| FourDollars | hi, does anyone know which code fragment generates http://www.flickr.com/photos/fourdollars/6891707522/in/photostream/ ? | 05:57 |
|---|---|---|
| FourDollars | OK. I find it. partman-partitioning/no_bootable_gpt_efi | 06:22 |
| FourDollars | No, I didn't find it. Only the message, but not source code. | 06:25 |
| FourDollars | It seems relative to http://bazaar.launchpad.net/~ubuntu-core-dev/partman-partitioning/ubuntu/revision/913 . | 06:28 |
| infinity | Which source code are you referring to? | 06:28 |
| infinity | You mean "what triggers the error", or "what's displaying it"? | 06:28 |
| infinity | For the former, it's the revision you just pointed out. | 06:28 |
| FourDollars | what triggers the error | 06:29 |
| infinity | Exactly as it implies. The system uses only GPT-partitioned disks, and you have no EFI boot partition. | 06:30 |
| infinity | (And the test for that is in the revision you just linked) | 06:30 |
| FourDollars | yes, but the system does have EFI boot partition. | 06:31 |
| infinity | Is it marked as such? | 06:31 |
| FourDollars | However, EFI boot partition is created by myself. | 06:32 |
| FourDollars | So I am wondering how it detect 'EFI boot partition'. | 06:32 |
| infinity | It detects it based on it being set to that type. | 06:32 |
| infinity | 'This partition should be marked for use as an "EFI boot partition"' | 06:33 |
| FourDollars | Can we overwrite it by d-i preceed? | 06:33 |
| infinity | I'm sure it's preseedable, but you'll want cjwatson to get more into the details of that. | 06:34 |
| FourDollars | Thanks. | 06:35 |
| infinity | cjwatson: Also, is there a way to mark strings in debconf templates as unsplittable? In the above screenshot, "/boot" gets split to "/\nboot", which is rather unfortunate. | 06:36 |
| GrueMaster | More fun with preseeding partitioning. I feel your pain. | 06:42 |
| * GrueMaster spent all weekend trying to create a recipe for mixed pv/lvm partitioning with little success. Failing on blank partitions. | 06:43 | |
| FourDollars | lol | 06:43 |
| antarus | I just avoid anything complicated ;P | 06:44 |
| GrueMaster | I usually dive it and beat my head against the bottom until I either succeed or suffer brain hemorrhages. Speaking of, I'm heading to bed. | 06:45 |
| GrueMaster | My preseed if anyone wants to tell me how far off I am: http://paste.ubuntu.com/911134/ | 06:46 |
| GrueMaster | cjwatson: ^^^ | 06:47 |
| twb | In old d-i builds, installing the network-console udeb and completing the install over SSH would automatically tick the "OpenSSH server" tasksel task. | 07:33 |
| twb | This appears to not be the case in precise -- is this deliberate? | 07:34 |
| twb | Tested with http://archive.ubuntu.com/ubuntu/dists/precise/main/installer-amd64/current/images/netboot/ubuntu-installer/amd64/linux as at 2012-03-31 06:51:39 +1100, not sure how old my udeb mirror is offhand. | 07:35 |
| twb | Also, is the bug where partman defaults to GPT for ≥2TB disks (instead of 2TiB disks) fixed for precise? | 07:41 |
| twb | Looks like a bunch of the preseed options wrt. DNS domain name, locale, keyboard have changed, too :-/ | 08:09 |
| twb | Oh and NTP is picking Ubuntu's instead of the one supplied by the DHCP server | 08:10 |
| cjwatson | FourDollars: please file a bug with installer syslog and partman attached, and I'll see if I can work something out. Preseeding isn't the right answer here. | 08:39 |
| cjwatson | infinity: unsplittable> no, I'm afraid not | 08:39 |
| cjwatson | GrueMaster: how do you mean, "failing on blank partitions"? | 08:40 |
| cjwatson | twb: it was only ever intended to install the openssh-server package, not specifically check the task; the latter sounds like an emergent effect. have you checked whether it still installs the package? | 08:40 |
| twb | I didn't think to try that. I managed to set the root password wrong in the install, so I went back to doing the production lucid install that this box is gonna have | 08:41 |
| twb | What I actually care about is openssh-server being installed – if you're reasonably confident that's what happens, that's good enough for me. | 08:42 |
| cjwatson | twb: Yes, I fixed the TB/TiB confusion in partman-partitioning 83 and that's in precise. | 08:42 |
| twb | Awesome, thanks (re TB/TiB) | 08:42 |
| cjwatson | openssh-server *should* still be installed, looking at the code, but I've not tried that install path myself at all recently. | 08:43 |
| twb | btw did you know that if you mke2fs /dev/sda, partman will let you install to it without a partition table? This is a handy feature/bug that I use all the time now :-) | 08:43 |
