck | boot
hostname | boot
hwclock | boot
keymaps | boot
loopback | boot
microcode_ctl | boot
modules | boot
mtab | boot
net.eth0 | boot
procfs | boot
root | boot
swap | boot
swapfiles | boot
sysctl | boot
termencoding | boot
tmpfiles.setup | boot
urandom | boot
zfs | boot
|
|
|
Back to top |
|
|
tnt Veteran
Joined: 27 Feb 2004 Posts: 1222
|
Posted: Tue Aug 04, 2015 2:42 pm Post subject: |
|
|
it runs OK so far.
I guess I'll keep it like that for couple of weeks, to make sure everything is fine, and then go for 'zpool upgrade'.
btw, should I do/enable anything special on my pool after I get access to those new flags? _________________ gentoo user |
|
Back to top |
|
|
mrbassie l33t
Joined: 31 May 2013 Posts: 772 Location: over here
|
Posted: Wed Aug 05, 2015 6:59 am Post subject: |
|
|
I didn't, I just run home computers though. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
Manuel.G n00b
Joined: 07 Jan 2015 Posts: 8
|
Posted: Sat Oct 10, 2015 12:16 pm Post subject: |
|
|
After syncing today emerge would downgrade sys-fs/zfs from 0.6.5.2-r2 to 0.6.4.2 (including spl and zfs-kmod), anyone know the reason for this?
Edit: Well, after syncing again a day later it went away. Some hiccup at the server I synced from yesterday I guess. |
|
Back to top |
|
|
desultory Bodhisattva
Joined: 04 Nov 2005 Posts: 9410
|
Posted: Tue Oct 13, 2015 3:42 am Post subject: |
|
|
Manuel.G wrote: | After syncing today emerge would downgrade sys-fs/zfs from 0.6.5.2-r2 to 0.6.4.2 (including spl and zfs-kmod), anyone know the reason for this?
Edit: Well, after syncing again a day later it went away. Some hiccup at the server I synced from yesterday I guess. | It was probably not a hiccup, as such. |
|
Back to top |
|
|
Manuel.G n00b
Joined: 07 Jan 2015 Posts: 8
|
Posted: Tue Oct 13, 2015 6:31 am Post subject: |
|
|
Oh, I see. Thanks for the pointer, will look around the mailing lists for information in the future as well! |
|
Back to top |
|
|
goertzenator n00b
Joined: 26 Sep 2014 Posts: 12
|
Posted: Fri Oct 23, 2015 5:02 pm Post subject: zfs kernel source question |
|
|
I've happily used FreeBSD on ZFS root for many years now and would like do the same for my Gentoo system. One thing that is giving me pause however is the things you have to do to get a linux+zfs source tree. I'd like my system to be fairly easy to maintain going forward, and adding a tricky source setup step is counter to that.
Is there a kernel ebuild that adds spl and zfs? And a secondary question: is there any reason why the sources can't be integrated into say gentoo-sources, perhaps under a zfs USE?
PS, thanks for bringing ZFS to Gentoo. I am currently using BTRFS and am tired of scraping through random documentation to figure how to do things that are trivial in ZFS. I've had enough and want to switch. |
|
Back to top |
|
|
Dorsai! Apprentice
Joined: 27 Jul 2008 Posts: 285 Location: Bavaria
|
Posted: Sun Oct 25, 2015 1:01 pm Post subject: |
|
|
There really is no need to have ZFS in the kernel source tree except if you absolutely don't want an initramfs.
Just:
-compile your kernel with "genkernel kernel"
-emerge zfs (incl. spl and zfs-kmod)
-build modules with "genkernel modules"
-for automation set CMD_CALLBACK="emerge --quiet @module-rebuild" in /etc/genkernel.conf
The last step will do the first three steps for you with a simple "genkernel all" in the future. The same should also be possible with dracut. |
|
Back to top |
|
|
goertzenator n00b
Joined: 26 Sep 2014 Posts: 12
|
Posted: Mon Oct 26, 2015 1:40 pm Post subject: |
|
|
Thanks for the guidance. I currently don't have an initramfs and enjoy the simplicity of not having one, but maybe initramfs is the path of least resistance for getting zfs. |
|
Back to top |
|
|
ZZZZZZZ n00b
Joined: 01 Jan 2016 Posts: 2 Location: 1111
|
Posted: Sun Jan 10, 2016 12:28 pm Post subject: |
|
|
Hello,
I recently upgraded to the kernel 4.x and zfs 0.6.5 combo and I have a couple of questions.
A drive that I have exclusively with zfs is no able to show the datasets. Even prior to this upgrade lately I had to bring it up with zpool export -f pool and zpool import pool. Because apparently it was imported but halfway.
