haobug | hi hi, got problem again. http://pastebin.com/XxaqxmbJ after about 3 days of waiting... | 01:37 |
---|---|---|
=== jml_ is now known as jml | ||
cjwatson | haobug: You don't need to redo it all from scratch, at least, but there's some manual work to do to recover from a failure there without re-downloading - it should be possible to look at /usr/share/lxc/templates/lxc-ubuntu and search for "Installing updates" to find where it got to, but you'll have to trace the function call sequence up to there for yourself | 14:41 |
cjwatson | http://people.canonical.com/~cjwatson/tmp/lp-livefs-built.png progress! | 21:09 |
cjwatson | five-ish to-do items remaining, but pretty close. | 21:12 |
cjwatson | unfortunately will need another lp-buildd deployment. | 21:12 |
cjwatson | wgrant: Perhaps we could try to squeeze a launchpad-buildd update in with RT#69868, since the risk of buildd upgrades seems to be more strongly correlated with how many of them we do rather than how much they contain? I'd like to include https://code.launchpad.net/~cjwatson/launchpad-buildd/abort-race/+merge/218222 if possible, as well as the fairly obvious stuff that's in trunk. | 21:43 |
_mup_ | Bug #69868: Contrast setting is bad <gxine (Ubuntu):Invalid> <https://launchpad.net/bugs/69868> | 21:43 |
cjwatson | (I haven't yet done a QA pass on the whole lot.) | 21:45 |
cjwatson | I've been meaning to do that abort-race stuff for months, but doko keeps running into it so I thought I'd better make an effort to sort it out ... | 21:46 |
wgrant | cjwatson: Looking | 21:48 |
wgrant | cjwatson: Is that screenshot of a full test run with a local buildd? Nice. | 21:49 |
cjwatson | It is indeed. | 21:54 |
cjwatson | Not that that took ages to set up or anything. | 21:54 |
cjwatson | Remaining to-dos: some bits of view need to be disabled if the feature flag is off; latest builds table for +livefs; UI for requesting a build; a way to delete a LiveFS (much like for recipes) | 21:58 |
wgrant | I should probably rewrite the docs, as they don't really work well when not running LP in a heavyweight VM. | 21:58 |
cjwatson | I'm afraid I ignored the docs as I was fairly sure I knew which bits I needed | 21:59 |
wgrant | Do we need UI for requesting a build? | 21:59 |
wgrant | Deletion is probably easy, as the LiveFS can just take all the LiveFSBuilds and their files with it. | 21:59 |
cjwatson | I expect to eventually get to the point where some builds don't require cdimage, at which point a UI would be useful | 22:00 |
cjwatson | deletion - I was expecting to do what we do with recipes, i.e. leave the builds intact but disconnected | 22:00 |
wgrant | Why? | 22:02 |
wgrant | It's basically impossible to find a build if you don't know the livefs. | 22:02 |
cjwatson | just seems odd to delete builds that happened | 22:02 |
wgrant | We avoid deleting SPRBs because that would leave SPRs without a signature and with no way to trace the owner. | 22:02 |
cjwatson | well, ok, I guess I can do either way | 22:03 |
cjwatson | my local setup was basically sampledata, initseries trusty, run the initialize job, feed in the production chroot, install launchpad-buildd listening on a port not one of those already in sampledata, add that builder | 22:04 |
wgrant | What value does a LiveFSBuild with no LiveFS provide? How would you use it, or even find it (except by looking through all the builds for an archive or builder)? | 22:04 |
wgrant | Hm, you were able to get it working in a container? | 22:05 |
cjwatson | wouldn't be any good for normal package builds, but it worked for this | 22:05 |
cjwatson | I had to make the container allow arbitrary mounts, with some help from hallyn | 22:05 |
wgrant | Ah, right. | 22:05 |
cjwatson | almost exploded my system on the first attempt since it enabled qemu-x86_64 in my host :-) | 22:05 |
wgrant | Heh | 22:06 |
wgrant | I normally just run a KVM instance which lives on lxcbr0. | 22:06 |
cjwatson | LiveFSBuilds> I was thinking builder history views and the like, but you're probably right. I didn't know there was a specific reason for SPRBs. | 22:06 |
cjwatson | The use of old LiveFSBuilds is their logs. | 22:07 |
wgrant | Right, but if you want the logs you probably shouldn't be deleting the thing that lets you find them :) | 22:07 |
cjwatson | Perhaps I should have a way to disable a LiveFS | 22:07 |
cjwatson | Not deleted, but it wouldn't allow new builds any more | 22:07 |
cjwatson | Basically to avoid stupid mistakes | 22:08 |
cjwatson | Given that I'm not even sure we particularly need deletion | 22:09 |
wgrant | We need to be able to delete builds individually, and also the LiveFS and its builds as a whole unless there's a good reason not to | 22:10 |
cjwatson | What's the point of deleting individual builds? | 22:10 |
cjwatson | I don't think any other build types are individually deleteable, are they? | 22:11 |
wgrant | There doesn't need to be UI or API for it | 22:11 |
wgrant | but I'm working on the other types as we speak. | 22:11 |
cjwatson | Just the principle that all objects should be deleteable? | 22:11 |
wgrant | Though those are just being neutered, rather than deleted, since they produce artifacts that are referenced elsewhere. | 22:12 |
wgrant | Mostly for proper PPA deletion | 22:12 |
cjwatson | A LiveFSBuild associated with a PPA *could* just be neutered, since it doesn't actually build something *in* the PPA as such, it just uses it as a source; but there's probably not much use for such a build | 22:13 |
wgrant | Right | 22:13 |
wgrant | We delete where we can, neuter where we can't. | 22:14 |
wgrant | Keeping useless data around isn't something we should be doing. | 22:14 |
cjwatson | But I thought the database was too small | 22:14 |
cjwatson | I guess we do erase evidence of builds that happened in some other situations albeit in different ways, e.g. cancel/retry | 22:16 |
cjwatson | Anyway, thanks. Bed then bank holiday, I think | 22:16 |
wgrant | Right. It's theoretically nice to have a full history of everything, but in practice it's somewhat onerous and not terribly useful. | 22:17 |
wgrant | It's rare that I need to look way back in a builder's history. | 22:17 |
wgrant | And for short-term stuff I mostly look at logs. | 22:17 |
=== bradm1 is now known as bradm |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!