meena | ebal[m]: i have to look at that code again, because i don't remember if we correctly handle the complete failure of fallocate | 07:17 |
---|---|---|
meena | else: | 07:25 |
meena | try: | 07:25 |
meena | create_swap(fname, size, "fallocate") | 07:25 |
meena | except subp.ProcessExecutionError as e: | 07:25 |
meena | LOG.warning(errmsg, fname, size, "dd", e) | 07:25 |
meena | LOG.warning("Will attempt with dd.") | 07:25 |
meena | create_swap(fname, size, "dd") | 07:25 |
meena | it should fallback to dd | 07:25 |
ebal[m] | meena: | 09:14 |
ebal[m] | * meena: do you feel this is a bug ? | 09:14 |
meena | ebal[m]: sooo, somebody was saying this is a bug in Linux… which, i guess we can agree on, buuuuuutttt, we have enough workarounds for other software's bugs, so, i dunno… | 09:17 |
meena | ebal[m]: is this ⬆️the code you have too on your system? | 09:18 |
* ebal[m] sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/aJtMUKQOJyBmujmecvNJECUy/message.txt > | 09:21 | |
meena | so, the code looks the same, and it should fallback to `dd` | 09:22 |
meena | ebal[m]: can you post again the log? | 09:26 |
ebal[m] | <meena "ebal: can you post again the log"> https://pastebin.com/Wu7cdBCC | 09:27 |
meena | so, Unexpected error is not subp.ProcessExecutionError | 09:29 |
meena | https://github.com/canonical/cloud-init/blob/master/cloudinit/subp.py#L82 ← at the end it throws IOError | 09:37 |
meena | oh, that's from ProcessExecutionError | 09:37 |
otubo | Can't believe I missed the first day! My mind really tricked me that it would be today and tomorrow. AHHH | 11:38 |
otubo | lucasmoura: Good to see more BR joining the project! :-) Bem vindo! | 11:39 |
otubo | I'm recaping the slides from yesterday, looks like I need to bother rick_h now instead of Odd_Bloke :-) | 11:40 |
otubo | Oh just saw the emails correcting the dates, so I'm no completely crazy, not yet. | 11:42 |
lucasmoura | thanks otubo :) | 12:44 |
rick_h | otubo: apologies, they're all updated now | 12:58 |
ananke | hmm, trying to wedge repo key on debian is becoming a challenge. I can't simply drop a gpg key in /etc/apt/trusted.gpg.d/, because those need to be in a keyring format | 12:58 |
ananke | the proper format can be obtained by piping the gpg key via 'apt-key add', but that fails on lack of gnupg | 12:59 |
rick_h | otubo: there were a couple of questions to the community for input, especially on commit-bits and new user ramp up that if you get time to think about would be great to get your input. | 13:00 |
rick_h | you can see where we got in the running open notes document link | 13:00 |
otubo | rick_h: I'll definitely check on that! I believe I already got access to all docs listed on the post. | 13:01 |
rick_h | ok cool | 13:06 |
ananke | hah, I may have to resort to using AllowUnauthenticated | 13:17 |
Odd_Bloke | ananke: What version of apt are you using? Even oldstable should support ASCII-armored keys in /etc/apt/trusted.gpg.d/, per https://manpages.debian.org/testing/apt/apt-key.8.en.html#SUPPORTED_KEYRING_FILES | 13:31 |
Odd_Bloke | meena: ebal[m]: So I don't think that we should be working around what should be a short-term kernel bug in a permanent way in cloud-init, but there _is_ a bug: `create_swap` (the function defined within `create_swapfile`) catches ProcessExecutionError but doesn't re-raise it. This means that the `create_swap(fname, size, "fallocate")` call will never raise a ProcessExceptionError and so we will never | 13:37 |
Odd_Bloke | fallback from failed fallocate swap creation to dd swap creation. | 13:37 |
ebal[m] | <Odd_Bloke "meena: ebal: So I don't think th"> Do you need me to file a proper bug report for this ? | 13:38 |
Odd_Bloke | ebal[m]: Yes, please! | 13:40 |
ebal[m] | I also was thinking if it is of value something like this (just a draft - I am sure this does not work) regarding the swap method | 13:41 |
ebal[m] | https://github.com/ebal/cloud-init/commit/58f1819a2c8b880a325ddce9119032d38a0f1603 | 13:41 |
ananke | Odd_Bloke: I'm doing this on Debian 10.5 AMI, so I imagine it's fairly up to date | 13:45 |
ebal[m] | .... so .... to file a bug , should canonical hire me first ? what's up with this ubuntu one account ? do I need to pledge my loyally to Mark as my lord and savior ? | 13:45 |
ananke | Odd_Bloke: the gpg key I have is in ' PGP public key block Public-Key (old)' format, while the existing keys in trusted.gpg.d are in 'PGP/GPG key public ring (v4) created Sun Apr 14 07:48:55 2019 RSA (Encrypt or Sign) 4096 bits MPI=0xb15c6294ca2e5d16' | 13:47 |
ananke | but I may need to try that gpg --armor --export format | 13:49 |
Odd_Bloke | ananke: Yeah, I would _expect_ exporting like that to work. | 14:00 |
