The Linux Questions Thread: a bunch of pitfalls, but technically it's possible

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Linux Questions Thread: a bunch of pitfalls, but technically it's possible

«‹›1023 »

Quaint Quail Quilt: Jun 19, 2006; Ask me about that time I told people mixing bleach and vinegar is okay

This may be in visualization thread territory, and is very much "cowboy" territory, but for all the gamers or people who would totally switch to linux except for games.

How about this? I know it's reddit, but It has me wanting to build something conceivably compatible with all OS's for this specific use case.
https://www.reddit.com/r/pcmasterrace/comments/2z0evz/gpu_passthrough_or_how_to_play_any_game_at_near/

https://www.reddit.com/r/pcmasterrace/comments/3lno0t/gpu_passthrough_revisited_an_updated_guide_on_how/
This is the same page as above in case the reddit goes down later http://bufferoverflow.io/?p=1

My understanding is that by dedicating your main GPU to windows, and (if you need a gui) using your onboard CPU's graphics for linux, you can finally achieve the holy grail of good performance windows running inside linux (via VM)
I'll be keeping appraised of this issue, hopefully this becomes very easy to do with certain hardware by the time I build my next computer.

No dual booting required, I think grub messed with me the most for some reason when I was installing gentoo as a linux newb like an insane person.
(also my current hardware doesn't even support Windows 8.1, Windows 10 broke some stuff and it's not very linux friendly hardware)

# ? Feb 12, 2016 10:59

Adbot: ADBOT LOVES YOU

# ? Apr 18, 2024 10:21

HPL: Aug 28, 2002; Worst case scenario.

Some motherboards, especially older ones, won't let you use both on-board and an add-on video card at the same time.

# ? Feb 12, 2016 15:04

evol262: Nov 30, 2010; #!/usr/bin/perl

galahan posted:

This may be in visualization thread territory, and is very much "cowboy" territory, but for all the gamers or people who would totally switch to linux except for games.

Yes, this works, though the idea of a "holy grail" system with a bunch of xrandr fuckery and your keyboard passed through is weird.

SecureBoot also wasn't enabled on the VM, and it should be (this is easy to do).

There's literally zero reason not to just put Windows in a VM with a passthrough device, and use Steam streaming to play your games. Even if there are artifacts (and there might be), you can just change the monitor input connected to the Windows VM and ignore the choppy graphics on Linux. But then audio is perfect and you don't need to pass through keyboards/mice/etc.

HPL posted:

Some motherboards, especially older ones, won't let you use both on-board and an add-on video card at the same time.

Passthrough is a non-starter without VT-d, and it's basically useless without an EFI system for VFIO. Yes, you can do it on legacy boot. But don't.

Boris Galerkin posted:

I'm having a really frustrating time trying to dual boot Windows 10 and CentOS 7. First I wiped the hard drive and then installed Windows 10. This automatically created (off the top of my head) the following partitions: 1) Windows recovery, 2) 100 MB EFI boot partition, 3) a 16 MB Windows partition, and 4) a 100 GB data partition. I left the rest of the hard drive un partitioned at this point. Windows booted fine etc. Then I installed CentOS 7 and created a swap, home, root, and what I thought was a EFI boot partition. Now CentOS boots fine, but Windows has disappeared from grub. From searching around I found one recent thread where some other guy had the exact same problem as me and basically everything he's saying is 100% what I'm experiencing.

You don't need an extended boot partition (the 16MB "windows" partition) if you're booting from EFI. Are you booting from EFI?

Also, since EFI is plain FAT32, the expectation is that everything will just share the same partition. EFI executables are small. CentOS does not need its own EFI partition.

A couple of things:

First is that "linux" isn't a command in EFI grub. "linuxefi" is.

Boris Galerkin posted:

I think what happened was that CentOS overwrote the EFI boot partition that Windows created and I can't get my system to boot back into Windows. I've edited the grub.d/40... Config file like that thread said. I installed os prober and ntfs-3g. I can mount my Windows partition and I can see all the files in there. I just can not boot into Windows.

Please don't make assumptions like this.

Again, because EFI is FAT32, this is ridiculously easy to check.

