/dev/chat - 2014-08-13 (page 1 of 2)

« first day (1404 days earlier) ← previous day next day → last day (3552 days later) »

12:21 AM

@derobert bleh, most LC_COLLATE settings mix uppercase and lowercase

@FaheemMitha but sadly abandoned mid-story @derobert

@FaheemMitha occasionally. Finding reliable references can be a lot of work.

12:41 AM

Can anyone help me find a official/minor for Ubuntu studio 11.10 - 7.10?

12:54 AM

@Andrew ?

what do you mean by ?

I don't understand what you're asking for.

Who I forget download

can't find any thou...

???

And I want something that is sable, and I can trust.

12:56 AM

are you perhaps looking for old-releases.ubuntu.com/releases

8

A: Where can I download the latest release of Ubuntu?

On the Windows Operating System the Table may not rader right! All Ubuntu releases can be found at the releases and/or old releases pages respectively. ┏━━━━━━━━━━━━━━━━━━━━━━━━┯━━━━━━━━━━━━━━━┯━━━━━━━━━━━━━━━┯━━━━━━━━━━━━━━━┓ ┃ Version of Ubuntu + Edition │ Directly │ Torrent...

I am making that for every offical fork of Ubuntu like Kubuntu....

@Andrew Kubuntu is not a fork

I know.... But I don't know how to spell the right word thou.

direvtive... something like that

derivative?

it's not a derivative either

Yes it is

ubuntu.com/about/about-ubuntu/derivatives

1:01 AM

no, it's not. they're misusing the word derivative (slightly).

notice that Kubuntu is under the list of "flavors".

it's a flavor because it uses the same repo.

Yea.... but flavors is not right in the sense of real English, and not Linuxish English.

a derivative uses different repos e.g. Mint, Elementary OS, gNewSense

But they are not official.

@Andrew yes it is...

@Andrew doesn't matter. they're still derivatives.

Ubuntu is not an official derivative of Debian. but it's still a derivative.

Okay but still, can you help me with finding downloads links?

Yea

1:03 AM

7 mins ago, by strugee

are you perhaps looking for http://old-releases.ubuntu.com/releases/

Yea they don't acrive for any other than K,X, and EDu.

All thou, I can find a lot of voodoo download links for it, but I can't trust their stableness, and how fast the download link is.

2 hours later…

3:28 AM

@Gilles Yes, most mix uppercase & lowercase. But I consider that sane sorting, in ls or almost anywhere else.

(Be back later, need to go shut down my server/firewall to replace a disk.)

3:40 AM

BTW: error is probably another burnination candidate

4:34 AM

@strugee They're referred to as derivatives, whether they are actually that or not ;)

1 hour later…

5:36 AM

"It appears you've just replaced a hard disk to repair your degraded RAID array. Would you like to add it? Ha ha ha, just kidding, enjoy some fsck!"

2

6:20 AM

@Gilles Sadly abandoned, indeed.

@strugee That might be true.

@Gilles Good to know. Why SE in preference to WP? Because SE is less likely to be vandalized?

@derobert ?

Ok, RAID is messing with you?

6:34 AM

@strugee Months might be a slight exaggeration.

@FaheemMitha No, I just shut down the server to replace a failed disk. Then when it booted it, it decided to do half an hour of fscks before I could add the new disk to the arrays.

Doesn't quite have its priorities in order.

@derobert It should add, then fsck?

Up late or up early? Or do you wake up in the middle of the night like me?

@FaheemMitha Yeah, should worry about the rebuild before worrying about the every-x-days fsck

@derobert Oh, that's a routine fsck? Bug report?

Yeah, that would be a useful thing to add...

6:47 AM

@derobert What would? Bug report?

This is sw raid or hw raid?

@FaheemMitha It'd be useful to add some logic to the boot up fsck scripts to not do an fsck on a degraded array... But that logic is hard to get right.

And its mdraid

unix.stackexchange.com/review/suggested-edits/57879 ← very confused editor. Not only invalid edit, but suggests a question is a duplicate of itself

Good news though, no one voted to approve

@derobert Ok, but a bug report would do no harm. I guess a reproduction recipe is difficult?

