hardwire | I uh.. was sorta ushered over here from debian-boot ;) | 07:48 |
---|---|---|
hardwire | http://techbits.bogomip.com/2008/09/how-i-installed-ubuntu-amd64-server-on.html | 07:48 |
hardwire | just wrote that.. and sorta maybe just discovered that the hostname and domain is being overridden by the reverse dns lookup for the static IP during the automatic install | 07:49 |
hardwire | anyways to .. stop that? | 07:49 |
hardwire | cause it's bad m'kay? | 07:49 |
cjwatson | hardwire: unfortunately, this is a long-standing bug, namely bug 218965 (see the linked Debian bug there for more detail). The best I can offer as a workaround is to sed the hostname and domain you want into place in the relevant files in /target from a preseed/late_command hook. | 09:11 |
=== superm1|away is now known as superm1 | ||
=== superm1 is now known as superm1|away | ||
cr3 | is there a way to hook into the alternate installer to run some command when an error occurs? | 15:29 |
=== superm1|away is now known as superm1 | ||
superm1 | well CIA decided to not submit a few of my last commits again. i fixed an issue that I just saw with the build process introduced recently preventing the GTK frontend from building | 15:48 |
superm1 | i'm not sure if the last ubiquity upload (the .11) is affected | 15:48 |
evand | cia-submit? | 15:50 |
evand | but thanks | 15:50 |
superm1 | ah no it wasn't affected because it was FTBFS :) http://launchpadlibrarian.net/17247922/buildlog_ubuntu-intrepid-i386.ubiquity_1.9.11_FAILEDTOBUILD.txt.gz | 15:50 |
CIA-44 | ubiquity: superm1 * r2793 ubiquity/ (5 files in 4 dirs): update mythbuntu frontend video driver installation to use mythbuntu-common and abstract the versioning | 15:50 |
evand | ah, hooray | 15:50 |
superm1 | there should be 4 others or so though | 15:50 |
superm1 | hum | 15:50 |
superm1 | the others were more relevant | 15:50 |
CIA-44 | ubiquity: superm1 * r2793 ubiquity/ (5 files in 4 dirs): update mythbuntu frontend video driver installation to use mythbuntu-common and abstract the versioning | 15:50 |
superm1 | oh well | 15:50 |
superm1 | there looks to possibly be another issue going on with partman-base from something that cjwatson committed to the tree a long time back but that was never released | 15:52 |
superm1 | well when luke rev'd partman-base, that commit from way back when got added | 15:53 |
evand | hrm | 15:59 |
evand | it appears to have built fine. What's the issue, exactly? Is the install failing? | 16:01 |
superm1 | yeah | 16:02 |
evand | alternate or desktop? | 16:02 |
superm1 | desktop, but only using builds that are 1.9.11 or later | 16:02 |
superm1 | since they pull that in | 16:02 |
superm1 | it turns out that preseed_command isn't found | 16:02 |
superm1 | which i'm not sure where it was supposed to be provided from anyhow? I quick grep didn't indicate where it's supposed to come from | 16:03 |
evand | preseed | 16:04 |
* evand digs | 16:05 | |
superm1 | well the variable partman/early_command should come from preseed, but the executable/shell script what not called preseed_command doesn't seem to be sitting around in the usual places | 16:05 |
cjwatson | hm, really? | 16:07 |
cjwatson | oh, of course, it isn't in ubiquity, bleh | 16:07 |
cjwatson | evand: might need to suck that into ubiquity's pile of imported source packages. We shouldn't need the rest of it but preseed_command looks like it should be usable standalone | 16:07 |
evand | ok, will do | 16:09 |
=== superm1 is now known as superm1|away | ||
hardwire | anybody else noticed that hardy current d-i installers seem to ignore domain and hostname settings in /proc/cmdline as well as the preseed file if there is RDNS available? | 17:38 |
hardwire | using static IP? | 17:38 |
hardwire | here's my preseed: http://techbits.bogomip.com/2008/09/how-i-installed-ubuntu-amd64-server-on.html | 17:39 |
hardwire | in actuality, I'm not using an internal IP.. those are example settings. | 17:39 |
CIA-44 | ubiquity: evand * r2794 ubiquity/.bzrignore: Add mythbuntu build directory to bzrignore | 17:39 |
cjwatson | hardwire: I answered you this morning | 17:43 |
cjwatson | 09:11 <cjwatson> hardwire: unfortunately, this is a long-standing bug, namely bug 218965 (see the linked Debian bug there for more detail). The best I can offer as a workaround is to sed the hostname and | 17:43 |