Check /boot/efi/EFI

You should have "/boot/efi/EFI/centos" and "/boot/efi/EFI/Microsoft". The default EFI bootloader (which is probably shim.efi for grub2) is under "/boot/efi/EFI/BOOT/BOOTX64.EFI" (you can md5sum it or something to be sure)

Boris Galerkin posted:

http://askubuntu.com/questions/22698/update-grub2-not-finding-windows7-partition

The first reply, he mentions something about os prober being confused about the Linux /boot and Windows /Boot folders and mentions a solution but I don't know how to remove /boot like he says.

This is not relevant, and don't follow any steps in that thread. You're on an EFI system. grub was not installed on the Windows partition (which was a hacky chainloading solution to use ntldr in the distant past).

So, here's what you do:

Check to see whether you really have two EFI system partitions (two FAT32 partitions).

If you do, mount them both (under /tmp/efi1 and /tmp/efi2, whatever).

Find out which one the Microsoft bootloader is in. It's probably not the same one as the centos bootloader. But these are plain files.

Copy the Microsoft bootloader to the EFI system partition you want (cp -rpv /tmp/efi1/Microsoft /tmp/efi2/). Or do it the other way around and change fstab so /boot/efi points at the right partition.

Run grub2-mkconfig again. os-prober-efi should now pick up the other bootloader.

Delete the other EFI system partition so you never make this mistake again.

Read this so you understand how EFI works.

# ? Feb 12, 2016 15:56

some kinda jackal: Feb 25, 2003; �
�

everythingWasBees posted:

Is there any good reason to use 6.7 over 7, besides that being the version the school uses? Or is it really not going to matter much.

I don't know what level of coding you're doing; there's always the risk that you'll have version mismatches on libraries if you're bringing binaries over, but that's likely to be the case if you're running different patch levels of 6.7 as well. I'd just go with the latest 7 release and call it a day. Everything else should be under-the-hood administrative differences. I can't think of any day to day use differences between any of my 6 and 7 VMs.

# ? Feb 12, 2016 16:01

evol262: Nov 30, 2010; #!/usr/bin/perl

Actually, the api in packaged libraries is guaranteed. You shouldn't have version mismatches unless you're building on 6.7 and running on 6.3 or something (and even then, the version should be the same, or the same symlink, just missing symbols, maybe).

There aren't a lot of day-to-day differences other than systemd and firewalld, plus it boots a lot faster. Most of the libraries from 6 should be available as compat-..., unless the school system has compat libraries from centos/RHEL 5 installed.

To match your school, I'd probably still use 6, but you'll be OK with 7

# ? Feb 12, 2016 16:07

some kinda jackal: Feb 25, 2003; �
�

Oh right, yeah it's been a while since I did any coding so I goofed there. Good to hear that patchlevels shouldn't affect.

# ? Feb 12, 2016 16:09

evol262: Nov 30, 2010; #!/usr/bin/perl

We intentionally keep it the same in major versions where possible (there was a systemd rebase in 7.2, and rebases happen every so often, but they're avoided if at all possible)

So libfoo-1.2.3-X should be libfoo-1.2.3-X+ for the entire release cycle. This is easily observable with the kernel, where it has the same version the whole way through a release cycle, with -release incrementing as fixes are backported from upstream

# ? Feb 12, 2016 16:17

Boris Galerkin: Dec 17, 2011; I don't understand why I can't harass people online. Seriously, somebody please explain why I shouldn't be allowed to stalk others on social media!

evol262 posted:

You don't need an extended boot partition (the 16MB "windows" partition) if you're booting from EFI. Are you booting from EFI?

The Windows 10 installer created the 16 MB partition automatically so I had no say in that, and yes.

evol262 posted:

Check /boot/efi/EFI

You should have "/boot/efi/EFI/centos" and "/boot/efi/EFI/Microsoft". The default EFI bootloader (which is probably shim.efi for grub2) is under "/boot/efi/EFI/BOOT/BOOTX64.EFI" (you can md5sum it or something to be sure)

code:

$ sudo ls -l /boot/efi/EFI
total 4
drwx------. 2 root root 2048 Feb 11 21:37 BOOT
drwx------. 3 root root 2048 Feb 12 08:26 centos

$ sudo ls -l /boot/efi/EFI/BOOT
total 1338
-rwx------. 1 root root 1296176 Dec  7 13:19 BOOTX64.EFI
-rwx------. 1 root root   73240 Dec  7 13:19 fallback.efi

I don't have a /boot/efi/EFI/Microsoft, that's the problem.

quote:

Check to see whether you really have two EFI system partitions (two FAT32 partitions).

