=== Cale [n=cale@CPE004854805910-CM000e5cdd834a.cpe.net.cable.rogers.com] has joined #upstart
CaleHi, how do I get upstart to display what it is doing as it does it, during bootup and shutdown? I'd just like to be made aware of what's actually happening as it goes on.01:43
CaleThe little progress bar is cute, but it's not really all that informative.01:45
Cale(I'm using Edgy)01:47
_ionless /var/log/boot02:00
Caleyeah, how do I get that printed as the machine is actually booting?02:01
_ionProbably by adding 'console output' to /etc/event.d files that contain 'exec /etc/init.d/rc...'02:03
=== cortana [n=sam@62-31-146-25.cable.ubr12.azte.blueyonder.co.uk] has joined #upstart
=== cortana [n=sam@62-31-146-25.cable.ubr12.azte.blueyonder.co.uk] has joined #upstart
=== Cale [n=cale@CPE004854805910-CM000e5cdd834a.cpe.net.cable.rogers.com] has left #upstart ["Leaving"]
=== hunger [n=tobias@pd95b0676.dip0.t-ipconnect.de] has joined #upstart
=== nibil [n=nibil@adsl-static-3-105.uklinux.net] has joined #upstart
=== juergbi [n=juerg@80-219-26-249.dclient.hispeed.ch] has joined #upstart
=== johnnybuoy [n=johnny@unaffiliated/johnnybuoy] has joined #upstart
=== Md [i=md@freenode/staff/md] has joined #upstart
=== Md_ [i=md@freenode/staff/md] has joined #upstart
=== Md_ is now known as Md
=== jc-denton [n=nils@ztnw648.ztl.ch] has joined #upstart
jc-dentonhi all11:18
jc-dentoni have the strange problem that / is sometimes suddenly mounted read only11:19
jc-dentonand i cannout mount it rw with -o remount11:19
jc-dentonalso / is uncommented in /etc/fstab11:19
=== Md [i=md@freenode/staff/md] has joined #upstart
=== Keybuk [n=scott@quest.netsplit.com] has joined #upstart
johnnybuoywhy is commented?01:15
KeybukI don't understand?01:51
jc-denton# /dev/hda1 UUID=2232ef2d-5684-429b-8618-d31a881e55ab /               ext3    defaults,errors=remount-ro 0       101:51
jc-dentonit's commented out01:51
Keybukyou've edited the file and pressed "j" on that line, then saved it01:51
Keybuk(in vi)01:52
Keybukor edited it in some other editor, and joined the two lines, however you did that01:52
jc-dentoni did a fresh install of edgy01:52
jc-dentonand did nothing with /etc/fstab until i got trouble01:52
Keybuk"until you got into trouble" ?01:52
jc-dentonhowever i installed it when rc2 was out01:52
jc-dentonit was suddenly mounted read only01:52
Keybukso you edited /etc/fstab01:53
Keybukand broke it01:53
jc-dentonthe lin break is just there because i copied that thing01:53
jc-dentonlook i also checked it out on some other edgy installtion01:53
Keybukall I'm hearing here is "it was working fine until I broke it"01:53
Keybukif the edgy system booted the first time, then the file must have been correct on first installation01:54
Keybuk(and, btw, this has nothing to do with upstart)01:54
=== johnnybuoy [n=johnny@unaffiliated/johnnybuoy] has joined #upstart
=== Amaranth_ [n=travis@ubuntu/member/amaranth] has joined #upstart
jc-dentonso is /etc/fstab generated by something02:27
jc-dentonor can i simply edit it by hand02:27
jc-dentonand also what is this UUID?02:27
Keybukgenerated by the installer, but then edited by hand02:30
Keybukthe UUID is the UUID of the filesystem you want to mount02:30
Keybukyou can get it with "vol_id", e.g. "vol_id /dev/hda1"02:30
=== johnnybuoy [n=johnny@unaffiliated/johnnybuoy] has joined #upstart
=== johnnybuoy [n=johnny@unaffiliated/johnnybuoy] has joined #upstart
=== Amaranth [n=travis@ubuntu/member/amaranth] has joined #upstart
