View previous topic :: View next topic |
Author |
Message |
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Mon Jun 01, 2015 1:56 pm Post subject: Systemd trying to get in... |
|
|
I have a clean install of Gentoo 64bit with systemd masked and "-systemd" in my global use flags. I also mask "sys-fs/udev". Now it wants to change to eudev, but it claims that there is a circular dependency and the only fix is to change virtual/libudev-215-r1 to use the systemd flag. What the heck? I am using the default profile, not a KDE/Gnome/whatever profile. The system is shell-only and is using btrfs in RAID1 for a file-server. Gentoo lives on an old, 40GB IDE disk separate from the array. So why is libudev trying to depend on systemd? _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
charles17 Advocate
Joined: 02 Mar 2008 Posts: 3664
|
Posted: Mon Jun 01, 2015 2:12 pm Post subject: Re: Systemd trying to get in... |
|
|
Hm, according to Code: | RDEPEND="
!systemd? ( || (
>=sys-fs/udev-208-r1:0/0[${MULTILIB_USEDEP},static-libs?]
>=sys-fs/eudev-1.3:0/0[${MULTILIB_USEDEP},static-libs?] )
) | it should be satisfied with >=sys-fs/eudev-1.3
Could you post or pastebin the output of the message you are receiving? |
|
Back to top |
|
|
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Mon Jun 01, 2015 2:33 pm Post subject: |
|
|
I figured it out. Portage was ruined. When I installed the OS my disk was sda, but it was now sdc in the actual OS. I guess I need to use UUIDs, but they're hard to write down and then type in. I'll just use UUIDs from now on. Thanks for the help, but this one is on me. _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
depontius Advocate
Joined: 05 May 2004 Posts: 3509
|
Posted: Mon Jun 01, 2015 3:38 pm Post subject: |
|
|
The_Great_Sephiroth wrote: | Ever picture systemd as what runs "The Borg"? |
That quote seems quite appropriate, hopefully in more ways than just the obvious one. (Think "Voyager" finale.)
As for RAID-1 btrfs, watch out. A few weeks ago I lost a filesystem like that. Actually, I was using the btrfs native RAID-1. We had a power outage. I know I should be using a UPS, but it died a bit back. Still, losing data being written at the moment is normal. Losing last-written data is understandable. Having a filesystem go so bad that the "mount" command goes into an infinite hang, needing a reboot to recover, is unforgivable. Plus I couldn't even find the filesystem with "btrfs-scan". Luckily I had a full backup from less than 12 hours earlier. I'm now running ext4 on mdadm RAID-1.
I need to start a separate thread on that failure, if only to solicit opinions on UPS models. _________________ .sigs waste space and bandwidth |
|
Back to top |
|
|
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Wed Jun 03, 2015 12:57 pm Post subject: |
|
|
I am using BTRFS due to bit-rot protection. I am duplicating both the checksum data and the physical data. The only other system that protects this way is XFS, and it is a resource-hog.
Also, reinstalling and using UUIDs to mount fixed the issue. For some reason my kernel saw things differently than the live CD last time. Lesson learned, use UUIDs! _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
depontius Advocate
Joined: 05 May 2004 Posts: 3509
|
Posted: Wed Jun 03, 2015 3:12 pm Post subject: |
|
|
You can do that with mdadm as well, you just have to put a scan into a cron job and let it run in the background.
White reminds me... I haven't done that yet, thanks for the jog. _________________ .sigs waste space and bandwidth |
|
Back to top |
|
|
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Wed Jun 03, 2015 3:31 pm Post subject: |
|
|
I have never seen a thing about mdadm protecting against bit-rot. I would like to see some references to that. How does mdadm store checksums and such? Where? What RAID levels does it work with? At that point, since mdadm sets up the array, what file-systems does this work on? How would mdadm protect against bit-rot if I used exFAT or NTFS? Seems unlikely to me. _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
depontius Advocate
Joined: 05 May 2004 Posts: 3509
|
Posted: Wed Jun 03, 2015 4:33 pm Post subject: |
|
|
Hard disks store checksums, that's not the issue with mdadm. The problem is that mdadm RAID-1 reads from the fast/easy drive, and if the read succeeds, doesn't question the other drive. It's possible for the second drive to have an error, and nobody knows.
I'm not sure why filesystems are starting to store checksums, hard disk drives do that. Are the hardware checksums for some reason insufficient? _________________ .sigs waste space and bandwidth |
|
Back to top |
|
|
John R. Graham Administrator
Joined: 08 Mar 2005 Posts: 10589 Location: Somewhere over Atlanta, Georgia
|
Posted: Wed Jun 03, 2015 4:40 pm Post subject: |
|
|
ZFS and BTRFS store cryptographic hashes instead of mere CRCs. They are (marginally) more reliable. Some see the benefit in some use cases (e.g., long-term archival storage).