If you do, mount them both (under /tmp/efi1 and /tmp/efi2, whatever).

Find out which one the Microsoft bootloader is in. It's probably not the same one as the centos bootloader. But these are plain files.

Copy the Microsoft bootloader to the EFI system partition you want (cp -rpv /tmp/efi1/Microsoft /tmp/efi2/). Or do it the other way around and change fstab so /boot/efi points at the right partition.

Run grub2-mkconfig again. os-prober-efi should now pick up the other bootloader.

Delete the other EFI system partition so you never make this mistake again.

Read this so you understand how EFI works.

I only see a single fat16 partition. I can not mount the 16 GB partition.

code:

$ sudo parted /dev/sda p
Model: ATA INTEL SSDSC2BF24 (scsi)
Disk /dev/sda: 240GB
Sector size (logical/physical): 512B/512B
Partition Table: gpt
Disk Flags: 

Number  Start   End    Size    File system     Name                          Flags
 1      1049kB  473MB  472MB   ntfs            Basic data partition          hidden, diag
 2      473MB   578MB  105MB   fat16           EFI System Partition          boot
 3      578MB   595MB  16.8MB                  Microsoft reserved partition  msftres
 4      595MB   105GB  104GB   ntfs            Basic data partition
 5      105GB   105GB  262MB   ext4
 6      105GB   110GB  5369MB  ext4
 7      110GB   113GB  2147MB  linux-swap(v1)
 8      113GB   215GB  102GB   ext4

Here's the relevant mount output:

code:

[root@multnomah ~]# mount
/dev/sda2 on /boot/efi type vfat (rw,relatime,fmask=0077,dmask=0077,codepage=437,iocharset=ascii,shortname=winnt,errors=remount-ro)
/dev/sda5 on /boot type ext4 (rw,relatime,seclabel,data=ordered)
/dev/sda6 on / type ext4 (rw,relatime,seclabel,data=ordered)
/dev/sda8 on /home type ext4 (rw,relatime,seclabel,data=ordered)

Again partitions 1-4 were created automatically by Windows, 5-8 were created 'automatically' by CentOS although I resized the partitions and changed the filesystem from xfs to ext4 where applicable but I don't think that matters.

If I mount /dev/sd4 (the Windows partition) then I can find a file called 'bootmgr.efi' and 'bootmgfw.efi' in the path '[sd4]/Windows/Boot/EFI' and I can copy bootmgfw.efi into where you said to copy it to:

code:

[root@multnomah ~]# mount /dev/sda2 /tmp/efitmp/
[root@multnomah ~]# mount /dev/sda4 /tmp/wintmp/

[root@multnomah ~]# mkdir /tmp/efitmp/EFI/Microsoft

[root@multnomah ~]# cp /tmp/wintmp/Windows/Boot/EFI/bootmgfw.efi /tmp/efitmp/EFI/Microsoft/.

But when I do grub2-mkconfig nothing regarding Windows is getting picked up:

code:

[root@multnomah ~]# grub2-mkconfig > /dev/null
Generating grub configuration file ...
Found linux image: /boot/vmlinuz-3.10.0-327.4.5.el7.x86_64
Found initrd image: /boot/initramfs-3.10.0-327.4.5.el7.x86_64.img
Found linux image: /boot/vmlinuz-3.10.0-327.el7.x86_64
Found initrd image: /boot/initramfs-3.10.0-327.el7.x86_64.img
Found linux image: /boot/vmlinuz-0-rescue-356521c4a2d04d97b7f9fa91f1736c3d
Found initrd image: /boot/initramfs-0-rescue-356521c4a2d04d97b7f9fa91f1736c3d.img
done

I did notice you say os-prober-efi specifically. I only have os-prober installed and it is the only thing that comes up when I do 'yum search os-prober'.