@FaheemMitha Not really, the problem is figuring out the logic. And honestly, I'm not sure if a better option wouldn't just be to turn off the every X days fsck entirely.

I write bug reports all the time which people ignore. the trick is to forget about them immediately after you've sent them. I'm usually quite surprised to get responses to them, because I don't remember sending them.

@derobert Is this RAID specific?

@FaheemMitha In general... I'm not sure there is any real reason for routine fsck.

7:00 AM

@derobert I see. Not sure the people who put it in would take it out, regardless. People have fixed ideas about things.

Ok, off to lunch. Later, everyone.

1 hour later…

8:03 AM

Is there anyone on?

8:28 AM

Yeah... still hanging on...

What today topic of discussion???

I am here

2

Q: Where do I get all of the versions of Ubuntu + Flavors

This took me several days to do. If it doesn't render right, I made and test it for Firefox(with zoom 100%) version 31 since it comes default with Ubuntu. Please note, that after 14.10 comes out to stable, some of the links will be broken, and I didn't add PPC because it is dead on the desktop....

downloads flavors

Good?

Nice. Make it community plz. helpful stuff!!!

How do I do that?

I think you get an option to add as community stuff. Check the help!!!

Can I do it to the question?

@Rituraj I am not seeing a option too thou.

8:35 AM

@Andrew shouldn't you be asking this in Ubuntu chat?

Sorry. Tz called community wiki. Check out the below link

unix.stackexchange.com/help/privileges/community-wiki

faheem : Agreed. a bit off topic.

So.... :( Can some one do it for me?

Oh you only have 22 rep

:((

When did you hear about this website?

@Andrew what do you want done, exactly?

8:40 AM

Using the website as anonymous for a long time, yesterday got time to create a account.

Wow...

@Faheem : I am suugesting to make his link a

community wiki*

@FaheemMitha what is your rep on AskUbuntu?

@Andrew One sec.

400

I don't really use AU.

You only need 10000 - 400 rep

8:42 AM

I'm here, mostly.

@Andrew Not following.

What is the point of .... when Ubuntu can go in here just fine.

Its a perfect fit.

@Andrew Still not following. Is there something you wanted?

I want my question to turn into a community wiki post

3

Q: Where do I get all of the versions of Ubuntu + Flavors

This took me several days to do. If it doesn't render right, I made and test it for Firefox(with zoom 100%) version 31 since it comes default with Ubuntu. Please note, that after 14.10 comes out to stable, some of the links will be broken, and I didn't add PPC because it is dead on the desktop....

downloads flavors

@Andrew Ok. And how much rep is required for that? If you don't have the rep, I suggest just flagging it, and presumably a mod can take care of it.

10k

Why flagging it.... It will count agasin me.

8:46 AM

@Andrew Ok. Then ask on AU chat, or flag it for a mod. TBH, I'm not sure of the correct procedure here, but either of those sound reasonable.

Well Okay

@Andrew What? No, flag just means you are asking for attention. Or if you prefer, ping the mods in the chat room.

Assuming they are there, that is.

The AU mods, that is.

Okay

If it is a Community wiki post.... Whould I gain any rep if someone upvoted it?

@Andrew As far as I know, no. But nobody is forcing you to make it community wiki.

Okay

8:49 AM

I think people choose to make it community wiki so that it encourages other people to edit it. I'm not sure if it is really a useful concept.

Yea

As you probably know, if you edit a question a sufficient number of times, it automatically becomes community wiki.

Humm

3 hours later…

11:39 AM

@derobert don't you say ;)

@FaheemMitha nope, not anymore

1 hour later…

12:59 PM

@Braiam Oh, they changed the rules? Do you have a link?

1:18 PM

@terdon - you use cinnamon desktop but what's the underlying distro? The screensaver/desktop lock hasn't been working for me and is driving me nuts. I have a dual monitor setup too.

do you have something similar?

1:39 PM

Blog - Stack Exchange: Putting the Community back in Wiki

Grace Note on April 22, 2014

Ever seen this diagram?