| twb | Especially when sda is vda | 08:43 |
| cjwatson | Yep, I think so | 08:43 |
| twb | I don't remember seeing it documented anywhere | 08:44 |
| FourDollars | cjwatson: I am wondering what you check between 60 and 70 of check.d/biosgrub of http://bazaar.launchpad.net/~ubuntu-core-dev/partman-partitioning/ubuntu/revision/913 . | 08:44 |
| cjwatson | It's a consequence of the handling of what parted calls "loop" devices, which aren't loopback mounts but actually more like the way that you can have a partition on /dev/mapper/foo | 08:44 |
| twb | Ah | 08:45 |
| twb | Also I discovered sponge yesterday, and it saved me a few lines of sh - thanks :-) | 08:45 |
| cjwatson | FourDollars: "method" is what you've asked partman to use the partition as, or what it's guessed that it needs to be used as for itsel | 08:45 |
| FourDollars | cjwatson: Is partition id '0xef' ? | 08:45 |
| cjwatson | f | 08:45 |
| cjwatson | No | 08:45 |
| cjwatson | The id there will be more like START-END in bytes | 08:46 |
| FourDollars | cjwatson: If the disk already has EFI partition before partman is executed, what does partman do? | 08:47 |
| cjwatson | It's supposed to automatically assign it method 'efi' | 08:49 |
| cjwatson | Please file that bug for me | 08:49 |
| FourDollars | Roger that. | 08:49 |
| cjwatson | (on partman-partitioning, I guess) | 08:50 |
| FourDollars | I think so too. | 08:51 |
| FourDollars | Thanks a lot. | 08:51 |
| cjwatson | may actually be a bug in partman-efi or somewhere else, but I'll need to analyse the logs to tell | 08:52 |
| FourDollars | I will generate the log files once I get access the hardware with EFI BIOS. | 08:57 |
| * FourDollars can not get the hardware today. | 08:57 | |
| GrueMaster | cjwatson: Morning. Did you see my pastebin? The failure is that it prompts me for a filesystem for the blank LVM partition. If I give it partition info in the preseed then it prompts for a mount point. I just want to have a LVM partition that is blank, unformatted, and not mounted. method{ keep } appears to be ignored in partman-lvm. | 14:38 |
| cjwatson | I fear that may not be possible right now. | 14:39 |
| cjwatson | Though that does seem slightly odd and I'd welcome a bug on partman-auto-lvm, preferably with that preseed and a DEBCONF_DEBUG=developer syslog attached. | 14:40 |
| cjwatson | And I suppose a partman log too. | 14:40 |
| GrueMaster | Lovely. The documentation (for what little it is worth) indicates otherwise. I've spent all weekend working on this. Now I know why most of my google searches show examples where people have written their own replacements. | 14:40 |
| cjwatson | It probably wouldn't be that hard to fix. | 14:41 |
| cjwatson | (Well, assuming prior knowledge.) | 14:41 |
| GrueMaster | I'll see what I can do. The system is currently stuck in a loop. | 14:41 |
| GrueMaster | And I only have serial console access. | 14:41 |
| cjwatson | Well, preseed attached to a bug better than nothing. | 14:41 |
| cjwatson | Prompting for a mountpoint on the automatic partitioning path is bizarre; I'd like to see that for myself to see exactly what path it's following | 14:43 |
| GrueMaster | The preseed above isn't anything special (other than local mirror settings). Should be fairly easy to reproduce on x86 in a kvm. | 14:45 |
| CIA-32 | ubiquity: stgraber * r5349 ubiquity/ (bin/ubiquity-dm debian/changelog): Set default wallpaper for ubuntu studio in ubiquity-dm | 14:54 |
| GrueMaster | cjwatson: Bug 971608 filed. I don't have access to debian's bug tracking system to cross file. | 15:08 |
| ubot2 | Launchpad bug 971608 in partman-auto-lvm "Unable to create blank lvm volume in preseed install" [Undecided,New] https://launchpad.net/bugs/971608 | 15:08 |
| cjwatson | GrueMaster: you shouldn't anyway, unless you've reproduced it in Debian. | 15:09 |
| cjwatson | (and the entire world has access to Debian's BTS, it has no authentication :-P ) | 15:10 |
| aperson_ | Hello, I'm trying to find out if the documentation at: http://d-i.alioth.debian.org/manual/en.i386/apbs02.html#preseed-auto applies to Ubuntu Oneiric. I'm using "auto url=http://ubuntu-builder/ubuntu/preseed.d/preseed.cfg" on the kernel command line but at install loading the preseed file fails. The syslog has: Mar 29 23:33:03 main-menu[354]: (process:1340): wget: bad address 'ubuntu-builder' The DHCP server did give the | 17:21 |
| aperson_ | leases file has the correct DOMAIN in it. Also if I manually put the DOMAIN in the resolv.conf file the build will succeed. | 17:22 |
