|
pip install matplotlib failed with the following:quote:* The following required packages can not be built: I can't find anything via search.... edit: nm got it to work with apt-get Liam Emsa fucked around with this message at 04:37 on Feb 20, 2015 |
#
?
Feb 20, 2015 04:28
|
|
|
|
| # ? Jun 4, 2024 14:03 |
|
|
Megaman posted:It was a typo, I've since fixed it and it comes up with the same result, it can't detect the iso I can't tell the formatting from your post. Can you post /proc/cmdline from the installer?
|
|
#
?
Feb 20, 2015 04:40
|
|
|
evol262 posted:I can't tell the formatting from your post. Can you post /proc/cmdline from the installer? BOOT_IMAGE=/boot/grub/debian/vmlinuz file=/hd-media/preseed-work.cfg iso-scan/filename=/mini.iso locale=en_US keyboard-configuration/xkb-keymap=us languagechooser/language-name=English countrychooser/shortlist=US
|
|
#
?
Feb 20, 2015 04:48
|
|
|
Megaman posted:BOOT_IMAGE=/boot/grub/debian/vmlinuz file=/hd-media/preseed-work.cfg iso-scan/filename=/mini.iso locale=en_US keyboard-configuration/xkb-keymap=us languagechooser/language-name=English countrychooser/shortlist=US God,  ebian:Try asking on their lists. I'd try booting from "linux (loop)/...", but it's probably an artifact of Debian's terrible installer.
|
|
#
?
Feb 20, 2015 05:25
|
|
|
evol262 posted:God, I figured it out: code:
|
|
#
?
Feb 20, 2015 07:17
|
|
|
Megaman posted:I figured it out: Again, the Debian installer is incredibly stupid and nobody should use it. Also, all these ISO parameters are not standard. I suspect the preseed would need to be /cdrom/preseed.cfg, but again, you should ask someone on their list or an irc channel to explain this and why they make such terrible, barely-documented "features"
|
|
#
?
Feb 20, 2015 14:26
|
|
|
Megaman posted:I've downloaded the Debian unstable mini.iso installer and I can install SID just fine when I burn it to a USB key. Where did you get this? Last I checked there isn't an official one.
|
|
#
?
Feb 20, 2015 14:34
|
|
|
RFC2324 posted:Tried both this and rtorrent, transmission was ideal for running on a headless server. I've been running rtorrent on a headless server for years now with no issues, what makes transmission preferable?
|
|
#
?
Feb 20, 2015 14:40
|
|
|
wooger posted:Where did you get this? Last I checked there isn't an official one. http://ftp.debian.org/debian/dists/unstable/main/installer-amd64/current/images/netboot/ Not sure how that isn't standard, it's on their ftp.
|
|
#
?
Feb 20, 2015 14:48
|
|
|
evol262 posted:Again, the Debian installer is incredibly stupid and nobody should use it. Also, all these ISO parameters are not standard. I'll ask them. So if Debian installer is broken and no one should use it, are you claiming I shouldn't use Debian? I don't understand what you're getting at by saying that, you can't install Debian it. If you're not saying that, then can you explain how everyone using Debian automates a sid install? Because that's what I'm trying to do. Or are you telling me no one on the planet automates a sid install? I'm a little confused.
|
|
#
?
Feb 20, 2015 14:54
|
|
|
ToxicFrog posted:I've been running rtorrent on a headless server for years now with no issues, what makes transmission preferable? The fact that it is designed to run as a daemon, and has easy management tools built in. rtorrent, unless you also install the separate rutorrent package, has no convenient management interface(transmission has a handy little web page it auto generates), and has to be run in a session manager like screen or tmux, while transmission has the transmission-daemon service explicitly designed for usage this way. I had to write up an init script by hand to manage it, and all of the pre-made scripts I googled had the same ugly kludge of running it in a screen and pretending that was a daemon. Both packages required some tweaking of config files, but the rtorrent package(which I got from EPEL) didn't appear to have any default config file, either auto generated or provided by the install package, which made setting it up to work headless much more of a pain, and looked like it would be more painful to maintain overall.
|
|
#
?