Boris Galerkin fucked around with this message at 20:51 on Feb 12, 2016

# ? Feb 12, 2016 20:45

evol262: Nov 30, 2010; #!/usr/bin/perl

Your earlier post read like you had two EFI partitions (which is ok, but not optimal). However...

This really looks/sounds like you installed Windows in Legacy mode. That extended boot partition and the fact that there's no Microsoft bootloader installed in the EFI system partition.

Are you sure Windows was installed in EFI mode? That executable cannot be directly copied from the Windows partition and work.

The EFI os prober expects an entry for the bootloader in the EFI variables. If you can't see it in efibootmgr, it won't work.

Try booting your system in legacy mode (you can set this preference from the UEFI firmware) and see what happens. If Windows comes up, it's in legacy mode, and you'll need to convert it to EFI, which should be easy, since you already have an EFI partition.

You can also try directly chainloading that EFI executable. It should be bootx64.efi, but you can chainload the other and see if it works.

evol262 fucked around with this message at 21:56 on Feb 12, 2016

# ? Feb 12, 2016 21:52

Docjowles: Apr 9, 2009

evol262 posted:

We intentionally keep it the same in major versions where possible (there was a systemd rebase in 7.2, and rebases happen every so often, but they're avoided if at all possible)

So libfoo-1.2.3-X should be libfoo-1.2.3-X+ for the entire release cycle. This is easily observable with the kernel, where it has the same version the whole way through a release cycle, with -release incrementing as fixes are backported from upstream

There was a breaking change with OpenSSL (I want to say between CentOS 6.4 and 6.5?) where the version was bumped to 1.0 from 0.9. Which sucked rear end because you could no longer use any packages that targeted any earlier version of CentOS and everything depends on OpenSSL.

But other that that one time which I am still really salty about, RHEL is good about compatibility in minor upgrades!

# ? Feb 12, 2016 23:28

evol262: Nov 30, 2010; #!/usr/bin/perl

I don't remember this. There was a bump for heartbleed, but I think the update over the RHEL 6 cycle was from 1.0.0 to 1.0.1 in 6.4 or something. IIRC, openssl 0.9 has been shiped as openssl098e for the whole cycle, but it's been a long time...

# ? Feb 12, 2016 23:42

Combat Pretzel: Jun 23, 2004; No, seriously... what kurds?!

galahan posted:

This may be in visualization thread territory, and is very much "cowboy" territory, but for all the gamers or people who would totally switch to linux except for games.

Been there done that. Plenty of voodoo with the QEMU command-line to get it to work with maximum performance. Doesn't help that the documentation isn't forthcoming, like what all the cmdline options on the virtual devices do and such.

It's nice from a cause-I-can perspective, but you might as well dualboot.

# ? Feb 12, 2016 23:50

BlackMK4: Aug 23, 2006; wat.; Megamarm

I'm having a weird issue with a cron job.

The crontab entry for root:

code:

0 23 * * * /home/ubuntu/ec2-automate-backup -v vol-b30b901c -k 5 -p -r us-west-1 2>&1

The cron log for the job:

code:

Feb 12 23:00:01 ip- CRON[10854]: (root) CMD (/home/ubuntu/ec2-automate-backup -v vol-b30b901c -k 5 -p -r us-west-1 2>&1)

It's not actually working, though. I can run this command:

code:

ubuntu@name:/var/log$ sudo /home/ubuntu/ec2-automate-backup -v vol-b30b901c -k 5 -p -r us-west-1 2>&1
Snapshots taken by ec2-automate-backup will be eligible for purging after the following date (the purge after date given in seconds from epoch): 1455750097.
Tagging Snapshot snap-cfb0b4ed with the following Tags: Key=CreatedBy,Value=ec2-automate-backup Key=PurgeAfterFE,Value=1455750097 Key=PurgeAllow,Value=true
Snapshot Purging is Starting Now.

and it runs fine. The file is:

code:

ubuntu@name:~$ ls -al
-rwxrwxrwx 1 ubuntu ubuntu  11K Feb  4 00:37 ec2-automate-backup

What am I doing wrong here?