- John _________________ I can confirm that I have received between 0 and 499 National Security Letters. |
|
Back to top |
|
|
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Wed Jun 03, 2015 5:05 pm Post subject: |
|
|
As far as I understand it, hard-drive checksums do not protect against bit-rot. Heck, they may not be able to. A single drive, even if it detected an error, could not recover the lost data. ZFS/BTRFS can due to the multiple disks. I am looking to preserve data for ages. I wish BTRFS would clone the checksums and such to more than two disks in RAID1. I figure a BTRFS RAID1 on four disks would be insanely secure. _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
John R. Graham Administrator
Joined: 08 Mar 2005 Posts: 10589 Location: Somewhere over Atlanta, Georgia
|
Posted: Wed Jun 03, 2015 5:26 pm Post subject: |
|
|
The trick pony that ZFS and BTRFS can provide is error detection above the physical layer intelligently integrated with traditional RAID. An undetected physical layer read error will be detected by the mismatch of the cryptographic hash, at which point the traditional RAID mechanisms can transparently correct the error and re-write the bad physical block.
- John _________________ I can confirm that I have received between 0 and 499 National Security Letters. |
|
Back to top |
|
|
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Wed Jun 03, 2015 6:20 pm Post subject: |
|
|
Which makes it more valuable to me than mdadm with ext4 on top. I hope the BTRFS team gets the rest done quickly, such as cloning everything on all RAID1 disks. It looks promising to me. _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
depontius Advocate
Joined: 05 May 2004 Posts: 3509
|
Posted: Fri Jun 05, 2015 11:55 am Post subject: |
|
|
I don't know if you read my whole post when I jumped in on this.
If you're going to use btrfs, make darned sure you've got a good UPS, and have it properly set up to shut down your system. As I mentioned, I didn't just lose data, I lost the entire btrfs filesystem. After the crash, "btrfs-scan" couldn't even find it.
I've had UPS setups in the past, but as of now both have died, and I've been wanting to get another, only this time I want to get a good one. I want one that can monitor my battery life, and one with electronics that won't crap out on me.
One expects to lose open files on a crash or power event without a UPS. One doesn't expect to lose the entire filesystem, beyond recovery. I've never had more than the expected crash issues with ext-N or xfs, only with btrfs. Nor have I ever had the mount command hang, requiring a reboot, except with btrfs. Maybe someday it will be ready. _________________ .sigs waste space and bandwidth |
|
Back to top |
|
|
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Fri Jun 05, 2015 2:12 pm Post subject: |
|
|
It is working fine for us, so far. I have unplugged the running PC while files are being used on the BTRFS partition and still no issue. I will do some more testing but it seems OK to me. _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
depontius Advocate
Joined: 05 May 2004 Posts: 3509
|
Posted: Fri Jun 05, 2015 3:16 pm Post subject: |
|
|
Things are generally OK until they're not. I'm just glad I had a full backup less than 12 hours old, at the time. _________________ .sigs waste space and bandwidth |
|
Back to top |
|
|
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Fri Jun 05, 2015 3:36 pm Post subject: |
|
|
I have two external drives hooked up, so we have backups. If BTRFS gives me trouble I will determine whether or not it is worth using. _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
depontius Advocate
Joined: 05 May 2004 Posts: 3509
|
Posted: Sun Jun 28, 2015 12:05 am Post subject: |
|
|
Next attempted systemd encroachment - actually it's a gnome encroachment, but that's effectively the same thing.
On the update check I just did, icedtea-bin-7.2.5.3 wants to upgrade to icedtea-bin-5.2.5.5, but that requires gsettings-desktop-schemas-3.14.2, which of course brings us into the systemd era.
I looked at the icedtea-bin ebuild, and see the following line:
Code: | # gsettings-desktop-schemas is needed for native proxy support. #431972 |
I'm not using native proxy support or any sort of proxy. I haven't yet taken the time to see if I can tweak the ebuild, perhaps turn off the proxy support.
I just checked gsettings-desktop-schemas-3.14.2.ebuild, and the problem appears to be that it wants >gdm-3.8, while I've got the pre-systemd gdm-2.20.11-r1 installed. For now, and to get through today's updates I'm going to mask the icedtea-bin update. In the longer run, I need to either look at tweaking my own copy of the gsettings-desktop-schemas ebuild, or migrate off of gdm. I've pretty much quit using xdcmp, so that may not be too hard any more.
I'm now looking at the gdm ebuilds, which I've been assuming require systemd. From what I can see, systemd appears to be an optional USE flag. I know others have stuck to gdm-2.20.11 like me, but maybe that limitation is not real.
For now I'm masking the icedtea-bin update, jsut so I can get everything else done. Then I think I'll take things one at a time, starting with either a gdm update, or switching to another display manager. _________________ .sigs waste space and bandwidth |
|
Back to top |
|
|
steveL Watchman
Joined: 13 Sep 2006 Posts: 5153 Location: The Peanut Gallery
|
Posted: Sun Jun 28, 2015 5:32 pm Post subject: |
|
|
The_Great_Sephiroth wrote: | Also, reinstalling and using UUIDs to mount fixed the issue. For some reason my kernel saw things differently than the live CD last time. Lesson learned, use UUIDs! |
Actually the more recent PARTUUID's are more robust, since they're kernel-side, so a) fixed until you wipe the partition at the table level, not fs, and b) available without any fs modules required, thus more useful for the early-boot phase.