Feb 20, 2015 15:31
|
|
|
RFC2324 posted:The fact that it is designed to run as a daemon, and has easy management tools built in. rtorrent, unless you also install the separate rutorrent package, has no convenient management interface(transmission has a handy little web page it auto generates), and has to be run in a session manager like screen or tmux, while transmission has the transmission-daemon service explicitly designed for usage this way. I had to write up an init script by hand to manage it, and all of the pre-made scripts I googled had the same ugly kludge of running it in a screen and pretending that was a daemon. RFC2324 posted:Both packages required some tweaking of config files, but the rtorrent package(which I got from EPEL) didn't appear to have any default config file, either auto generated or provided by the install package.
|
|
#
?
Feb 20, 2015 16:00
|
|
|
Megaman posted:I'll ask them. So if Debian installer is broken and no one should use it, are you claiming I shouldn't use Debian? I don't understand what you're getting at by saying that, you can't install Debian it. If you're not saying that, then can you explain how everyone using Debian automates a sid install? Because that's what I'm trying to do. Or are you telling me no one on the planet automates a sid install? I'm a little confused. You can use Debian if you want to. It's a great distro. The Debian installer isn't broken, it's just arcane and terrible, like every old installer. Anaconda's not a lot better, but at least it's very well documented. 99% of people automating installs of Linux do so over PXE. If they can't do it over PXE because of network issues, they're shipping CDs to retail branches, attaching them over IPMI, or dd-ing them onto flash drives. What you're trying to do is automate a Debian install through a loopback-mounted ISO on a flash drive (instead of dd-ing it) and getting repeatedly bitten by the disastrously unclear semantics of the Debian installer when doing this. It's not a very common use case, and nobody can give you a better idea of what's wrong and how to fix it than the Debian community.
|
|
#
?
Feb 20, 2015 17:38
|
|
|
sm00th posted:Yes, rtorrent is harder to set up and requires some kludges to work on a headless machine, however there are tons of manuals/tutorials/howtos all over the interwebs that will give you a working setup in no time. Besides last time I checked rtorrent had a much smaller footprint than transmisson/deluge resource-wise which is a big deal if you want to run it on a raspberry pi or an older machine. I'm not saying rtorrent is bad, or anything like that, it just didn't meet my use case of an easy drop in for a torrent server my technically illiterate girlfriend/roommate can use. I wanted to put a stop to the fighting over bandwidth, and was having more than enough trouble fighting with the samba part(windows sharing never works easily for me, I just don't get it for some reason). Something I can easily drop in, set the download/watch folders up for, and forget about was my goal. I have a 12 core sun server with 56gig of ram sitting in my office, mostly idle, so resources are not a real issue.
|
|
#
?
Feb 20, 2015 17:44
|
|
|
I had a powerloss when updating all the packages using Pacman -Syu. When the system reboots it dumps / to readonly because of errors. How do I run fsck against / when it needs to be unmounted when fsck is running?
|
|
#
?
Feb 22, 2015 00:05
|
|
|
YouTuber posted:I had a powerloss when updating all the packages using Pacman -Syu. When the system reboots it dumps / to readonly because of errors. How do I run fsck against / when it needs to be unmounted when fsck is running? Rescue CD or pass fsck.mode=force on the kernel cmdline
|
|
#
?
Feb 22, 2015 00:08
|
|
|
evol262 posted:Rescue CD or pass fsck.mode=force on the kernel cmdline Ah, fsck wasn't the problem. Arch ARM changed cmdline.txt to require rw entered in manually on the kernel commandline. Looking up how to pass those arguments was helpful however. Is it safe to leave fsck.mode=force on the commandline or should it removed?
|
|
#
?
Feb 22, 2015 01:37
|
|
|
doublepost.
|
|
#
?