BlackMK4 fucked around with this message at 05:56 on Feb 13, 2016

# ? Feb 13, 2016 00:05

Cidrick: Jun 10, 2001; Praise the siamese

BlackMK4 posted:

What am I doing wrong here?

Is there anything in your .profile or .bashrc that set up environmental variables that your script needs to run? Or perhaps something in an .s3cmd folder?

9 times out of 10 the reason a cron won't fire when it runs from the command line, it's an environmental variable not existing or the path to the executable isn't fully defined (which yours is, so it's not that)

Edit: Assuming this is your script, it looks like it's using Amazon's EC2 command line tools. And their docs seem to suggest that your AWS key and secret are kept in your .bashrc.

Edit2: Yay!

Cidrick fucked around with this message at 00:13 on Feb 13, 2016

# ? Feb 13, 2016 00:09

BlackMK4: Aug 23, 2006; wat.; Megamarm

Ah, holy poo poo. You're completely right. I forgot to copy my aws config to /root from /home/ubuntu

# ? Feb 13, 2016 00:12

Docjowles: Apr 9, 2009

evol262 posted:

I don't remember this. There was a bump for heartbleed, but I think the update over the RHEL 6 cycle was from 1.0.0 to 1.0.1 in 6.4 or something. IIRC, openssl 0.9 has been shiped as openssl098e for the whole cycle, but it's been a long time...

I did have the details wrong, because yeah, it was a long time ago

The TLDR is this thread (not posted by me, just illustrates the same problem): http://serverfault.com/questions/563051/installing-a-third-party-application-package-on-centos-6-4-fails-due-to-missing

For ~manager reasons~ we couldn't upgrade off of 6.4 right away, but wanted to keep up to date on certain third party packages. But those third parties started building on CentOS 6.5 (and beyond) and we could no longer install the updates because the required libraries weren't present. This one OpenSSL upgrade is the only time I can think of that a .x release made packages totally unusable on existing systems of the same major version, although there's probably others I am unaware of.

Docjowles fucked around with this message at 03:07 on Feb 13, 2016

# ? Feb 13, 2016 03:04

evol262: Nov 30, 2010; #!/usr/bin/perl

The systemd rebase in 7.2 is the other biggy, though not a lot of stuff codes against it.

There's lots of optional channels/products which include other versions (cloud-init changes frequently), but I'm glad I don't remember the openssl swap

# ? Feb 13, 2016 04:33

Jo: Jan 24, 2005; Soiled Meat

I've got an HP Proliant DL580 with a serial terminal at 9600-baud. I can get as far (I think) as booting off a USB stick made with the Universal USB Installer. When the installed kicks in, though, it looks like the serial terminal drops out. I've tried modifying the required files recommended here: http://pcengines.info/forums/?page=post&id=E25612E9-84F0-4DCF-A876-1E92FD1D065C but it looks like I might still be dropping to VGA.

quote:

isolinux.cfg:
# D-I config version 2.0
CONSOLE 0
SERIAL 0 9600 0
include menu.cfg
default vesamenu.c32
prompt 0
timeout 0

txt.cfg:
default install
label install
menu label ^Install Ubuntu Server
kernel /install/vmlinuz
append file=/cdrom/preseed/ubuntu-server.seed vga=788 initrd=/install/initrd.gz -- console=ttyS0,9600n8 quiet �

syslinux.cfg:
# D-I config version 2.0
CONSOLE 0
SERIAL 0 9600 0

default menu.c32
prompt 0
menu title UNetbootin
timeout 100

label unetbootindefault
kernel /install/netboot/ubuntu-installer/amd64/linux
append initrd=/install/netboot/ubuntu-installer/amd64/initrd.gz tasks=standard pkgsel/language-pack-patterns= pkgsel/install-language-support=false vga=788 -- console=ttyS0,9600n8 -- quiet

Any guesses about how I can get the system installed when I don't have a keyboard or monitor I can hook up? What's the lowest :effort:

way?

Jo fucked around with this message at 07:56 on Feb 17, 2016

# ? Feb 17, 2016 07:47

ToxicFrog: Apr 26, 2008

Jo posted:

I've got an HP Proliant DL580 with a serial terminal at 9600-baud. I can get as far (I think) as booting off a USB stick made with the Universal USB Installer. When the installed kicks in, though, it looks like the serial terminal drops out. I've tried modifying the required files recommended here: http://pcengines.info/forums/?page=post&id=E25612E9-84F0-4DCF-A876-1E92FD1D065C but it looks like I might still be dropping to VGA.