That’s the visual elevator pitch for Stack Exchange. We were the little dot in the middle, a potent mix of useful traits from other tools, a wiry mutt full of hybrid vigor. The purpose of this blend was to allow and encourage the construction of a library of solutions, by providing communities with the tools they needed to share their experiences and challenges with others who might struggle with the same issues.

The diagram illustrated where we stole drew inspiration for the design of those tools, and their influence occasionally shows up in the results. Sometimes, …

@Braiam Thanks for the link. Probably a good change.

I've got some volume groups. there have been some posts (maybe only one) about finding out the underlying physical volumes for a volume group. Anyone got a link?

3

Q: How to find the physical volume(s) that hold a logical volume in LVM

I have a volume group (VG) that contains two physical volumes (PV). Several logical volumes (LV) in the VG are likely to use extents on both PVs. Is there a way to tell which LVs occupy space on which PVs?

on serverfault, but I thought there was something here.

Similarly how does one find what physical volumes are in a software raid, i.e. md0, md1?

2:34 PM

Ok, mdadm --detail has that information.

Anyone know a way, given say a physical device like /dev/sda to show what partitions it has. I'd have thought fdisk would work, but I'm not finding it.

@FaheemMitha, mount?

@Ramesh Er, what?

@FaheemMitha lsblk?

Never, mind, fdisk -l works. Not sure what I was thinking.

@Ramesh For what?

@Ramesh Oh, to show the structure? Yes, that is actually pretty handy, thanks.

3:06 PM

@FaheemMitha Not always.

root@Einstein:/home/anthony# fdisk -l /dev/sdb

WARNING: GPT (GUID Partition Table) detected on '/dev/sdb'! The util fdisk doesn't support GPT. Use GNU Parted.

@derobert I see.

@derobert does lsblk cope in this case?

root@Einstein:~# grep sdb /proc/partitions
   8       16 2930266584 sdb
   8       17     128488 sdb1
   8       18    9767520 sdb2
   8       19  146480670 sdb3
   8       20 1308759322 sdb4
 259        0        992 sdb128

that works, though, as long as Linux has the partition table read.

Anyone here aware of switching loop? I encountered a case where if I just connect the 2 ends of the network cable to the 2 wall ports, it will create a switching loop.

@Ramesh Yep. Great way to take out an Ethernet network that uses cheaper switches (more expensive ones will detect this and disable the ports, at least if configured to)

Sometimes you can even pull it off with a loopback plug!

@derobert, how does that work?

3:09 PM

@Gilles there you go meta.unix.stackexchange.com/q/3102/41104

@Ramesh You make a Ethernet plug that has the TX+ to RX+, TX- to RX-.

i.e, pin 1 to pin 3; pin 2 to pin 6

juniper.net/techpubs/en_US/junos13.3/topics/task/operational/…

They're actually useful for testing interfaces.

@derobert thanks. I am still confused on how the packets are transmitted in this case. I believe there is no IP protocol or something of that sort.

If am trying to create a loopback, can I trace what packets are being transmitted using tcpdump?

@Ramesh It's not an IP thing. A loopback plug works on the physical level.

You have physically connected the TX (transmit) pairs to the RX (receive) pairs.

Is there an easy way to see how many HD slots my mb has? I could look up the specs, but I'm lazy...

@FaheemMitha /sys/class/ata_port ?

3:21 PM

root@orwell:/sys/class/ata_port# ls
ata1 ata2 ata3 ata4 ata5 ata6 ata7 ata8 ata9

i guess i should just look at the mb specs

if you do an ls -l it'll show you where they are located...

