| mpt | Okay, back to LVM | 11:48 |
|---|---|---|
| mpt | cjwatson, xnox: Did you have any further thoughts on the LVM sketches I posted yesterday? | 11:48 |
| xnox | mpt: =) | 11:49 |
| xnox | nothing from me. | 11:50 |
| cjwatson | I didn't, though mostly got eaten by figuring out how to move Kubuntu to universe | 11:51 |
| mpt | If we go for the first option, then the funnels between the two lists would slide around as you scrolled either list. | 11:51 |
| mpt | (Even down to nothingness, if an LVG scrolled out of view altogether.) | 11:52 |
| xnox | mpt: http://meldmerge.org/images/meld-mary.png ? | 11:53 |
| mpt | xnox, yes, like that. I was looking for suitable screenshots of kdiff3 yesterday. :-) | 11:53 |
| xnox | meld is gtk app =) | 11:53 |
| xnox | mpt: http://docs.fedoraproject.org/en-US/Fedora/15/html/Installation_Guide/images/diskpartitioning/ddmain.png | 11:55 |
| xnox | fedora's implementation in anaconda installer | 11:55 |
| xnox | http://docs.fedoraproject.org/en-US/Fedora/15/html/Installation_Guide/s1-diskpartitioning-x86.html | 11:55 |
| mpt | xnox, that looks like my second sketch -- with "Hard Drives" as their equivalent of my "Disks not in any LVM volume group". | 11:56 |
| xnox | yeap. | 11:56 |
| xnox | mpt: I like funnels, but I don't know how slick the implementation will look | 11:57 |
| mpt | "The top pane contains a graphical representation of the hard drive, logical volume, or RAID device selected in the lower pane." -- Well, that's obviously upside down then, isn't it | 11:57 |
| xnox | and if funnels will be friendly, if the are not slick. | 11:57 |
| mpt | xnox, is it actually possible to have an LVG that contains only one disk? | 11:58 |
| xnox | notice the naming as well VolGroup & lv_MOUNTPOINTNAME | 11:58 |
| xnox | yes. | 11:58 |
| xnox | I do | 11:58 |
| xnox | let me show you in a sec | 11:58 |
| xnox | mpt: LVG with one disk, because I can remove/add/resize/snapshot partitions without dataloss | 11:59 |
| xnox | as a regular disk repartitioning requires defragmentation and is constrained by physical layout. | 12:00 |
| mpt | Huh. | 12:00 |
| mpt | Sorry if this is a silly question, but: If LVM makes repartitioning easier, why don't we do it by default? | 12:01 |
| xnox | =)))) | 12:02 |
| xnox | because Windows/MacOSX cannot read any partitions inside LVM | 12:02 |
| cjwatson | We looked at it ages ago. Another problem was that the desktop-side management tools weren't good enough last time we looked. | 12:02 |
| xnox | and Windows/MacOSX cannot be run on top of LVM | 12:03 |
| xnox | they have their own solutions. | 12:03 |
| cjwatson | Although in principle it ought to be way better. | 12:03 |
| mpt | ok | 12:03 |
| xnox | Windows Snapshot Service thing and not sure on MacOS X side. | 12:03 |
| xnox | https://picasaweb.google.com/105922848292507689403/June152012?authuser=0&authkey=Gv1sRgCOrolvqC0qOKPg&feat=directlink | 12:03 |
| mpt | Is Ubiquity still embedding gparted to the extent that the LVM changes you're making here will be picked up by gparted? | 12:03 |
| xnox | mpt ^^^ | 12:03 |
| xnox | my layout of the disks | 12:03 |
| * xnox nevermind the ugly colouring of the labels | 12:03 | |
| xnox | mpt: gparted is not in ubiquity, but LVM will show / display work with gparted fine | 12:04 |
| * xnox or is gparted used by partman cjwatson ? | 12:04 | |
| mpt | The current advanced partitioning UI is based on gparted | 12:05 |
| mpt | I don't know whether it's a copy or an embed | 12:05 |
| cjwatson | xnox: No. | 12:06 |
| cjwatson | mpt: It's not really. | 12:06 |
| cjwatson | It used to be pre-feisty. | 12:06 |
| cjwatson | That was awful and I rewrote it. | 12:06 |
| cjwatson | So no, it's not based on gparted in any meaningful sense any more, except maybe some tenuous inspiration or something. | 12:07 |
| cjwatson | Even then I don't think that's so very much. | 12:07 |
| * xnox gparted doesn't seem to do anything sensible with LVM2, the disks application can though. | 12:08 | |
| mpt | cjwatson, ok, I apologize for not noticing a huge amount of work on your part. :-) | 12:08 |
| * xnox is off to lunch with my mate. be back soon | 12:08 | |
| mpt | A benefit of that Anaconda design is that it doesn't use a separate dialog for LVM, it's all just in the disk/partition table | 12:09 |
| mpt | The tradeoff is that their "Create Storage" dialog is hideously complex | 12:11 |
| mpt | cjwatson, so is a "physical volume" either a whole disk or a partition that isn't inside an LVG already? | 12:18 |
| cjwatson | (I wouldn't use the abbreviation "LVG". LVM's objects are PV, VG, LV) | 12:20 |
| mpt | ok | 12:21 |
| cjwatson | But yes - or, strictly, it's such a thing that has been prepared for use by LVM | 12:21 |
| mpt | but we don't need to expose that preparation at all in the UI, right? just do it when we need to | 12:22 |
| cjwatson | Right. | 12:22 |
| cjwatson | At the moment we have a step where you tell the partitioner that a partition's "use as" method is "for use by LVM" or some such | 12:22 |
| cjwatson | But that doesn't really need to be exposed | 12:23 |
| mpt | ok | 12:28 |
| mpt | oh poop | 12:36 |
| mpt | If ordinary everyday partitions can be part of VGs, then we can't reorder the physical volumes to match their use in VGs, without breaking the grouping of partitions under their parent disks | 12:37 |