Any guesses about how I can get the system installed when I don't have a keyboard or monitor I can hook up? What's the lowest way?

Is it networked? Does the Ubuntu Server installer support piloting it over ssh? I know the SUSE one does.

# ? Feb 17, 2016 10:28

evol262: Nov 30, 2010; #!/usr/bin/perl

Jo posted:

I've got an HP Proliant DL580 with a serial terminal at 9600-baud. I can get as far (I think) as booting off a USB stick made with the Universal USB Installer. When the installed kicks in, though, it looks like the serial terminal drops out. I've tried modifying the required files recommended here: http://pcengines.info/forums/?page=post&id=E25612E9-84F0-4DCF-A876-1E92FD1D065C but it looks like I might still be dropping to VGA.

Any guesses about how I can get the system installed when I don't have a keyboard or monitor I can hook up? What's the lowest way?

iLO. iLO. iLO

# ? Feb 17, 2016 13:59

some kinda jackal: Feb 25, 2003; �
�

evol262 posted:

iLO. iLO. iLO

I'm can't remember how I did servers without OOB. Or maybe I'm just blocking it out of my memory.

some kinda jackal fucked around with this message at 14:34 on Feb 17, 2016

# ? Feb 17, 2016 14:32

Docjowles: Apr 9, 2009

Martytoof posted:

I'm can't remember how I did servers without OOB. Or maybe I'm just blocking it out of my memory.

physically walking up to the thing and putting a CD-ROM in the drive :rip:

# ? Feb 17, 2016 15:23

RFC2324: Jun 7, 2012; http 418

Martytoof posted:

I'm can't remember how I did servers without OOB. Or maybe I'm just blocking it out of my memory.

My home server has a finicky as hell iLO, due to an old embedded version of java.

Its still worth fighting with it for an hour to get it running when I need to get in instead of plugging in peripherals.

# ? Feb 17, 2016 18:45

Boris Galerkin: Dec 17, 2011; I don't understand why I can't harass people online. Seriously, somebody please explain why I shouldn't be allowed to stalk others on social media!

What's a good email client and pdf annotator?

For email the only thing I really know about is Thunderbird and I downloaded it and it looks like poo poo. I guess there's N1 that looks nice and Mac-like, but has anyone actually used it?

For the pdf bit, I spent about an hour looking into it today and it seems like my only choice is ocular if I want annotations (highlighting text, dropping in notes, don't necessarily care about putting shapes and stuff into the pdf) but my only concern is I'm using gnome desktop so if I could avoid installing half of a kde desktop environment just to run a pdf viewer then that would be great. There was one I found that started with a z- but from what I could tell people didn't like this one because the annotations existed "on top" of the pdf rather than "with" the pdf? Apparently editing pdfs is a hard thing to do or something, why? I guess I'm a bit spoiled with OS X's 'Preview' which is a native app that has pretty good built in pdf editing.

# ? Feb 17, 2016 22:46

Jo: Jan 24, 2005; Soiled Meat

ToxicFrog posted:

Is it networked? Does the Ubuntu Server installer support piloting it over ssh? I know the SUSE one does.

Yes, it's networked, but there's no OS on there at the moment to SSH into.

evol262 posted:

iLO. iLO. iLO

I was going to say, "Looks like iLO needs a license." but HP's website says there's one included with the server. :woop:

# ? Feb 18, 2016 00:33

some kinda jackal: Feb 25, 2003; �
�

Jo posted:

I was going to say, "Looks like iLO needs a license." but HP's website says there's one included with the server.

You'd be shocked how many people try to sell iLO licenses on eBay by posting a photo of the code.

# ? Feb 18, 2016 00:48

ToxicFrog: Apr 26, 2008

Jo posted:

Yes, it's networked, but there's no OS on there at the moment to SSH into.