Surprised the nubskool didn't jump all over them, but instead everyone bangs on about the FS UUID, which is much less "persistent", and how udev renames blah-blah.
Anyhow, I'd go for that, unless you have some odd setup. |
|
Back to top |
|
|
NeddySeagoon Administrator
Joined: 05 Jul 2003 Posts: 54237 Location: 56N 3W
|
Posted: Sun Jun 28, 2015 5:52 pm Post subject: |
|
|
steveL,
PARTUUIDS on MSDOS disklabels need to be used with care. Take for example Code: | /dev/sdb1: UUID="9392926d-6408-6e7a-8663-82834138a597" TYPE="linux_raid_member" PARTUUID="0553caf4-01" |
Here the PARTUUID="0553caf4-01" is the volume label (0553caf4) which is good until you write a new partition table, or change it yourself, followed by the partition number, here -01.
This is safe for partitions 1..4. After that it gets more interesting as logical partitions are stored as a linked list. This means that on a drive that has partitions 5, 6, 7 and 8, when you delete partition 5, partitions 6, 7 and 8 get renumbered to 5, 6 and 7, so PARTUUID breaks horribly but filesystem UUID continues to work.
MSDOS partition tables have always been a hack, since the time they were introduced, when drives reached 40Mb. (Yes, Megabytes.)
On GPT it just works. _________________ Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
Back to top |
|
|
depontius Advocate
Joined: 05 May 2004 Posts: 3509
|
Posted: Sun Jun 28, 2015 7:00 pm Post subject: |
|
|
NeddySeagoon wrote: |
MSDOS partition tables have always been a hack, since the time they were introduced, when drives reached 40Mb. (Yes, Megabytes.)
On GPT it just works. |
I didn't know the partuuid worked at all on MSDOS, if I really want it, I use GPT. _________________ .sigs waste space and bandwidth |
|
Back to top |
|
|
steveL Watchman
Joined: 13 Sep 2006 Posts: 5153 Location: The Peanut Gallery
|
Posted: Mon Jun 29, 2015 1:35 am Post subject: |
|
|
Ah OK, thanks jrg; so they're not quite as robust if you use the old-style MS-DOS disks.
Seems ok if you have a fixed partition scheme, and lvm on top of it, as you only shove space around via the device-mapper.
I'll stick with good old /dev/sda and motherboard chipset-drivers in the kernel.
IDK what it is, but I find that positively luxuriant compared to the old days of bindists with thousands of modules. |
|
Back to top |
|
|
The_Great_Sephiroth Veteran
Joined: 03 Oct 2014 Posts: 1602 Location: Fayetteville, NC, USA
|
Posted: Mon Jun 29, 2015 1:45 pm Post subject: |
|
|
I had not heard of PARTUUIDs before this thread. I only use GPT unless the system is so old it REQUIRES DOS partition tables. Did you know that some old boards even REQUIRE the partition to start on sector 63? If you put an advanced format disk in one of those systems you'd be hosed.
I will look into PARTUUIDs for my GPT system. Thanks for the info. _________________ Ever picture systemd as what runs "The Borg"? |
|
Back to top |
|
|
steveL Watchman
Joined: 13 Sep 2006 Posts: 5153 Location: The Peanut Gallery
|
Posted: Mon Jun 29, 2015 6:44 pm Post subject: |
|
|
Oops for some reason I though neddy was jrg; sorry about that.
GS: good one; informing you of the option is all I was after. |
|
Back to top |
|
|
NeddySeagoon Administrator
Joined: 05 Jul 2003 Posts: 54237 Location: 56N 3W
|
Posted: Mon Jun 29, 2015 6:48 pm Post subject: |
|
|
The_Great_Sephiroth,
Even in the dim and distant past, where it was worthwhile to edit the partition table by hand, to recover some of the "wasted" space, I've never come across a board that required the first partition to start at CHS 0,1,0.
Monterboards have no idea of the partition table on the disc, they all load CHS 0.0.0 into RAM and some broken BIOSes check for the signature word at the end of the partition table and the bootable flag on ore or more of the primary partitions in the position where the MSDOS partition table would be.
GTP provides all of there features except setting the bootable flag on the 'protective' MSDOS partition table it provides. You need to use Code: | fdisk -t dos /dev/sd... |
Advanced format drives fake 512b sectors to the operating system, so should still work. _________________ Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
Back to top |
|
|
John R. Graham Administrator
Joined: 08 Mar 2005 Posts: 10589 Location: Somewhere over Atlanta, Georgia
|
Posted: Mon Jun 29, 2015 6:55 pm Post subject: |
|
|
steveL wrote: | Oops for some reason I though neddy was jrg; sorry about that. | I'm flattered.
- John _________________ I can confirm that I have received between 0 and 499 National Security Letters. |
|
Back to top |
|
|
|