cjwatson | domain you want into place in the relevant files in /target from a preseed/late_command hook. | 17:43 |
hardwire | cjwatson: apologies.. I had to log in to my irc proxy and scroll through the logs this morning.. | 17:43 |
hardwire | I didn't immediately see anything hardwire oriented :) | 17:44 |
hardwire | but thanks. | 17:44 |
cjwatson | no problem, sorry I can't offer anything better | 17:44 |
hardwire | long standing means it's still in intrepid eh? | 17:45 |
cjwatson | yes | 17:45 |
hardwire | doheth | 17:45 |
hardwire | hmm.. cjwatson if I could theorize for a moment.. couldn't I just "echo 10.0.0.22 servername.domain servername >> /etc/hosts" on the installers root? | 17:46 |
hardwire | the dns resolver would prefer that anyways. | 17:47 |
hardwire | not much help for DHCP folk.. | 17:47 |
hardwire | but hey.. | 17:47 |
cjwatson | if you did it before it copied those files into /target, sure | 17:47 |
hardwire | I'm gonna give it a go. | 17:47 |
cjwatson | and *after* it has generated them itself, otherwise it'll just blat what you did | 17:47 |
cjwatson | so you can't do it in preseed/early_command | 17:47 |
hardwire | oh snap | 17:48 |
cjwatson | I'm surprised the DNS resolver cares what the local hostname is | 17:48 |
cjwatson | oh, I suppose if you're giving a non-FQ name as the mirror or something | 17:48 |
cjwatson | you could fully-qualify the mirror as a workaround | 17:48 |
hardwire | I'm surprised that the installer cares at all. | 17:49 |
cjwatson | it shouldn't, unless you're using a non-FQ mirror | 17:49 |
cjwatson | even then I'm surprised that it cares | 17:49 |
hardwire | so the hosname and domain name of the installed workstation for some reason depends on if the mirror is fully qualified or not? | 17:49 |
hardwire | that seems like chaos. | 17:49 |
cjwatson | no, obviously not | 17:49 |
cjwatson | I meant that if the mirror wasn't fully-qualified then it might use the local domain as a base for resolution | 17:50 |
hardwire | ahha | 17:50 |
hardwire | "search" in /etc/resolv.conf | 17:50 |
cjwatson | but looking at your blog I see you're saying it's doing a reverse DNS lookup on the local host name, so that's something different, ignore me | 17:50 |
hardwire | you're allowed | 17:50 |
cjwatson | does that only happen when you try to ssh in? | 17:50 |
hardwire | ssh into the installer? | 17:51 |
cjwatson | that's one step that I would expect to do an RDNS check | 17:51 |
cjwatson | oh, you're not doing that either | 17:51 |
hardwire | yar | 17:51 |
cjwatson | therefore I'm confused. Are you sure the RDNS check is happening during installation, not later? | 17:51 |
hardwire | the installer is simply... guessing.. what my hostname should be. | 17:51 |
cjwatson | sure, I know | 17:51 |
hardwire | it's all happening during the install | 17:51 |
cjwatson | but I'm asking whether just fixing it up at the end of the installation would be sufficient | 17:51 |
hardwire | I've tested it in subnets that have no RDNS.. and things work fine | 17:52 |
cjwatson | no, stop, listen :) | 17:52 |
cjwatson | if it doesn't make the installer go significantly more slowly, it doesn't matter whether you arrange for the installer to get it right to start with, or if you just have a preseed/late_command hook fix it up at the end | 17:52 |
cjwatson | therefore, it matters whether the RDNS check that's causing the delay is happening during installation, or after first boot | 17:52 |
cjwatson | in your blog, you're talking about waiting for the server to come up by polling port 22, which wouldn't distinguish between those two cases | 17:53 |
cjwatson | that's why I'm asking | 17:53 |
hardwire | ah. | 17:53 |
hardwire | the end step was an attempt at humor | 17:53 |
hardwire | booting a remote server with no i/o access to it, and waiting for it to install ubuntu then completely reboot itself with openssh installed. | 17:54 |
hardwire | the tick tock of the clock while you panic wondering if your evil plan worked or not | 17:54 |
hardwire | that's all | 17:54 |
cjwatson | not unreasonable, people do do that :) | 17:54 |
cjwatson | but if you aren't doing that, do you have more information about which stage the delay happens at? | 17:54 |