Odd_Bloke | ebal[m]: Not to sign up for an Ubuntu One account; there is a CLA you'd need to sign to contribute code, if that would be a blocker for you. | 14:00 |
ebal[m] | it's too late now, all my personal data belongs to canonical ! | 14:01 |
Odd_Bloke | ebal[m]: In terms of that proposed change, I think we need to think about that interface a little more: as that patch is currently written, specifying "fallocate" would mean "fallocate unless it fails then dd" whereas "dd" would just mean "dd". | 14:02 |
ananke | Odd_Bloke: so after testing it, it appears that this won't work, and I'm now guessing that apt-key's manual refers to _input_ files when adding new keys via apt-key | 14:09 |
ebal[m] | Done, hopefully provides all the necessary info to everybody | 14:13 |
ebal[m] | https://bugs.launchpad.net/cloud-init/+bug/1897099 | 14:13 |
ebal[m] | thanks | 14:13 |
ubot5 | Ubuntu bug 1897099 in cloud-init "create_swap do not fallback to dd when fallocate fails" [Undecided,New] | 14:13 |
ebal[m] | <Odd_Bloke "ebal: In terms of that proposed "> I totally agree with you on that, this was written at around 01.30 in the night - just a rough idea to get some initial feedback - if it make sense. It does for me to have an option to use the swap method - but it may not be important or useful for anybody else. So instead of spending time there - I could spend time to work on a workaround for personal use. | 14:16 |
ananke | argh, so for the gpg file to be valid in /etc/apt/trusted.gpg.d/, you have to do something like: | 14:23 |
ananke | wget -q -O - https://deb.parrotsec.org/parrot/misc/parrotsec.gpg | gpg --no-default-keyring --keyring ./tempfile.gpg --import | 14:23 |
ananke | gpg --no-default-keyring --keyring ./tempfile.gpg --export > valid_keybox_file.gpg | 14:24 |
ananke | the problem is, the resulting file is a binary | 14:24 |
ananke | ahh, write_files supports base64! I may do it that way then | 14:25 |
Odd_Bloke | ananke: Aha, that makes sense; apologies for the misdirection! | 14:32 |
Odd_Bloke | Yeah, base64 is probably the way to go. | 14:32 |
ananke | no worries. It's a bit big, close to 10k after base64 encoding, but we'll see if that works | 14:38 |
ananke | 9665 bytes base64 encoded, 9457 bytes gzipped + base64 | 14:39 |
ananke | well I'll be damned, this is also not possible: ==> amazon-ebs: Error creating launch template for spot instance: InvalidUserData.Malformed: User data is limited to 16384 bytes. | 15:02 |
=== tds9 is now known as tds | ||
ananke | going to trim my cloud-init yaml as much as possible, see if I can fit under that limit | 15:08 |
otubo | ebal[m]: Odd_Bloke Does raising an exception from inside create_swap() make sense? https://pastebin.com/2Qv4Npwe | 15:21 |
otubo | Odd_Bloke: just saw your PR which is basically the same thing but better. | 15:48 |
Odd_Bloke | otubo: Yep, apologies, was "just" testing it before I dropped the link here (and then I went into a meeting), hope you didn't duplicate too much work! | 15:49 |
otubo | Odd_Bloke: nah :) | 15:49 |
blackboxsw | community notice: hey folks "day 2" of cloud-init virtual summit is today in +30 mins. If you are interested in participating, feel free to drop in for any time you have available https://meet.google.com/rup-mbis-kyr | 16:31 |
blackboxsw | virtual cloud-init summit detailed schedule/intent is here: https://discourse.ubuntu.com/t/cloud-init-online-summit-sept-23-24/17867 | 16:32 |
blackboxsw | community notice: Just about to kick off day 2 virtual cloud-init summit: https://meet.google.com/rup-mbis-kyr see you there | 17:01 |
=== ijohnson is now known as ijohnson|lunch | ||
smoser | falcojr: this is probably a personal rant. but personally, 'client.exec("your command here")' is way inferior to client.exec("your", "command", "here") | 17:30 |
smoser | in that you forced shell escaping on the user (and you have not identified what the shell is or should be) | 17:30 |
=== ijohnson|lunch is now known as ijohnson | ||
meena | Odd_Bloke: oy. from reading the code i didn't catch that (ha ha ha) | 17:52 |
Odd_Bloke | meena: I was happy to raise it, I won't take exception. | 17:53 |
Odd_Bloke | meena: Do you happen to know if netifaces works on BSD? | 17:54 |
meena | gimme a sec | 17:55 |
Odd_Bloke | It works on OS X, that's basically the same thing, right? | 17:56 |
meena | Odd_Bloke: meena@fbsd12-1 /u/h/meena> pkg search netifaces | 17:56 |
meena | py27-netifaces-0.10.9 Getting network addresses from Python 3 | 17:56 |
meena | py37-netifaces-0.10.9 Getting network addresses from Python 3 | 17:56 |
meena | soon to be only py37- ;) | 17:56 |
Odd_Bloke | Well that's an encouraging sign, at least. | 17:56 |