Feb 22, 2015 01:46
|
|
|
YouTuber posted:Ah, fsck wasn't the problem. Arch ARM changed cmdline.txt to require rw entered in manually on the kernel commandline. Looking up how to pass those arguments was helpful however. It's safe to leave if you're ok fscking on every single reboot.
|
|
#
?
Feb 22, 2015 02:46
|
|
|
I have an mdadm raid5 array with three WD Red 3Tb drives (1 6gb ext4 partition) and whenever I mount it the drives start making a clicking sound every second. It sounds like a head parking. I checked that the drives are set to the max 300s idle park time, and smart does not report an unusually large number for load/unload cycle count - about 80 on each drive vs 20 power cycles and start/stop counts each. So what the hell is going on? The noise alone drives me batty, and I don't want to damage the drives. The array is set to 512kb chunk size and the drives correctly report 512k logical/4M physical sector sizes. Should I try to set it to 4M chunk size? I just don't get it. e: OK, crazy idea but I just formatted it as NTFS instead of ext4 and the noise stopped. Whaaat the hell? This is Fedora 21 in case it matters. Precambrian Video Games fucked around with this message at 11:00 on Feb 24, 2015 |
|
#
?
Feb 24, 2015 10:55
|
|
|
Was the array resyncing? Are you sure you don't have a dying drive?
|
|
#
?
Feb 24, 2015 14:13
|
|
|
Clicking once per second certainly sounds like a dying drive.
|
|
#
?
Feb 24, 2015 14:44
|
|
|
I wouldnt run raid5 with disks that size, takes way too long to rebuild in case of failure.
|
|
#
?
Feb 24, 2015 14:45
|
|
|
My ultimate goal is to change my disk scheduler to BFQ. However, I seem to be unable to even set it to CFQ or find out what scheduler it's currently using, if any. Running "cat /sys/block/sda/queue/scheduler" returns "none" instead of something like "deadline noop [cfq] bfq", which is what I expect. Doing "echo cfq > /sys/block/sda/queue/scheduler" doesn't work (permission denied, even via sudo" and doing "sudo tee /sys/block/sda/queue/scheduler <<< cfq" doesn't work (cat shows it's still just "none"). I'd like to know why my disk scheduler says "none" and how to fix it. I'm running Ubuntu 14.04 with liquorix kernel 3.18, if that helps.
|
|
#
?
Feb 24, 2015 17:25
|
|
|
Vinigre posted:My ultimate goal is to change my disk scheduler to BFQ. However, I seem to be unable to even set it to CFQ or find out what scheduler it's currently using, if any. Try "elevator=somescheduler" from the cmdline. But since you're running weird kernels with "cool" names like it's Android instead of your distribution's kernel or any of the kernels from upstream (including -mm), I'd suggest that you either figure out who maintains "liquorix" and ask them what schedulers are included or look at the config (probably in /proc/config.gz, but they may have disabled that, too) and see what schedulers it's built with. Also note that this was a kernel behavior change around 3.12 or 3.13. If you're running a virtualized system, "none" is normal. blk-mq does the same thing, and if you're running a system where scsi blk-mq comes into play (like SSDs), "none" will also be the default. You can rebuild without CONFIG_SCSI_MQ_DEFAULT set if you want the old behavior back, but you'll probably lose performance.
|
|
#
?
Feb 24, 2015 18:36
|
|
|
Crosspost from the Ubuntu thread: So I have a desktop and a laptop that both have Ubuntu. I have a folder shared on the desktop. I can browse to it on the laptop, but when I try to drag a folder to it, it won't let me. What settings do I need to change where?
|
|
#
?
Feb 24, 2015 19:16
|
|
|
Liam Emsa posted:Crosspost from the Ubuntu thread: Is it shared with SMB or NFS? When you're browsing, what are you browsing with? Does it prompt you for authentication? You're probably coming in as nfsnobody or whatever's mapped to "nobody" in Samba, which doesn't have permissions to write to the shared folder. Fix the permissions.
|
|
#
?