Yes, I know, but the installer is itself a linux livesystem. The SUSE installer, at least, lets you pre-configure the install media with network settings and an SSH password; insert install media, boot, wait a few minutes, then ssh into it from another computer and the installer starts up automatically. You run the install remotely over ssh and then reboot into the installed system. I was wondering if Ubuntu Server had something similar, since it's extremely useful.

# ? Feb 18, 2016 15:07

HPL: Aug 28, 2002; Worst case scenario.

Normally I use Xfce out of necessity because I need something light and compact, but Xfce has always kind of been lacking in the looks department. The Manjaro folks have done a really nice job of actually making it look good.

EDIT: Holy crap is this slow. I can't get anything done. Back to Fedora/LXDE.

HPL fucked around with this message at 05:04 on Feb 19, 2016

# ? Feb 19, 2016 04:35

Boris Galerkin: Dec 17, 2011; I don't understand why I can't harass people online. Seriously, somebody please explain why I shouldn't be allowed to stalk others on social media!

Boris Galerkin posted:

What's a good email client and pdf annotator?

For email the only thing I really know about is Thunderbird and I downloaded it and it looks like poo poo. I guess there's N1 that looks nice and Mac-like, but has anyone actually used it?

For the pdf bit, I spent about an hour looking into it today and it seems like my only choice is ocular if I want annotations (highlighting text, dropping in notes, don't necessarily care about putting shapes and stuff into the pdf) but my only concern is I'm using gnome desktop so if I could avoid installing half of a kde desktop environment just to run a pdf viewer then that would be great. There was one I found that started with a z- but from what I could tell people didn't like this one because the annotations existed "on top" of the pdf rather than "with" the pdf? Apparently editing pdfs is a hard thing to do or something, why? I guess I'm a bit spoiled with OS X's 'Preview' which is a native app that has pretty good built in pdf editing.

Ended up uninstalling Thunderbird completely and just resorting to webmail for now, and as far as the pdf annotation goes, okular had to go because the annotations weren't saved to/visible to other pdf readers which was a big problem for me. I ended up just going with Foxit, seems to work decently well.

So this is more of a networking question but here in the office I need to use a proxy if I want to use the LAN. There's a proxy setting in CentOS but it looks like I can't set it per network so I have to manually disable the proxy if I were to take my laptop home or connect to a wifi network instead of the LAN. Is there a way to automate this?

# ? Feb 21, 2016 12:23

evol262: Nov 30, 2010; #!/usr/bin/perl

nm-dispatcher+proxydriver

# ? Feb 21, 2016 14:42

Thermopyle: Jul 1, 2003; ...the stupid are cocksure while the intelligent are full of doubt. �Bertrand Russell

On phone so google it yourself, but Linux Mint site got hacked (running WordPress, lol) and the hackers replaced the ISO downloads with their own compromised version.

# ? Feb 21, 2016 16:51

HPL: Aug 28, 2002; Worst case scenario.

Thermopyle posted:

On phone so google it yourself, but Linux Mint site got hacked (running WordPress, lol) and the hackers replaced the ISO downloads with their own compromised version.

I guess if there's a silver lining, it's that Mint is largely a desktop distro, so most of the people affected were individual users. Imagine if RHEL got hacked.

# ? Feb 21, 2016 21:07

evol262: Nov 30, 2010; #!/usr/bin/perl

Everything we distribute is gpg signed, and the primary signing key owner is incredibly paranoid.

redhat.com being hacked would be annoying. They'd also need to break the CDN infrastructure (to redistribute modified ISOs), and figure out/break into the signing infrastructure to resign/distribute modified packages. Mint didn't suffer vulnerable packages either, so let's limit it to that...

Sure, if they broke into the CDN, they could edit ISOs. Anaconda is pretty smart, but a trivial systemd service that runs at shutdown could find the installed root and dump their package in it. If they were really good, they'd also add an selinux policy which started it at boot.

But the images from the CDN (and the customer portal) are URLs which include the sha of the image. I don't honestly know if it's even possible to touch the CDN (or upload images to it/modify it) without kerberos. So they'd need to also break that. And the download page is populated dynamically based on hashes of releases. So they'd need to break that system.

It would need to be a lot more sophisticated.

CentOS or Debian are larger worries (suse's release engineering is also pretty complex), but neither one is running vulnerable applications on the same server as the images are hosted on.