Now with the upgrade one drive works flawlessly and the above has weird behaviors.
zpool is able to see and import it but ALL datasets are missing. And zfs also reports the name foo/bar/ but after that all datasets don't show up.
Looking more into it I used the program zbd, this gave me a glimmer of hope as zbd DOES report the correct structure with ALL the datasets inside that pool. Moreover testing some zbd options apparently is able to list each file inside it.
I am not familiar with zbd but since I saw reporting all files with their size and all I thought I was smart and wanted to extract them into another drive. But it looks is not that simple.
The only change that I made after upgrade was to upgrade zfs features of only the problematic drive.
So before trying things blindly like a monkey in the hopes to fix it I wanted to ask the proper steps to proceed. Moreover, whether there's a chance of getting the data from that drive at all.
Basically the datasets went missing after the kernel+zfs upgrade.
Oh and a last question, after zfs upgrade emerge has a warning against upgrading zfs features of the root drive. Is this warning only for people with booting with GRUB is for every case?
This zfs update is pretty amazing, so far has handled a massive upgrade without choking the ram like earliest versions and I haven't even needed to restrict ram for zfs in the modules options.
thanks |
|
Back to top |
|
|
mrbassie l33t
Joined: 31 May 2013 Posts: 772 Location: over here
|
Posted: Wed Jan 13, 2016 7:52 am Post subject: |
|
|
@ZZZZZZZ
Although they're officially supported I've had similar troubles with 4.x kernels, so I'm back on 3.x for as long as need be.
There's also an issue with 0.6.5.3-r1 and -r2 and grub. Iirc the emerge message said something about having to send your data to a new pool if you're booting from it.
Hope that's of help to you for what it's worth. |
|
Back to top |
|
|
Yamakuzure Advocate
Joined: 21 Jun 2006 Posts: 2284 Location: Adendorf, Germany
|
Posted: Thu Jan 14, 2016 7:39 am Post subject: |
|
|
I have my laptop set up using zfs-0.6.5.3 with gentoo-sources-4.2.8 and everything works fine.
The emerge message means, that if your boot device is on ZFS, you must not do an 'zfs upgrade' to that, as grub2 can not handle the new features, yet. Everything else can be upgraded as you wish.
If you upgrade your datasets, make sure not to try anything on them with any previous versions of zfs. That means, that after upgrading, you have to make sure, that zfs, zfs-kmod and spl are all the newest version you used for upgrading.
Maybe your kernel still uses an older version?
See kernelOfTruth' ZFSonLinux, SystemRescueCD up-to-date kernel-modules for a SysRescCD with the current modules. You can do some checks, and whether you can import your problematic zpools there. _________________ Important German:- "Aha" - German reaction to pretend that you are really interested while giving no f*ck.
- "Tja" - German reaction to the apocalypse, nuclear war, an alien invasion or no bread in the house.
|
|
Back to top |
|
|
janos666 n00b
Joined: 15 Nov 2015 Posts: 30
|
Posted: Sat Jul 30, 2016 3:01 am Post subject: Re: Gentoo on ZFS |
|
|
ryao wrote: | [...] I intend to release documentation detailing how to install Gentoo Linux on ZFS when it is ready.
Does anyone have any questions? |
Is it possible to boot from ZFS root (single device, single volume on a SATA SSD) with an EFI Stub kernel (loaded from a separate partition and FAT32 filesystem in real EFI boot mode on x86_64 UEFI platform) without an initramfs (which seems to be inconvenient to set up with an EFI Stub kernel and I never used initramfs/initrd with BIOS and GRUB2 either, I don't like it).
This would be a transition from Btrfs root to ZFS root (reboot from USB, fresh backup from the root fs, partition wipe, zfs creation, file restore, reboot).
I already have the SPL and ZFS modules built-in to the kernel (I don't like loadable modules, so I always build everything statically) and the system is already booting in EFI mode (from single/single/single profile Btrfs).
The root is currently mounted by the built-in kernel command line with it's PARTUUID (GPT) (first read-only, then remounted read-write by the "root" RC script but that's only because there is a several years old bug which causes a kernel panic if you set "noatime" in the kernel command line, otherwise I would simply mount the rootfs read-write with the kernel command line and delete that rc script - which I hope would work with ZFS where noatime can be defined as a property and it's probably handled by the ZFS modul, not the kernel code which has problems with noatime). |
|
Back to top |
|
|
otakugeek n00b
Joined: 26 Jun 2014 Posts: 50
|
Posted: Wed Aug 31, 2016 12:33 pm Post subject: |
|
|
I've tried for few days the ZFS version 0.6.5.7 and work fine until the kernel 4.6.x, all system work fine, but the ZFS service use alot of memory, I have a laptop with 16GB and almost all time zfs use 10GB~12GB, I have a SSD and tmpfs so was not slow, the I/O is ok too, I could copy or move above 300mbps or more depend the type of file.