| GrueMaster | aperson_: I assume ubuntu-builder is a local server? Does your dns resolve that normally? | 17:58 |
| aperson_ | ubuntu-builder is a local server and it resolves correctly if it's a FQDN ex: auto url=http://ubuntu-builder.example.com/ubuntu/preseed.d/preseed.cfg but not if it's just the hostname. The docs state that it should pick up the domain name from DHCP | 18:01 |
| aperson_ | The syslog shows Mar 29 22:02:03 netcfg[1243]: DEBUG: DHCP domain name is 'sub.example.com' so to me that indicates that it knows of the dowmain | 18:04 |
| aperson_ | Also if I look at the leases file it shows the domain. | 18:05 |
| aperson_ | What seems to be missing is that netcfg or main-menu isn't add the domain token to the resolv.conf file. | 18:06 |
| GrueMaster | aperson_: Sorry, got distracted. Something is not right in your dhcp/dns setup. I have no problems here. I do netboot & preseed installations daily in my own private network. My preseed/tftp/pxe/pool server is mirror.gruenet. Everything is shortened to http://mirror/... (http://mirror/preseeds/, etc). And it just works. | 18:50 |
| aperson_ | Here's the complete PXE target being used: | 18:54 |
| aperson_ | label desktop-jim | 18:54 |
| aperson_ | kernel ubuntu/linux-64 | 18:55 |
| aperson_ | append initrd=ubuntu/initrd-64.gz BOOT_DEBUG=2 -- locale=en_US debian-installer/locale=en_US kbd-chooser/method=us interface=eth0 netcfg/dhcp_timeout=60 priority=critical auto=true auto url=http://ubuntu-builder/ubuntu/preseed.d/preseed.cfg | 18:55 |
| aperson_ | I'm having to do this remotely and I'm waiting for them to re setup the test system. Once that's done I can get the leases file from the building host. | 18:56 |
| aperson_ | That way you can see what DHCP is providing | 18:57 |
| aperson_ | I do have the output of a syslog from a previous attempt. | 18:57 |
| aperson_ | I've been looking at the netcft source code and I don't see how it could work. Unless you pass netcfg/get_domain via the kernel command line. | 18:59 |
| GrueMaster | My pxe config for amd64 is this: append vga=788 initrd=ubuntu-installer/amd64/initrd.gz url=http://mirror/preseed/precise-amd64.cfg -- debconf/priority=critical locale=en_US console-setup/ask_detect=false console-setup/layoutcode=us live-installer/enable=false | 19:00 |
| GrueMaster | As I said, I think the fault lies in your dhcp configuration. | 19:00 |
| GrueMaster | My dhcp server has: | 19:01 |
| GrueMaster | option domain-name "gruenet"; | 19:01 |
| GrueMaster | option domain-name-servers 192.168.0.1; | 19:01 |
| aperson_ | I'll change it if I can figure out what to change. The syslog shows that it gets: Mar 29 22:02:03 netcfg[1243]: DEBUG: DHCP domain name is 'asubdomain.example.com' | 19:02 |
| aperson_ | shouldn't your option domain-name be "gruenet.com" or org? | 19:03 |
| GrueMaster | No, because it is neither a .com or .org. It is internal only. | 19:04 |
| aperson_ | ahh | 19:04 |
| GrueMaster | It isn't the internet, it is the gruenet. :P | 19:04 |
| aperson_ | :D | 19:04 |
| aperson_ | as in plonk? | 19:05 |
| GrueMaster | Something like that. | 19:05 |
| aperson_ | looking at the source code for netcfg I don't see how it would get the correct domain name unless it's on the kernel command line. it reads it from /tmp/domain_name, but I don't see it filled in. | 19:07 |
| GrueMaster | In your install? | 19:07 |
| aperson_ | from dhcp.c line 445 to 455 | 19:08 |
| aperson_ | the dhcp.c from the netcfg tar ball | 19:09 |
| aperson_ | 444 /* 445 * Default to the domain name returned via DHCP, if any 446 */ 447 if ((d = fopen(DOMAIN_FILE, "r")) != NULL) { 448 char domain[_UTSNAME_LENGTH + 1] = { 0 }; 449 fgets(domain, _UTSNAME_LENGTH, d); 450 fclose(d); 451 unlink(DOMAIN_FILE); | 19:10 |
| GrueMaster | Well, only thing I can suggest is trying to run with the fqdn address for the preseed server. | 19:10 |
| aperson_ | ugly sorry | 19:10 |
| aperson_ | Yeah, I've been trying to avoid that as the combination of PXE targets and preseed files will explode to a few hundred quickly. I have to have dozens of mirrors for remote locations | 19:11 |
| aperson_ | I can get a anycast address, but that causes other issues. | 19:12 |
| aperson_ | Thanks for the help | 19:12 |
| === stgraber_ is now known as stgrber | ||
| === stgrber is now known as stgraber | ||
| === infinity1 is now known as infinity | ||
| antarus | hrm anyone know what the actual fields in partman recipies refer to? | 23:31 |
| antarus | ahh min max priority? | 23:32 |
| antarus | weird | 23:32 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!