Feb 24, 2015 19:36
|
|
|
evol262 posted:Try "elevator=somescheduler" from the cmdline. But since you're running weird kernels with "cool" names like it's Android instead of your distribution's kernel or any of the kernels from upstream (including -mm), I'd suggest that you either figure out who maintains "liquorix" and ask them what schedulers are included or look at the config (probably in /proc/config.gz, but they may have disabled that, too) and see what schedulers it's built with. It looks like it's blk-mq in my case. The config shows that deadline, noop, cfq, and bfq are all included, and that "CONFIG_SCSI_MQ_DEFAULT=y". However, only my sdb is an SSD, and this problem applies to both my sda and sdb. Is that intended? Is there some way I can get the old behavior back for at least my spinning disk drive without doing a kernel rebuild?
|
|
#
?
Feb 24, 2015 19:37
|
|
|
Vinigre posted:It looks like it's blk-mq in my case. The config shows that deadline, noop, cfq, and bfq are all included, and that "CONFIG_SCSI_MQ_DEFAULT=y". However, only my sdb is an SSD, and this problem applies to both my sda and sdb. Is that intended? Is there some way I can get the old behavior back for at least my spinning disk drive without doing a kernel rebuild? blk-mq is better in every single way. There is no reason to want the old behavior back. And no, you can't get it back without recompiling.
|
|
#
?
Feb 24, 2015 21:18
|
|
|
evol262 posted:blk-mq is better in every single way. There is no reason to want the old behavior back. I'd like if I could get some solid numbers comparing blk-mq to the other schedulers on both SSD and HDD, specifically using BFQ, before I accept that claim. Unfortunately, I can't seem to find any. This report seems to indicate that this may be unintended? And contains an anecdote where blk-mq did worse, which prompts my hesitation to continue using blk-mq for the time being.
|
|
#
?
Feb 24, 2015 22:39
|
|
|
Vinigre posted:I'd like if I could get some solid numbers comparing blk-mq to the other schedulers on both SSD and HDD, specifically using BFQ, before I accept that claim. Unfortunately, I can't seem to find any. Bluntly, test it yourself or ask on the lkml. I don't work on the kernel, and I can't give you comparison numbers. There's an absolute ton of whitepapers and benchmarks which you can trivially find by looking up "blk-mq benchmark" on your search engine of choice. In general, it's also dramatically faster for spinning media as long as NCQ is supported (i.e. SATA 2), and it's much, much worse if you're trying to directly write to a device with no command scheduling. Like dd-ing flash drives, though filesystem access in flash drives is better. It may or may not have been an intentional change in whatever kernel you're running. Again, you should ask whoever maintains it. But also, in general, reports like "I'm seeing problems with kernel configuration A but not B" need a hell of a lot more information then supplied in order to make a value judgment on anything (and should be reported on bugzilla.kernel.org anyway, since Canonical doesn't have the time, desire, or engineers to fix kernel issues), and almost certainly won't affect you anyway. You're spending all this effort trying to optimize something that's almost guaranteed to not be a problem in your environment anyway. The point of blk-mq and scsi-mq is using multiple cpus and increasing performance. As long as the block layer can keep up, it's much lower CPU usage and much faster. Really bluntly: if the block layer can't keep up, you wouldn't be running a cleverly-named "enthusiast" kernel build on a system with one spinning drive and one SSD, because you'd be doing serious things.
|
|
#
?
Feb 24, 2015 23:17
|
|
|
evol262 posted:Was the array resyncing? Are you sure you don't have a dying drive? it was long since done resyncing. These are brand new drives and they all passed extended smart tests so I don't see any evidence that they're failing. I mounted a couple of them on their own earlier and they made no such sounds, and they're not making any unusual noises formatted as NTFS. Is there any reason why I shouldn't stick with NTFS? I've never had issues with Linux NTFS support. For what it's worth, I tried benchmarking the drive and it starts at around 250mb/s write 70 read, but after about a minute it drops to 150 and 50. I'm not sure if that's typical.
|
|
#
?