but the memory consuming is horrible, if I have 32GB maybe I could continue, but close all my memory is impossible.
but I'll wait the next updates, with better memory manager, support to kernel mainline and playmouth I'll use again. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
Yamakuzure Advocate
Joined: 21 Jun 2006 Posts: 2284 Location: Adendorf, Germany
|
Posted: Thu Sep 01, 2016 1:11 pm Post subject: Re: Gentoo on ZFS |
|
|
janos666 wrote: | Is it possible to boot from ZFS root (single device, single volume on a SATA SSD) with an EFI Stub kernel (loaded from a separate partition and FAT32 filesystem in real EFI boot mode on x86_64 UEFI platform) without an initramfs | Short answer: No.
Long answer: You need the zfs module just to actually *find* the zfs pool you want to boot into. And as this is external, it has to be put into an initramfs.
"But zfs can be built in!": Well, go ahead and try to patch the zfs module and the spl code into your kernel.
But: You can embed your initramfs into the stub kernel using (e.g.) dracut like it is outlined on the EFI Stub Kernel wiki page.
Here is a step-by-step howto: EFI stub kernel booting ZFS root filesystem with builtin initramfs. _________________ Important German:- "Aha" - German reaction to pretend that you are really interested while giving no f*ck.
- "Tja" - German reaction to the apocalypse, nuclear war, an alien invasion or no bread in the house.
|
|
Back to top |
|
|
janos666 n00b
Joined: 15 Nov 2015 Posts: 30
|
Posted: Fri Sep 02, 2016 2:56 pm Post subject: Re: Gentoo on ZFS |
|
|
[quote="Yamakuzure"] janos666 wrote: | Is it possible to boot from ZFS root (single device, single volume on a SATA SSD) with an EFI Stub kernel (loaded from a separate partition and FAT32 filesystem in real EFI boot mode on x86_64 UEFI platform) without an initramfs
"But zfs can be built in!": Well, go ahead and try to patch the zfs module and the spl code into your kernel. |
I am not sure what you mean exactly. I think I do have it (and everything else I use) built-in "properly".
I have the whole runtime module loading functionality disabled in the kernel config and ZFS (as well as everything else) works just fine.
Portage keeps nagging me about this, even with the built-in use flag enabled for the ZFS package (I guess that's a bug!), so I usually enable the module loading, process the ZFS package, disable module loading, and build the fully monolithic kernel (without runtime module loading support).
Indeed, although that doesn't satisfy my preference towards having a monolithic kernel without any kind of initramfs/disk or the likes of those. (Actually, that's my biggest practical problem with systemd: I think it would make initramfs absolutely mandatory which I don't wish to use in any case if avoidable).
---
Although, in the mean time, I had to go back using the UEFI-CMS BIOS-compatible boot mode with GRUB2 because I realized this Asrock board can't be used in EFI mode on a headless setup. It treats the lack of a display as an utter disaster, reboots in recovery mode and resets the boot order with the BIOS compatible boot options (all the possible ones including those I previously disabled, like network boot) placed at the top of the boot list and consequently fails to boot the .efi kernel from the GPT disk. (And I figure it thinks my PCI-E 4x network card placed in a 16x slot is a display adapter with no EFI GOP support. (Well, I am using an inexpensive consumer PC board for a home server, so these kind of glitches are expected for some degree. I probably should get a relatively cheap second-hand Xeon workstation system.) |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
djdunn l33t
Joined: 26 Dec 2004 Posts: 810
|
Posted: Mon Nov 14, 2016 3:39 am Post subject: |
|
|
fyi i had a problem using ~amd64 latest genkernel to boot zfs as root for 4.8.7 kernel had to go to genkernel-9999 to work _________________ “Music is a moral law. It gives a soul to the Universe, wings to the mind, flight to the imagination, a charm to sadness, gaiety and life to everything. It is the essence of order, and leads to all that is good and just and beautiful.”
― Plato |
|
Back to top |
|
|
alinefr Tux's lil' helper
Joined: 05 Jul 2009 Posts: 113 Location: São Paulo, Brasil
|
Posted: Sun Nov 27, 2016 2:09 pm Post subject: |
|
|
What would be the best way to set zfs_arc_max? As a genkernel user I did try many different ways like adding options on /etc/modprobe.d/zfs.conf inside initramfs, or even adding straight on where genkernel line calls modprobe zfs, like modprobe zfs zfs_arc_max=3435973837 but no way. I ended up adding Code: | echo 3435973837 > /sys/module/zfs/parameters/zfs_arc_max | to /etc/local.d/zfs.start. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
|