# ? Feb 21, 2016 22:41

Mr Shiny Pants: Nov 12, 2012

I just upgraded to Debian stretch because I needed the newest QEMU bits. The problem is that NFS performance seems to have tanked from Jessie to Stretch. Write performance on my infiniband cards went from 200MB/Sec to 118K Sec....

My NAS is still the same so I am blaming the Linux NFS stuff. I've tried various options, (noatime, async vers=3 etc etc. ) but it is still crap and I could not find anything that would make it better.

So I've tried mounting NFS over RDMA and this works and gets me 400 MB/Sec on an async zpool. Yay!

One thing that is not clear to me: Do I need xprtrdma.ko? This seems to be the Sun RDMA RPC Driver, but Stretch uses RPCRDMA and this works fine.

Can I just keep using RPCRDMA or do I also need the other stuff? The SUN driver gets referenced a lot in some older faqs and stuff.

If someone could shed some light on this I would be grateful.

# ? Feb 21, 2016 22:55

evol262: Nov 30, 2010; #!/usr/bin/perl

AFAIK, Mellanox cards require it, but it's easy just to see if the driver is in use.

# ? Feb 22, 2016 16:16

HPL: Aug 28, 2002; Worst case scenario.

What's the advantage of file systems like GlusterFS or Ceph versus NFS, DFS or whatever else regular systems use?

# ? Feb 23, 2016 01:31

evol262: Nov 30, 2010; #!/usr/bin/perl

Gluster and ceph are not orthogonal to NFS.

DFS is similar in that it's a distributed filesystem. Ceph and gluster are alternatives to lustre, and, broadly, to traditional fileservers and SANs (though ceph rbd can be used in a similar way, conceptually).

The advantages are resiliency, geo-replication, easy tie-in with many virt products, and performance (using native clients, or particularly if you have a lot of storage nodes).

The disadvantage is that you're essentially implementing a network-level RAID, probably with local disk controllers to worry about. Also, performance is strictly worse than single purpose fileservers if you don't have many nodes, and special purpose storage systems often blow them out of the water.

I'd suggest you read about distributed filesystems (gluster, cephfs, dfs) vs network filesystems (NFS, cifs) and where they overlap. Distributed storage architectures (ceph rbd, many commercial SAN vendors which allow adding extra shelves/heads which talk to others) are an exercise for the reader.

# ? Feb 23, 2016 03:26

Cingulate: Oct 23, 2012; by Fluffdaddy

I have a static IP somewhere. I want to be able to have a permanently open ssh connection that appears to the remote as coming from that static IP, but actually comes from some arbitrary place. (There won't be a system permanently available from the static IP.)
Is this possible?

I'll probably have to wire up something less obviously retarded than this in the long term, but for now, I'm just wondering if this would work.

# ? Feb 23, 2016 15:28

evol262: Nov 30, 2010; #!/usr/bin/perl

Cingulate posted:

I have a static IP somewhere. I want to be able to have a permanently open ssh connection that appears to the remote as coming from that static IP, but actually comes from some arbitrary place. (There won't be a system permanently available from the static IP.)
Is this possible?

I'll probably have to wire up something less obviously retarded than this in the long term, but for now, I'm just wondering if this would work.

No. This doesn't even work conceptually. You cannot have a permanently open stateful connection if there's no system permanently available.

If there's no system available behind it (or VPN gateway which from which you can get behind it), it doesn't work, either.

What are you trying to do with this? There's probably a better way.

# ? Feb 23, 2016 15:41

Adbot: ADBOT LOVES YOU

# ? Apr 18, 2024 10:21

Cingulate: Oct 23, 2012; by Fluffdaddy

evol262 posted:

No. This doesn't even work conceptually. You cannot have a permanently open stateful connection if there's no system permanently available.

If there's no system available behind it (or VPN gateway which from which you can get behind it), it doesn't work, either.

What are you trying to do with this? There's probably a better way.

Ah okay. I can think of a few less insane ones, this would just be super lazy from my end.

# ? Feb 23, 2016 15:44

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Linux Questions Thread: a bunch of pitfalls, but technically it's possible

«‹›1023 »