=== wasabi [n=wasabi@ubuntu/member/wasabi] has joined #upstart
jc-dentonKeybuk: thx so far..03:37
jc-dentonbut mount doesn't get these UUIDs03:48
jc-dentonso what does edgy use to mount it03:49
jc-dentonif i remove the # at the beginning03:58
jc-dentonand type mount -a03:58
Keybukthen you'll get an invalid /etc/fstab03:58
jc-dentonso it cannot be mounted by mount03:58
Keybuk# /dev/hda103:59
KeybukUUID=2232ef2d-5684-429b-8618-d31a881e55ab  /  ext3  defaults,errors=remount-ro  0  103:59
jc-dentonmaybe i need to ask different03:59
Keybuk# /dev/hda203:59
jc-dentonwhat reads these lines?03:59
KeybukUUID=ab2da10b-b58d-4f52-9291-2c39391e9996  none  swap  sw  0  003:59
Keybukis what that should look like03:59
Keybukjc-denton: mount.03:59
Keybukthe problem isn't that there's a "#" before "/dev/hda1"04:00
Keybukthe problem is that you've joined that line ("# /dev/hda1" with the line that should appear below it "UUID=..../...ext3..." etc.)04:00
jc-dentonnow i'm completely confused04:01
jc-dentonso the line for /dev/hda2 is correct04:01
wasabi_Am I the only one who doesn't really like UUIDs?04:01
=== Amaranth [n=travis@ubuntu/member/amaranth] has joined #upstart
jc-dentonbut the one for /dev/hda1 not04:01
Keybukjc-denton: the TWO LINES for /dev/hda2 are correct04:01
Keybukwasabi_: what would you prefer?04:01
Keybukjc-denton: the fact there's only one line for /dev/hda1 is what's wrong04:02
wasabi_Keybuk: Unfortunatly I have no answer to that. Probably static device names. ;004:02
Keybukjc-denton: tip: make it look like the /dev/hda2 TWO LINES04:02
Keybukwasabi_: device names aren't static04:02
jc-dentonthat's not how the syntax for /etc/fstab is defined04:02
Keybukjc-denton: yes it is04:02
wasabi_Why shouldn't they be though?04:02
wasabi_My problem with UUIDs is they ignore some odd corner cases.04:02
wasabi_Such as md1 vols.04:02
Keybukwasabi_: because part of the name is a number, which implies enumeration of devices, which implies a fixed device set04:03
_ionwasabi: If we want everything to just work with static device names, the device names must contain some kind of a UUID. :-)04:03
wasabi_but they don't just ignore it in a way as to make it not work, it continues to work, wrong.04:03
Keybukwasabi_: hmm?  md volumes contain an equivalent to a UUID or LABEL which can be used to mount them04:03
wasabi_Sure... I guess my main complaint is that the UUIDs come from the FS itself.04:03
wasabi_Which may actually have multiple paths to reach it.04:03
Keybukwasabi_: multiple paths is irrelevant, multiple views is the problem there04:03
wasabi_Sure? You've got the underlying md device... which is mountable!04:04
wasabi_But you should never mount it.04:04
jc-dentonUUID=asdfasdf is a device or what04:04
Keybukwasabi: what decides that one md is 0 and the other is 1 ?04:04
jc-dentonand what i would like to know04:04
jc-dentonis this documented somewhere04:04
wasabi_Keybuk: Hmm?04:04
Keybukjc-denton: yes04:04
Keybukjc-denton: yes04:04
Keybukwasabi_: well, if you're going to rely on something which is md0 and something which is md104:05
Keybukwasabi_: what decides which is 0 and which is 1? :)04:05
wasabi_Well, either way, it's a problem. When edgy switched to UUIDs, I couldn't boot. With my weird ass md + lvm + evms setup. :004:05
Keybukwasabi_: we mucked up md/lvm/evms because the developer who's supposed to care about that is a lazy idiot04:05
wasabi_Keybuk: I'd name the md device statically based on the ID of the *md device*04:05