@derobert Great, thanks.

lrwxrwxrwx 1 root root 0 Aug 7 06:08 ata1 -> ../../devices/pci0000:00/0000:00:11.0/ata1/ata_port/ata1
lrwxrwxrwx 1 root root 0 Aug 7 06:08 ata2 -> ../../devices/pci0000:00/0000:00:11.0/ata2/ata_port/ata2
lrwxrwxrwx 1 root root 0 Aug 7 06:08 ata3 -> ../../devices/pci0000:00/0000:00:11.0/ata3/ata_port/ata3
lrwxrwxrwx 1 root root 0 Aug 7 06:08 ata4 -> ../../devices/pci0000:00/0000:00:11.0/ata4/ata_port/ata4
lrwxrwxrwx 1 root root 0 Aug 7 06:08 ata5 -> ../../devices/pci0000:00/0000:00:11.0/ata5/ata_port/ata5

Doesn't tell me a whole lot.

well, you can use lspci to see what those devices are

you have 5 on one device, and two on another, and two on a third

@derobert ok. lspci with what argument?

lspci | grep 00:11.0 should work

3:26 PM

@derobert ok

and | grep 02:00.0, etc.

dmidecode is also a good way, if it works on your board

@derobert yes, i just run that. will it tell me everything about the hookup of the drives to the slots? looking now

@derobert thanks. That makes sense.

@FaheemMitha If you want to know where each drive is, ls -l /dev/disk/by-path is good, if it exists on your system

@derobert It does, thanks.

I'm confused, overclockersclub.com/reviews/asus_sabertooth990fx_r2_review/… only mentions eight SATA 6Gb/s ports and two eSATA ports.

3:37 PM

@FaheemMitha older specs?

Can one plug internal HDs into eSATA, and does eight SATA means space for 8 drives? I think one is probably used for the DVD drive already.

@Braiam Hmm?

Is this an out of date MB? I just bought it last year.

4:00 PM

@FaheemMitha That should be 10, maybe I missed one on your output?

@derobert It does sound like it should be 10.

You can connect an internal HDD to an eSATA port, I'm pretty sure, at least if you get the right cables

I've currently got 4 drives in there, I think.

@derobert Ok, thanks.

You should also be able to use port multipliers.

You have a lot of drives. I don't want to think of how many millions of dollars you must have paid for those in India...

@derobert I mistyped. I have 4.

Not counting the dvd writer.

@derobert This seems to afford you much amusement...

2 old, 2 rather newer.

@derobert per an earlier conversation, do you think it is worth buying Hitachi HDs?

4:03 PM

@FaheemMitha Well, it's pretty clear you'd be rich if you paid rest-of-world prices for your electronics...

They are almost certainly more expensive.

@derobert I'd have more money than I do, granted.

@FaheemMitha Personally, I got WD Reds.

@derobert Why?

Much cheaper. And still reasonably reliable.

The Seagate ones really seem crap these days, so they are off the table.

@derobert Ah. And the Red means...?

4:05 PM

Red is their line intended for small NAS applications

@derobert So?

That's what Red means

You concur Seagate == crap?

Ok, but why would one prefer them, though?

Yes. The WD Red was to replace... another failed Seagate.

They're the step up from WD Green drives, which I'm not sure I trust... and not much more expensive.

@derobert Ok. Do they have a premium line?

4:07 PM

@FaheemMitha Yes, they have WD Black and also WD Enterprise

@derobert And those are supposed to be better?

Is this what you meant? amazon.in/Drives-WD30EFRX-Western-Digital-IntelliPower/dp/… ?

WD Black is supposed to be faster (and probably is). WD Enterprise is supposed to be more reliable. No idea if WD Enterprise actually is.

@derobert Hmm.

Yep, WD30EFRX is what I just installed (and actually ordered another of)

@derobert That's a lot of space. Are you using RAID 10?

4:10 PM

@FaheemMitha I have parts in RAID5 and parts in RAID10. Most of it is RAID5. Once the newest drive comes in, plan to convert to RAID6.

@derobert I like 10. It's conservative.

Though the older drives are 1.5TB, so I don't yet have the rest of that 3TB drive available. Once I get another...

I see that WP is calling RAID 1 what I call RAID 10

That must be just you...

RAID 1 is plain mirroring. You copy the exact same data onto n drives

Maybe I'm using RAID 1, then. I just have an exact copy on two disks.

4:13 PM

Yes. That's RAID1.

RAID10 requires 4+ drives (for stacked implementation) or 3+ (for mdraid's magic raid10)

Yes, I guess I was confused. Ok, fine. I like RAID 1.

