=== harlowja is now known as harlowja_away | ||
smoser | utlemming, https://code.launchpad.net/~utlemming/cloud-init/sdc-op-script/+merge/206065 | 00:52 |
---|---|---|
smoser | the only thing wrong with a boothook is that it doesnt guarantee network. | 00:52 |
=== harlowja_away is now known as harlowja | ||
=== pquerna_ is now known as pquerna | ||
=== harlowja_ is now known as harlowja_away | ||
=== shardy_afk is now known as shardy | ||
smoser | harlowja_away, one comment on datasource openstack | 12:01 |
smoser | can't 'wait_for_url' | 12:01 |
smoser | i know it sucks | 12:01 |
smoser | but you can't | 12:01 |
smoser | only ec2 can be annoying | 12:01 |
smoser | ie, we want to be able to have both OpenStack and EC2 enabled | 12:02 |
smoser | and not block boot for some stupid seconds | 12:02 |
smoser | harlowja_away, wake up. man. | 16:15 |
smoser | utlemming, did you see my half written response to the smart os change? | 16:15 |
utlemming | smoser: yeah, and it makes sense | 16:16 |
utlemming | smoser: I'll look at it later | 16:16 |
smoser | ok. i really, really, really want an upload of cloud-init today | 16:16 |
smoser | there is a bunch of stuff in | 16:16 |
harmw | smoser: will we see an new cirros release this month aswell? | 17:46 |
smoser | :) | 17:47 |
smoser | i've been meaning to do that. | 17:47 |
smoser | pathetically, one of my motivations for doing so is so i can get in a icehouse commit log :) | 17:47 |
harmw | nice, though how does cirros fit into icehouse commitlogs? ;p | 17:48 |
smoser | devstack | 17:49 |
smoser | i'll change the default cirros used to be the released version. | 17:49 |
harmw | nice nice | 17:52 |
smoser | what do you use cirros for harmw? | 17:55 |
harmw | testing openstack, or something else where I 'just want a vm shell' | 18:03 |
harmw | instead of booting fedora/centos | 18:03 |
harmw | which take considerably longer to boot :) | 18:05 |
=== harlowja_away is now known as harlowja_ | ||
harlowja_ | smoser waked up, lol | 18:17 |
harlowja_ | smoser wait_for_url, can be removed :-P | 18:22 |
mfisch | is $INSTANCE_ID only available at bootcmd? It didn't seem to be there at runcmd time | 18:23 |
=== shardy is now known as shardy_afk | ||
smoser | harlowja_, i just about have it i think. was testing that it actually worked | 18:43 |
smoser | then gonna merge it. | 18:43 |
smoser | hm.. | 18:44 |
smoser | mfisch, the doc says it is. | 18:44 |
smoser | i'm ok to put it in at runcmd also. | 18:44 |
harlowja_ | smoser did u get the overall change, basically unifying cfgdrive and osmetadata datasource, being backed by a common class, and then only really changing in the child classes how the data is fetched and from where (url or disk) | 18:45 |
smoser | mfisch, it does appear that it should be there for you in bootcmd and in boothooks, but it is only done there. | 18:46 |
smoser | harlowja_, yeah, it looks good. | 18:46 |
harlowja_ | smoser k | 18:46 |
harlowja_ | thx boss | 18:46 |
harlowja_ | :-P | 18:46 |
smoser | harlowja_, http://paste.ubuntu.com/6927014/ | 18:48 |
smoser | thats my changes so far. | 18:48 |
harlowja_ | smoser seems ok, although should i just remove 'wait_for_url' | 18:49 |
harlowja_ | cause i know we all love waiting for datasources to come online, lol | 18:50 |
harlowja_ | and watchign the log | 18:50 |
smoser | harlowja_, well, i had removed wait_for_url | 18:52 |
smoser | but then i put it back | 18:52 |
harlowja_ | :-P | 18:52 |
smoser | i figured its more useful being configurable | 18:52 |
harlowja_ | ah | 18:52 |
smoser | if you had a openstack where it just wasn't working | 18:52 |
smoser | ie, it didn't come up right way, you could turn max_wait up | 18:52 |
smoser | and it would retry | 18:52 |
harlowja_ | ya | 18:53 |
harlowja_ | looks good to me then | 18:53 |
harlowja_ | shall i patch or u? | 18:53 |
smoser | merged. | 18:55 |
harlowja_ | k | 18:57 |
harlowja_ | i know the heat folks and others wanted , https://code.launchpad.net/~harlowja/cloud-init/ec2-ssl ? | 18:57 |
harlowja_ | although in general since i don't use ssl for this stuff anyway (cfgdrive stil..) up to u :-P | 18:58 |
smoser | harlowja_, didn't i pull that lareday? | 18:58 |
harlowja_ | oh | 18:58 |