wasabi_not of the FS on it.04:05
Keybukwasabi_: right, where's that ID stored?04:05
wasabi_In the MD metadata hopefully.04:06
jc-dentonKeybuk: i did man fstab04:06
wasabi_Which is unviewbale unless the md vol is composed.04:06
Keybukwasabi_: so that should be picked up, and placed in /dev/disk/by-id ?04:06
jc-dentonand there  is still the page from 200404:06
jc-dentonKeybuk: where?04:06
Keybukjc-denton: yes... fstab has supposed mounting by UUID or LABEL for a long time04:06
_ionwasabi: If you're using LVM, why use UUIDs instead of /dev/vgname-lvname? They already have a unique, static identifier.04:07
wasabi_Maybe I'm mis understanding something. Where *does* the uuid come from?04:07
Keybukwasabi_: make sure you're subscribed to the udev-* specs for MTV :p04:07
wasabi__ion: Mostly because it forced me too. ;)04:07
Keybukwasabi_: the filesystem/device/block device/container/whatever's appropriate04:07
wasabi_I have subscribed.04:07
=== mbiebl [n=michael@e180101102.adsl.alicedsl.de] has joined #upstart
wasabi_I guess that's maybe my core question.04:07
wasabi_In my case, where does the uuid come from.04:07
wasabi_The FS, the container, etc?04:07
Keybukwasabi_: right now, errrr04:07
Keybukls -l /dev/disk/by-id04:07
wasabi_At some point in the past it came from the FS, because I had two, that conflicted.04:08
Keybukls -l /dev/disk/by-uuid04:08
wasabi_And it mounted the wrong one.04:08
wasabi_Looking in /dev/disk/bu-uuid looks right now.04:08
wasabi_only one pointer to md1, no pointers to it's underlying devices.04:08
Keybukyou had two, identical, Universally UNIQUE Ids? :p04:08
wasabi_Yes. Two paths to the same file system.04:08
wasabi_It's was a MIRROR04:08
wasabi_MD metadata is at the end.04:08
wasabi_The underlying devices are still usable.04:09
Keybukso the MD itself should have an ID different to any of its exposed filesystems04:09
KeybukMD = abcdef, composed of two disks04:09
jc-dentonKeybuk: ok04:09
Keybukhda1 = 12345604:09
Keybukhda2 = 12345604:09
jc-dentonbut this is the first time i see that04:09
Keybukyou'd mount root=UUID=abcdef04:09
wasabi_In the past, it wasn't like that.04:09
jc-dentonis that linux only?04:09
Keybukjc-denton: yes04:09
wasabi_Keybuk: In the past, I had UUID=12345, which was valid for 3 different devices... /dev/sda3, /dev/sdb3, and /dev/md004:09
Keybukunless, of course, there's a way to tell by just looking at hda1 that it's part of an MD04:09
wasabi_And I guess the link was whatever the last one detected was. ;)04:10
Keybuk(without having yet composed the MD)04:10
wasabi_Or first one?04:10
wasabi_Also, I use XFS, which does have it's own UUID.04:10
wasabi_Maybe it decided to use that.04:10
Keybukthis is precisely what we need to discuss at MTV, and understand how it fits together04:10
wasabi_It looks fine now though. ;)04:10
Keybukand then make sure it works04:10
jc-dentonok i think i got it04:10
jc-dentonthanks so far for the infos04:10
jc-dentonbut it should be mentioned somewhere04:10
jc-dentondoes edgy has release notes or so04:10
Keybukjc-denton: it is mentioned, in many places04:10
wasabi_It made no UUIDs to my evms devices at all though04:11
jc-dentonbut i could not find it04:11
wasabi_But it did overwrite fstab to use uuids.04:11
Keybukwasabi_: yeah, we don't migrate MD, EVMS, LVM, etc. because we took the safety approach04:11
Keybukjc-denton: it's in the man page for fstab!04:11
wasabi_Heh. You did migrate them. ;)04:11