@derobert Ok, no, I'm definitely not doing anything so complicated.

RAID10 combines RAID 0 (striping, for performance) with RAID1 (mirroring, for redundancy)

So you wind up with data striped across m disks, and each stripe existing in n different copies. With 4 disks, that'd be striped across 2, and each stripe having 2 copies.

@derobert I wonder if it is reliable.

RAID10? yes. It's fairly good on reliability. It also features fast rebuilds, with minimal impact on system performance during rebuild.

@derobert Ok.

So, for 4 drives you get two drives worth of space. So, exactly the same as RAID 1.

4:19 PM

No, with RAID1, you get 1 drive worth of space. Always.

A 4-drive RAID1 has 4 copies of the data.

@derobert I was comparing two RAID 1s (4 disks) vs one RAID 10 (4 disks).

An 8-drive RAID1 has 8 copies of the data. A 20-drive RAID1 has 20 copies of the data (actually have one of those here at work, for /boot...)

Why would anyone use a 4 disk RAID 1? Sounds like overkill.

At some point, some other hardware catastrophe becomes more likely.

@FaheemMitha Well, then somehow you have to combine your two RAID1s. If you want performance, you'd combine them with RAID0. That'd be... a stacked RAID10.

I've actually never lost a hard drive to disk failure.

4:21 PM

RAID0 - Blocks striped, no mirror, no parity.
RAID1 - Blocks mirrored, no stripe, no parity.

I found this always useful.

@FaheemMitha I have a 3-disk RAID1 in use, because sometimes data is really #@!#( important

@derobert Sure, so the main reason to go from 2 RAID 1s of two disks each to one RAID 10 with 4 disks is performance?

@derobert Well, in that case, I'd be looking at backups. But I've never crunched the numbers.

@FaheemMitha The uptime is important, too.

In any case, all my important text stuff (except email) in in version control, so i just push it around.

@FaheemMitha Yes. mdadm RAID10 is at least as reliable as whatever way you were going to combine those two RAID1s, and will perform better.

@FaheemMitha Yeah, the 3-disk RAID1 is a database server. It needs to not lose the data, and remain available.

4:24 PM

@derobert Good point, but the probability of 2 drives failing and the third being untouched seem very small.

@derobert But more complicated to set up, presumably?

If two drives go at the same time. that means something external probably went wrong.

@FaheemMitha Probably less. It's at least one fewer array to set up.

@derobert oh

I mean, it's mdadm --create /dev/md/whatever -l 10 -n 4 /dev/sd[a-d]2, etc.

that's pretty easy

@derobert Oh, is that all? Hmm.

The only complicated part is picking your layout

4:26 PM

@derobert Layout of what?

@FaheemMitha how it places the stripes and copies on the drives

That's a performance optimization

@derobert Also specified in the command line?

With the right layout for your workload, you can get the performance of a 4-disk RAID0, but with safety

@derobert I see. Have you ever had a raid member go bad and have to rebuild?

@FaheemMitha I had Seagate drives...

I'd say I've dealt with that at least a hundred times?

4:29 PM

@derobert Blimey.

I have had a lot of disks in production for a long while, though. That's over the course of over ten years.

I think I've replaced 4 or 5 disks so far this year.

@derobert I didn't realise you ran such a big operation. No wonder you know all this stuff. How many computers do you manage?

@FaheemMitha It's not actually that large. Maybe 20-25? Not sure, would have to count.

@derobert Hmm, that seems like quite a lot. Is it like a cluster?

Nope, for the most part, they're all pretty unique. Some are in redundancy pairs, other than that, they're all different.

4:33 PM

One other problem with RAID 10 is that (I suppose) the disks should be all of the same size, otherwise it gets untidy. And of course, you want to be constantly increasing the size of the disks. Usually every time I buy new disks the size doubles.

@derobert I see. Universities and research orgs are big on clusters. So is eg Google, probably.

@FaheemMitha I'm not sure if mdadm supports growing raid10 yet.

@derobert I've no idea either. But isn't that an issue in practice?