hardwire | my server is srv001.sb.brutetech.com and for some reason serverbeach set my RDNS to brutetech.com | 17:54 |
hardwire | when the machine comes back up after a fresh install it has the hostname "brutetech" and "search com" is in /etc/resolv.conf | 17:55 |
hardwire | you know this of course | 17:55 |
hardwire | There is no delay when dealing with RDNS | 17:55 |
hardwire | it's just odd that RDNS is being used above my own preferences, which you've already told me I need a fixup for. | 17:55 |
cjwatson | oh, right, I thought you were talking about the installer taking 30 minutes longer *due* to this bug or something like that. (That sort of thing does happen so I assumed that's what you were asking about.) | 17:57 |
hardwire | so are we still thinking the same bug on this? | 17:57 |
cjwatson | yes, same bug | 17:58 |
cjwatson | I suggest something like this: | 18:00 |
cjwatson | d-i preseed/late_command string sed -i 's/brutetech/srv001/' /target/etc/hostname; sed -i 's/brutetech.*/srv001.sb.brutetech.com srv001/' /target/etc/hosts; sed -i 's/search.*/search sb.brutetech.com/' /target/etc/resolv.conf | 18:00 |
cjwatson | lo-tech | 18:00 |
cjwatson | massage to taste | 18:00 |
hardwire | cjwatson: ok.. I'm gonna try a few things | 18:00 |
hardwire | Mainly because I don't like my lvm's being named wrong | 18:01 |
cjwatson | oh, well, that bit *is* overridable | 18:02 |
cjwatson | d-i partman-auto-lvm/new_vg_name string whatever | 18:03 |
hardwire | I like your fixup for my scenario | 18:03 |
hardwire | but it doesn't match other peoples rdns for their own IP's.. etc.. and I was hoping to get a howto out via that blog entry :) | 18:03 |
hardwire | which, thankfully, I can still edit. | 18:04 |
hardwire | I will succeed, mark my words. | 18:04 |
hardwire | are you marking them? | 18:05 |
cjwatson | :-) I have too much to do to fix this for 8.10, unfortunately, and Debian being more or less frozen makes it awkward too; but if nobody else beats me to it I expect to have a go at it for 9.04 | 18:05 |
hardwire | wowser | 18:16 |
hardwire | there is no if statement surrounding seed_hostname_from_dns | 18:16 |
hardwire | how absolutely strange | 18:17 |
* hardwire tries to find a good hook to use | 18:18 | |
hardwire | cjwatson: I'm moving on for now.. requesting my RDNS be set right in the first place, or not at all. | 18:24 |
hardwire | but it's a strange world in netcfg-1.40ubuntu5/static.s | 18:25 |
hardwire | c | 18:25 |
hardwire | I have no idea why there is no check if the hostname & domain is set up before seeding values from dns | 18:25 |
cjwatson | as Joey says in the linked bug, it is "quite broken" | 18:37 |
hardwire | cjwatson: I may be barking up the wrong tree here but.. even I could fix this. | 18:49 |
hardwire | so what's the holdup? | 18:49 |
hardwire | is there a decision pending on this very thing over at Debian HQ Global Inc? | 18:49 |
cjwatson | err, no, just nobody's got round to it | 18:56 |
cjwatson | personally, every time I remember about it, there's a freeze of some sort on so I go and do something else :) | 18:56 |
cjwatson | it's rather delicate since you need to take into account a variety of code paths and preseeding possibilities | 18:57 |
hardwire | I understand that all too well | 19:02 |
hardwire | d-i preseed/early_command string echo 127.0.0.1 localhost > /etc/hosts; echo 72.51.32.92 srv001.sb.brutetech.com srv001 | 19:45 |
hardwire | does anything need to be escaped/quoted there? | 19:45 |
hardwire | not sure how the parser for preseed stuff works | 19:45 |
hardwire | it appears that netcfg.../static.c writes /etc/hosts AFTER the dns lookup | 19:46 |
hardwire | so writing it before that time will cause the DNS lookup to resolve whatever I want. | 19:46 |
hardwire | store that in memory.. then override the /etc/hosts file with the correct information | 19:46 |
* hardwire waits for the remote install to verify his evil plan | 20:12 | |
cr3 | if I want to modify the content of an iso, I will need to sign the Release file but how can I make my key available? | 20:13 |
cr3 | a simpler way might be to set debian-installer/allow-unauthenticated | 20:26 |
=== superm1|away is now known as superm1 | ||
hardwire | cr3 you could.. | 22:57 |
hardwire | oh | 22:57 |
=== superm1 is now known as superm1|away |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!