harlowja_ | possibly | 18:58 |
harlowja_ | will double check | 18:58 |
smoser | if not i will | 18:59 |
smoser | seems not | 18:59 |
smoser | harlowja_, there are merge conflicts there | 19:01 |
smoser | can you resolve those really quick ? | 19:01 |
harlowja_ | k, one sec, writing up little speaker summary/abstract for atalanta summit stuff | 19:02 |
smoser | ooh. fancy. | 19:03 |
smoser | harlowja_, i'd also like to get the tox stuff too. but can't really pull it if it means it doesn't work "offline" | 19:06 |
harlowja_ | ya, understandable | 19:06 |
smoser | harlowja_, random question. | 19:21 |
harlowja_ | yo | 19:21 |
smoser | i'd like to have cloud-init write some "state" to /run/cloud-init | 19:22 |
harlowja_ | k | 19:22 |
smoser | i dont have any ideas for a mechanism on doing tha tthat doesn't result in potential race conditions on reading it | 19:22 |
smoser | ie, cloud-init writing the state file while a reader is reading it | 19:22 |
smoser | do you? | 19:22 |
harlowja_ | write to temp state file, then swap it? | 19:23 |
harlowja_ | mv is atomic i think in linux | 19:23 |
harlowja_ | who's the readers in this case? users | 19:23 |
harlowja_ | ? | 19:23 |
smoser | readers is users, yes (or tools) | 19:36 |
smoser | mv is atomic, yes. but if i write state to 'cloud-init.state', then go to update it. | 19:36 |
smoser | then that is not atomic. | 19:36 |
smoser | mv cloud-inti.state cloud-init.state.old && mv cloud-init.state.new cloud-init.state | 19:37 |
smoser | would leave a race. | 19:37 |
smoser | probably over thinking | 19:37 |
smoser | :) | 19:37 |
med_ | smoser, how do I ask cloud-init to re-run | 19:39 |
smoser | :) | 19:39 |
mfisch | med_ is asking because I want to test a feature w/o rebuilding an image | 19:40 |
med_ | (ie, on an image after I locally hack the cloud-init code0 | 19:40 |
* med_ is mfisch proxy | 19:40 | |
smoser | what i do very commonly is | 19:40 |
smoser | rm -Rf /var/lib/cloud /var/log/cloud-init* | 19:40 |
smoser | reboot | 19:40 |
med_ | He made me drive him to work and now he asks me to ask questions | 19:40 |
med_ | nod. that's it. | 19:40 |
smoser | you can be less invasive too | 19:40 |
* med_ knew there was a way | 19:40 | |
smoser | and you can run modules by themselves | 19:40 |
smoser | cloud-init single | 19:40 |
mfisch | smoser: I'm toying with the idea of having access to INSTANCE_ID inside the landscape module | 19:41 |
mfisch | although I wonder if landscape uses hostname by default, /me checks | 19:41 |
med_ | mfisch, sounds like you're still working for (C) | 19:41 |
med_ | j/k | 19:41 |
smoser | mfisch, so 2 things there.. | 19:44 |
smoser | a.) i'm not opposed to having INSTANCE_ID just in the environment whenever we set it | 19:44 |
smoser | b.) you do have access to the instance id in the landscape module | 19:44 |
mfisch | b) not currently | 19:44 |
mfisch | and its not a shell script like bootcmd is | 19:45 |
mfisch | so I'd have to do a pass of the dictionary we're building | 19:45 |
smoser | you do have access. | 19:45 |
smoser | you get 'cloud' passed in | 19:45 |
smoser | that has it. | 19:45 |
mfisch | ah cool | 19:46 |
mfisch | would a simple check like if the dict has a key of $INSTANCE_ID set it there work? | 19:46 |
mfisch | may want to make that more generic | 19:47 |
smoser | mfisch, ? | 19:49 |
smoser | the 'cloud' object has 'get_instance_id()' | 19:49 |
smoser | and it proxies most things through to cloud.datasource. | 19:50 |
mfisch | smoser: I mean in the config file, we need a token that says "substitute this with the instance ID" | 19:50 |
smoser | i'm confused now. | 19:51 |
mfisch | smoser: the landscape stuff gets loaded as a dictionary in that file | 19:53 |
mfisch | account: foo | 19:54 |
mfisch | for example | 19:54 |
mfisch | computer-title could default to instance id if unset | 19:54 |
mfisch | or something like if dict[computer-title] == "$INSTANCE_ID" then do the substitution | 19:55 |
mfisch | smoser: let me propose a fix and maybe it makes more sense | 19:55 |
smoser | http://paste.ubuntu.com/6927388/ ? | 19:57 |
mfisch | smoser: yes, exactly | 19:57 |
smoser | that doesn't do any template/rendering, but functional | 19:57 |