Well, if you add more disks or bigger disks, you'd want it to

@derobert Not sure what you mean.

You'd want it to what?

mdadm --grow is how you change the array geometry to, e.g., take advantage of more disks, or larger disks (or switch RAID levels, or use fewer disks, etc.).

4:38 PM

Oh, I see, you'd want to be able to grow the raid array.

@derobert All with the risk of breaking your array, presumably.

@FaheemMitha Increasing size is relatively safe. Decreasing size is relatively safe, as long as you remembered to resize the filesystem (etc.) first.

Minor tech question. If building a RAID 1 with 2 disks, should I check the individual disks for bad blocks without proceeding? And if so, how?

No need. You'll find out about bad blocks when the initial sync happens.

@derobert Sure, but accidents can happen. I usually try to avoid messing with filesystems as much as possible.

@derobert Will the mdm raid thingy tell me about them?

Sure. Any thime you're playing with the low-level management tools, you need to be careful.

4:41 PM

Are you the sole admin for those 20/25 machines?

@FaheemMitha They'll either be repaired/reallocated by the disk, or the initial sync will fail. Messages will be spewed by the kernel.

@FaheemMitha Yep.

@derobert Hmm, that sounds pretty fancy. And if I want to find out about them in advance?

Personally, I run a conveyance self-test on a new disk with smartctl, and then start to use it.

@derobert Sounds like a lot of work.

@FaheemMitha You could have badblocks scan the disks... but there isn't much reason to.

The initial sync starts when you create the array with mdadm. If one of the disks fails, you just replace it.

You can of course reasonably wait for the initial sync to finish before putting critical data on the array. Or at least, your primary copy of said data.

4:44 PM

@derobert Ok. The SMART thingy sounds like a high-level reporting thing.

@derobert Ok

Do you do all the sw installation on those 20/25 machines too? And what OS?

Yep, I do pretty much everything on them. All of them run Debian.

@derobert testing? stable?

Your choice? Someone elses?

My choice. There used to be RedHat boxes here...

@derobert Good choice. :-) So they give you leeway on that?

They run various stable releases, many of them run oldstable or worse :-( [bloody hardware]

@FaheemMitha Yep. Place is tiny.

4:47 PM

@derobert Do you spec out the hw yourself? Who does the purchase?

Custom building might be a good idea if you don't have a big operation.

I have my machines custom built.

@FaheemMitha I spec it out, someone else actually does the purchase.

@derobert OK, that's good. Off the shelf stuff?

And by purchase, I mean types a credit card number into the system builder's web site.

They're mostly custom-built, but nothing too exotic. Except the tel cards.

(BTW: mdadm manpage seems to say RAID10 does not support grow yet)

gets into @derobert house and steals the shinnies

If you do custom build, i'd be interested to know who. Currently some people in India do stuff for me. Before that it was someone call hpcteam or something.

@Braiam shinnies?

@derobert Surprising.

4:50 PM

@FaheemMitha it's a place called Silicon Mechanics ... but they're in Washington (state), I believe. Not very close to India.

I don't think it was hpcteam exactly, but I'm blanking on the exact name.

@derobert I didn't think they were in India. :-)

So, they do good work?

@FaheemMitha Yeah, been quite happy with the work they've done.

teamhpc, not hpcteam

changelog.complete.org/archives/777-review-silicon-mechanics

@FaheemMitha stuff that shines, of value

JG is a DD. Not very active these days, though. AFAICT.

@Braiam Not familar with that word.

4:59 PM

@FaheemMitha Well, it'd probably help if he had the appropriate number of n's in it. But since he was trying to steal my stuff, I wasn't going to help him....

@Braiam shinnies? I think that's Candian slang for a variant of hockey or something.

@derobert yeah, I only wrote it by how it sounds

@FaheemMitha another option, if you're feeling adventurous, is to use btrfs's built-in RAID

@derobert I'm not feeling that adventurous.

I'm now using btrfs on a backup drive! Deduplication is nice for that application.

If I run out of space on it, I'll get another drive and add it to the btrfs

btrfs has nice things like data checksums.