| mpt | So the funnel will be more like a braid | 12:37 |
| cjwatson | Indeed. | 12:38 |
| mpt | which, in turn, means that some of them will overlap each other | 12:42 |
| mpt | (or could) | 12:42 |
| cjwatson | Quite possible. | 12:42 |
| mpt | oh boy, this is going to be fun to implement | 12:44 |
| mpt | We're going to need clever icons for PV, VG, and LV too... | 12:45 |
| * mpt realizes that a logical volume group is a logical {volume group}, not a {logical volume} group | 12:52 | |
| mpt | I guess "logical group of physical volumes" was too wordy | 12:53 |
| cjwatson | Just don't use the term "logical volume group" and you'll be fine. | 12:59 |
| cjwatson | Since LVM doesn't. | 13:00 |
| mpt | ok | 13:02 |
| mpt | On the subject of terminology: What would be a good button label for dismantling a volume group altogether? | 13:05 |
| mpt | I had "Split" in the sketches, but that's not very good | 13:05 |
| mpt | "Dismantle"? | 13:05 |
| cjwatson | I would recommend preferring the verbs used by LVM itself, since they generally aren't too terrible and that way we have a better chance of not confusing people who already know LVM. | 13:06 |
| cjwatson | In this case that would be "Remove". | 13:06 |
| cjwatson | (If necessary, "Remove volume group" to distinguish from "Remove physical volume" or whatever.) | 13:07 |
| mpt | That LVM already uses "Remove" for removing individual volumes from a group is precisely why I don't want to use it again for something else :-) | 13:09 |
| cjwatson | "Remove logical volume", "Remove volume group" | 13:09 |
| cjwatson | Seems clear enough | 13:09 |
| mpt | And also because removing a volume from a group makes sense -- but what do you remove a volume group from? | 13:11 |
| cjwatson | The system | 13:12 |
| cjwatson | Much like, say, you remove a partition, or a file | 13:13 |
| mpt | You remove a partition from a volume, a file from a folder | 13:13 |
| cjwatson | s/volume/disk/ | 13:13 |
| mpt | Disk, or logical volume :-) | 13:13 |
| cjwatson | No | 13:13 |
| mpt | ehh | 13:13 |
| mpt | Disk, or volume group | 13:13 |
| cjwatson | You don't partition logical volumes unless you're insane | 13:14 |
| cjwatson | And you don't partition volume groups; you create logical volumes in them | 13:14 |
| mpt | Dammit I understood this until five minutes ago :-) | 13:14 |
| mpt | I thought that volume groups were equivalent to disks, in that you can partition them | 13:15 |
| cjwatson | Logical volumes are slightly analogous to partitions, but I wouldn't recommend using the same terms for them because LVs are much more fluid (less "partitioned"). | 13:15 |
| mpt | (That's why I was wondering earlier why we don't do LVM by default for every disk) | 13:15 |
| cjwatson | For example, nothing stops you extending a logical volume even if its current extents are enclosed by the extents for two other LVs on either side. | 13:16 |
| cjwatson | Anyway, I have to rush out for a bit | 13:16 |
| mpt | ok | 13:16 |
| mpt | I want to avoid having a UI for ... volumizing? volume groups that is a near-duplicate of the one for partitioning disks | 13:17 |
| xnox | mpt: disk and partitioning is like a bread roll, which you have to cut into pieces. LVM is more like a buckets of water: take physical volumes -> and make a bucket of water (Volume Group), and then you can pour as much as you like into different smaller buckets (logical volumes) | 13:19 |
| cjwatson | They can be in the same UI, but we shouldn't pretend they're exactly the same thing. | 13:19 |
| cjwatson | Because the constraints are quite different. | 13:19 |
| xnox | similarly you can pour water back into volume group, to resize/move/delete/increase size | 13:19 |
| cjwatson | For example, when creating a logical volume, there is no reason why you would put it in any particular position in the volume group, whereas that can be quite important for disks in some cases. | 13:20 |
| cjwatson | (In fact, you don't get to know what position a logical volume is in.) | 13:20 |
| mpt | ok, so there are variations that need speccing for how VGs behave in the main partition table | 13:22 |
| === ogra_ is now known as ogra | ||
| === ogra is now known as ogra_ | ||
| kentb | is it unusual for the very first page / plugin in oem-config to get loaded twice, for example: | 14:19 |
| kentb | Jun 15 12:04:36 dell-desktop ubiquity[2077]: switched to page language | 14:19 |
| kentb | Jun 15 12:04:36 dell-desktop ubiquity[2077]: switched to page language | 14:19 |
| xnox | kentb: i think it's ok. First it loads and checks if the langauge was preselected during boot, second time to check the preseed. | 14:41 |
| xnox | but better cjwatson to answer. I'm don't know oem-config that well. | 14:42 |
| kentb | xnox: Ok. thanks | 14:42 |
| * xnox grammar fail | 14:42 | |
| xnox | kentb: your welcome | 14:42 |
| cjwatson | It sounds a bit curious and at the very least could use some ambiguous debugging. | 14:49 |
| cjwatson | some LESS ambiguous debugging | 14:49 |
| xnox | cjwatson: s/ambiguous/ambitious/ would have worked as well =) | 14:55 |
| mpt | cjwatson, xnox, first draft done, details and questions mailed to ubuntu-installer@ <https://docs.google.com/a/canonical.com/document/d/1bZ4yQIVgGaUGSYu3qiUHnQt3ieBZoqunP_DcleHCr3I/edit#heading=h.v8wi3omt1z0> | 15:02 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