Feb 24, 2015 23:29
|
|
|
evol262 posted:Is it shared with SMB or NFS? When you're browsing, what are you browsing with? Does it prompt you for authentication? -I don't know. -File explorer into the network shared folders -No it doesn't
|
|
#
?
Feb 24, 2015 23:46
|
|
|
Vinigre posted:My ultimate goal is to change my disk scheduler to BFQ. However, I seem to be unable to even set it to CFQ or find out what scheduler it's currently using, if any. Ok, so I've finally solved my problem. In short, when scsi-mq is enabled by default, you can't choose what scheduler you use. The kernel I was using was compiled this way. To fix it without recompiling, add "scsi_mod.use_blk_mq=N" to the list of arguments for the kernel when booting. evol262 posted:you can't get it back without recompiling. Let it be known that you totally can. While I appreciate your explanation on why blk-mq is great, I simply can't find enough (any, actually) info out there comparing it to what I already use (BFQ), which suits my needs just fine. I'll definitely look into it later when distros start enabling it by default.
|
|
#
?
Feb 25, 2015 00:54
|
|
|
Vinigre posted:Ok, so I've finally solved my problem. In short, when scsi-mq is enabled by default, you can't choose what scheduler you use. The kernel I was using was compiled this way. To fix it without recompiling, add "scsi_mod.use_blk_mq=N" to the list of arguments for the kernel when booting. As with most things, it's best to do your own set of benchmarks before trusting that riced up kernels are actually useful or necessary. To answer your question about blk-mq, the whole point of that is to allow for high throughput by parallelizing the block queuing layer (currently linux only has a single queue which can be a huge bottleneck as you might imagine). As such, it operates at a different level than the IO scheduler which operates in/on the request queue. The reason that blk-mq currently lacks any IO scheduling is that for the workloads that it's targeting at its primary goal, IO scheduling would only add latency with little to no benefit. As it currently stands you should probably not be using blk-mq for any non solid state devices. It's possible that scheduling support will be added at a later time however. TLDR. If you were ever annoyed with the default linux block IO subsystem for maxing out around 1M IOPS, then blk-mq is definitely for you. This is also likely why a riced out kernel meant for "desktop, multimedia, and gaming workloads" use enabled it by default.
|
|
#
?
Feb 25, 2015 03:48
|
|
|
Why would flash work with Firefox but not Chromium?
|
|
#
?
Feb 25, 2015 04:56
|
|
|
Liam Emsa posted:Why would flash work with Firefox but not Chromium? You probably need to install the Pepper Flash plugin - I think Chromium no longer supports using the NPAPI version that Firefox uses.
|
|
#
?
Feb 25, 2015 05:23
|
|
|
Is there any way in the cron system to make a job run 1 hour after it ran the day before? I need to use cron settings for Jenkins for a 24 hour QA test but I think I might have to write a script OR just setup 7 different settings that start the test an hour after it started the day before and skip one day so it doesn't get stuck in some weird loop. Any ideas?
|
|
#
?
Feb 25, 2015 14:28
|
|
|
Virigoth posted:Is there any way in the cron system to make a job run 1 hour after it ran the day before? I need to use cron settings for Jenkins for a 24 hour QA test but I think I might have to write a script OR just setup 7 different settings that start the test an hour after it started the day before and skip one day so it doesn't get stuck in some weird loop. Any ideas? Is there a reason your cronjob couldn't "crontab -l > /tmp/crontab.$(date) && #edit run times && crontab < /tmp/crontab.$(date)"? Really, though using a different scheduler is probably better.
|
|
#
?
Feb 25, 2015 16:54
|
|
|
|
| # ? Jun 4, 2024 14:03 |
|
|
CaptainSarcastic posted:You probably need to install the Pepper Flash plugin - I think Chromium no longer supports using the NPAPI version that Firefox uses. And Adobe don't package it, so you'll have to install Chrome and point Chromium to the pepper flash .so file, or else grab it from an unofficial repo depending on your OS.
|
|
#
?
Feb 25, 2015 17:49
|
|