I also have it as the rootfs on my TV PC.

So, I'm slowly gaining confidence that it won't explode and destroy all data...

5:19 PM

Isn't btrfs still considered beta, though? Have you filed and btrfs bug reports yet?

@FaheemMitha No, haven't hit any btrfs bugs yet.

@derobert oh

I've hit various features that are known not to be implemented (and documented so), but not any bugs

cool btrfs.wiki.kernel.org/index.php/Conversion_from_Ext3

5:36 PM

Apparently Chris Mason now also works for Facebook. As do the principal Mercurial people. Facebook does a surprising amount of tech work.

@derobert How can you hit unimplemented features? Aren't they non-features by definiition?

@FaheemMitha Well, you get "hey, it'd be nice to do X" and find out X is on the roadmap, but not implemented yet

5:50 PM

@derobert Does the filesystem itself say "hey..." etc?

@FaheemMitha No, that's me :-/

@derobert Ah

6:09 PM

can i run smartctl on a hard drive that is in use?

@FaheemMitha yes

Apparently yes. It doesn't report any specific errors, but I don't understand much the output either.

@derobert Would SMART detect bad blocks

@FaheemMitha Most hard disks will, at least if offline surface scan is enabled.

I have two WD caviar black and two seagate barracuda.

-c will tell you:

6:15 PM

The barracudas are not RAIDed and are getting old. Maybe I should RAID them. Though my important stuff is on the WDs, which are raided.

Offline data collection
capabilities:                    (0x7b) SMART execute Offline immediate.
                                        Auto Offline data collection on/off support.
                                        Suspend Offline collection upon new
                                        command.
                                        Offline surface scan supported.
                                        Self-test supported.
                                        Conveyance Self-test supported.

@derobert smartctl -c?

Offline data collection status:  (0x82) Offline data collection activity
                                        was completed without error.
                                        Auto Offline Data Collection: Enabled.

@FaheemMitha yeah

You can also tell it to do a long self test (-t long), that'll typically do a full surface scan

@derobert smartctl -t long /dev/sda?

yep

6:18 PM

Please wait 104 minutes for test to complete.

Should I do one disk at a time?

yep, it'll take a while. -c will tell you the % left. Once its done, -l xselftest or -l selftest (if xselftest doesn't work) will tell you the status.

You can do as many disks as you want at once. Performance will be impacted somewhat.

Note that if you're using the drives, it'll take longer than the 104 minutes.

@derobert ok. I guess I'll do all 4 then.

90% of test remaining. <- faster than expected.

It rounds...

Hmm, not changing.

90% means somewhere between 90% and 100%

6:21 PM

@derobert That's a lousy piece of rounding.

Yes, indeed. But that's how it works...

The only way its going to take less than 104 minutes is if it fails.

it would be more useful if it's 9/10 test remaining :/

A progress bar would be even better. And the devs should learn basic arithmetic.

Oh, BTW, I think I figured out my ghost pending sectors! I found the sectors that are bad. They're towards the end of the partition. I bet they're unused by mdraid.

The adjacent readable sectors are zero-filled

I'm not insane enough to poke them to fill with 0s, but I bet if I replace the disk with the hot spare, wipe the superblock, then zero those sectors, it'll clear them...

What drives are these?

6:27 PM

Seagate :-(

@derobert Ah. The Bad Boys.

Ok. Kicked off the replacement with the hot spare.

Must be lunch time, then.

If you do all this stuff, who is your backup?

@FaheemMitha One of the other people here can do some of it... Other than that, there aren't that many of us...

@derobert Ok. So you do emergency admin work on vacation, then?

6:34 PM

@FaheemMitha I guess if no one else could. Thankfully, doesn't really come up.

@derobert Oh. Happened to me once. And was only an admin briefly.

« first day (1404 days earlier) ← previous day next day → last day (3552 days later) »

Transcript for

Aug '1413

/dev/chat

General discussion for unix.stackexchange.com. If you have a q...

join this room
about this room

00:00

06:00

12:00

18:00

all times are UTC

site design / logo © 2024 Stack Exchange Inc; legal

mobile