Keybukwasabi_: at first, yes04:11
Keybukthen we decided that it was a bad idea given we didn't understand them04:11
jc-dentonah ok04:12
wasabi_Well, to be clear, I don't mind UUIDs in the sense that I believe using a universally unique id is the right approach.04:12
wasabi_I just know I was burned a few times during the migration.04:12
wasabi_Shit. I'm a MS guy. I'm used to GUIDs all over the place.04:12
jc-dentonThis  will  make  the  system  more robust: adding or removing a SCSI disk changes the disk device name but not the filesystem volume label.04:12
Keybukthat's cause we don't understand LVM, EVMS, MD, etc.04:12
jc-dentonah ok04:13
wasabi_Can't wait to talk about em. :004:13
Keybukwe have specs to understand them, write it down, and then fix it :p04:13
jc-dentonso that ensures that another disk is not mounted on a specified mount point04:13
wasabi_I do some crazy shit with md/lvm/evms. I'd love to see that crazy shit supportable.04:13
Keybukjc-denton: and, more pointedly, it makes your system robust against installing (for example) another controller card04:13
Keybukit should be04:13
Keybukfrom what I understand, we basically just need to make sure04:13
Keybuka) uevents are generated04:13
Keybukb) which uevents mean we need to run certain tools04:14
Keybukc) how to obtain a unique identifier for the fs/container/device/wibble04:14
_ionjc-denton: Also, in (maybe not so) distant future, the kernel may switch to the libata PATA implementation. Then hd* will change to sd*.04:14
Keybuk_ion: RSN04:15
KeybukI believe it was merged into 2.6.1904:15
_ionOk, cool.04:15
_ionThe edgy kernel already uses libata for PATA disks connected to my SATA+PATA controller.04:16
wasabi_I guess the trick is figuring out which information from which devices to seed a uuid with.04:16
wasabi_For a md device, it should be the md metadata of the composed vol.04:16
wasabi_evms, it should be the evms metadata of the composed volume.04:17
wasabi_etc etc04:17
wasabi_Which is interesting because you can't actually obtain the MD metadata by looking at the md device itself.04:19
=== mbiebl [n=michael@e180101102.adsl.alicedsl.de] has joined #upstart
=== Amaranth [n=travis@ubuntu/member/amaranth] has joined #upstart
=== Jameson1234 [n=napkin@ip503c00d9.speed.planet.nl] has joined #upstart
=== Jameson1234 [n=napkin@ip503c00d9.speed.planet.nl] has joined #upstart
hungerKeybuk: Things get interesting with encrypted devices I think.05:52
hungerKeybuk: block device comes up -> need to check whether it contains files for loopback -> those turn into new block devs -> get mapped to yet another blockdev -> that needs to get mounted.05:52
Keybukthis is easy :)05:58
Keybukit's a nice logical sequence of events05:59
wasabi_Keybuk: I have some questions about UMV. Curious how this plays out. These specs are driving out schedules and stuff right?06:14
wasabi_What do we all just end up in a room at a certain time sitting around a table?06:15
wasabi_And then the guy who registered the specs speaks up or something?06:16
=== phsdv [n=kvirc@] has joined #upstart
=== johnnybuoy [n=johnny@unaffiliated/johnnybuoy] has joined #upstart
=== Amaranth [n=travis@ubuntu/member/amaranth] has joined #upstart
=== Md_ [i=md@freenode/staff/md] has joined #upstart
=== Md_ is now known as Md
=== phsdv [n=kvirc@] has left #upstart ["So]
=== mbiebl [n=michael@e180101102.adsl.alicedsl.de] has joined #upstart

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!