mfisch | smoser: I will test that | 19:58 |
mfisch | smoser: it screwed up the ordering in the config file to change it there, I'll look into it | 20:07 |
mfisch | also instance human readable name, whatever that is would work better there | 20:08 |
smoser | mfisch, yeah, yaml wont write in any predictable order | 20:15 |
mfisch | smoser: it's sticking it above [client] which confused landscape | 20:18 |
smoser | oh? | 20:18 |
mfisch | yeah it comes out as | 20:18 |
mfisch | computer-title = foo | 20:18 |
mfisch | [client] | 20:18 |
mfisch | ... | 20:18 |
smoser | oh. yeah, it would | 20:18 |
smoser | :) | 20:18 |
smoser | cause i put it there. | 20:18 |
smoser | i didn't realize that. | 20:18 |
smoser | just put it in the right place. | 20:18 |
smoser | mfisch, http://paste.ubuntu.com/6927527/ | 20:20 |
mfisch | smoser: I | 20:21 |
mfisch | smoser: I just moved it up and checked if it was in ls_cloudcfg | 20:21 |
mfisch | above the merge | 20:21 |
smoser | ewll that paste there should basically fill it into instance id if its not present. | 20:22 |
mfisch | smoser: yep and that explains why my idea didnt work | 20:27 |
mfisch | smoser: sorry for the delay, this is what I ended up doing based on the 12.04 version: | 20:48 |
mfisch | http://paste.ubuntu.com/6927655/ | 20:48 |
smoser | mfisch, well, if you do it after the mergeTogether | 20:58 |
smoser | then you ensure that you're lowest priority | 20:58 |
smoser | ie, merge all the stuff the way it does | 20:58 |
smoser | then only fill it if its not set elsewhere | 20:58 |
zooz | hi smoser, It's Vaidas. We've been talking over email re GCE data source | 20:59 |
mfisch | smoser: true | 21:00 |
smoser | zooz, hey. | 21:04 |
smoser | trunk has gce now. woohoo. | 21:04 |
smoser | (possibly not working as i might have broke your code :) | 21:05 |
zooz | cool, I am about to look at your improvements and add them to my changes | 21:05 |
zooz | I have got proper tests now, handling of user-data and more | 21:05 |
zooz | smoser, no worries, I will test in GCE before I do a pull request | 21:05 |
zooz | in fact I have been using cloud-init in GCE without any issues for some time now | 21:07 |
zooz | smoser, a merge request is on the way for your review. I have tested it in GCE and ran unit tests. No issues. | 21:20 |
smoser | zooz, thank you. | 21:22 |
smoser | i plan on a cloud-init upload to ubuntu tonight. | 21:22 |
zooz | great | 21:23 |
smoser | so this will make its way into cloud images for trusty tonight | 21:23 |
zooz | awesome! | 21:23 |
zooz | done | 21:28 |
zooz | let me know if any issues | 21:28 |
med_ | thanks smoser. | 21:42 |
med_ | does cloud-init get SRUd (or have a MRE for prior releases)? | 21:43 |
smoser | well, SRU. | 21:43 |
smoser | but for "hardware" enablement, sru team has let things in. | 21:44 |
med_ | nodz. | 21:44 |
smoser | it is honestly though, somewhat of apain to port back to 12.04 at this point. although datasources are fairly standalone. | 21:44 |
med_ | 'k | 21:44 |
* med_ will just take a look at rmadison in the morning | 21:44 | |
smoser | zooz, still there? | 21:52 |
zooz | smoser, yeah | 21:52 |
smoser | is there a reason you'd not want the dns resolution? | 21:52 |
smoser | the way i did it it actually works (should work) in gce | 21:52 |
smoser | and will fail quickliy (dns resolution) just about anywhere else | 21:52 |
zooz | smoser, no particular reason, I can test it using dns name | 21:53 |
smoser | without attempting to get access to a 169.254.169.254 that might be firewalled off (and socket level timeout) | 21:53 |
zooz | good point | 21:53 |
zooz | let me do a quick test | 21:53 |
zooz | smoser, it does work in GCE using dns name | 21:54 |
smoser | ok. so lets keep that part, that will save people elsewhere from doing a request to that url. | 21:55 |
smoser | (unless there is a dns entry external there, which would be unlikely) | 21:55 |
zooz | smoser, do you want me to quickly add it back? | 21:56 |
smoser | sure. | 21:56 |
smoser | i have to go now for a couple hours. | 21:56 |
smoser | so i'll take a look again later. | 21:56 |
zooz | thanks | 21:57 |
harlowja_ | smoser alrihgt, i got some free time | 23:27 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!