meena | lemme boot up netbsd or openbsd | 17:57 |
Odd_Bloke | rharper: Have you looked at netifaces at all? If so, do you know how it compares/relates to cloudinit.net? | 17:58 |
meena | or, i could look here: https://pkgsrc.se/search.php?so=netifaces | 17:58 |
meena | blackboxsw: what's the libpcap library we looked into? | 18:00 |
rharper | Odd_Bloke: yeah, i've used in the past; and originally wrote the net layer in that; but we wanted to avoid the depedency | 18:00 |
blackboxsw | meena: https://github.com/vmware/cloud-init-vmware-guestinfo which uses python3-netifaces | 18:01 |
rharper | and generally, the output shown doesn't show us anything we can't already discover in the net layer now; | 18:01 |
blackboxsw | so Andrew Kutz is looking at how best to upstream that new datasource and discovery mechanisms | 18:01 |
meena | blackboxsw: i meant scapy | 18:03 |
meena | i did look into netifaces previously, btw… for… something | 18:03 |
meena | libioc | 18:03 |
akutz | Hi all! | 18:22 |
smoser | o/ | 18:22 |
smoser | very happy to have a vmware representative here. | 18:22 |
smoser | so , wrt cloud-init-vmware-guestinfo | 18:23 |
smoser | i think it fits fine as a local datasource. | 18:23 |
smoser | the datasource declared (in network config format) that "you should dhcp on all the things". | 18:24 |
smoser | cloud-init should do that as it does ephemeral network on other providers. | 18:24 |
smoser | then read the data that is there | 18:24 |
smoser | and declare that that was the metadata config. | 18:24 |
smoser | err... declare that that was the network config. | 18:25 |
smoser | the only difference from what you have, is that you are calling the network data that is in your 'metadata.yaml' file network config. | 18:25 |
smoser | it is not. it is pre-network config. | 18:25 |
otubo | akutz: also, how are the plans to merge it into cloud-init main tree? | 18:27 |
akutz | @otubo - Um, I do not know. That is why I am here, to ask y'all :) | 18:28 |
akutz | @smoser - re: pre-network config -- interesting. Did I misunderstand https://cloudinit.readthedocs.io/en/latest/topics/network-config.html? | 18:30 |
smoser | akutz: you did not misunderstand. | 18:30 |
smoser | there is no such thing, really. | 18:30 |
otubo | akutz: I'm very interested in having that on RHEL :) Since the current OVF DataSource is kinda bogus, changing the instance-id and stuff | 18:31 |
smoser | but i'm saying that what you are providing is *not* network config. | 18:31 |
smoser | but that cloud-init needs to utilize the information you are providing to tell it how to get the network config. | 18:31 |
smoser | and then when it has it, it would exit local datasource, let the OS bring up the networking, and move on. | 18:31 |
akutz | Hmm, but the cloud-init docs explicitly refer to what is being provided as the network configuration... | 18:32 |
akutz | It's the intended config, and it brings up the actual network, with the observed config. But that's more of an issue with the way the cloud-init docs are written IMO. | 18:32 |
smoser | alternatively... i dont see any problem with providing current network information through the same general mechanism as instance-data is being provided. | 18:32 |
blackboxsw | community notice: reconvening for live community discussion for virtual cloud-init summit: https://meet.google.com/rup-mbis-kyr | 18:34 |
ananke | this is very peculiar. enabling an additional apt-repo via cloud-init causes apt to fail to resolve DNS entries. it does not make sense at all | 19:24 |
blackboxsw | hrm, depends on what deb package is being installed? Something upgrading network utils? | 19:25 |
ananke | ahh, wonder if it has something to do with resolvconf | 19:29 |
ananke | good idea, thank you | 19:30 |
ananke | looking at cloud-init-output.log it appears that the 'update' ran just fine, but then 'upgrade' module is what starts to fail | 19:31 |
blackboxsw | Thanks folks for the attendance on cloud-init virtual summit. we are dropping into separate hacking/discussion channels for the remainder. We'll post links to available google meet breakout sessions if interested | 19:34 |
akutz | I am taking a bio break and will look for the link to the break out for vSphere cloud-init | 19:35 |
falcojr | akutz: I'll post the link soon...also taking a short break | 19:35 |
blackboxsw | topic 1: (bug discussions, primary AWS support regions without IMDSv2 support https://bugs.launchpad.net/cloud-init/+bug/1896532): meet room: vhttps://meet.google.com/zqb-tewn-qpx | 19:35 |
ubot5 | Ubuntu bug 1896532 in cloud-init "Ec2Datasource fails in environments without IMDSv2" [Undecided,New] | 19:35 |
blackboxsw | topic2: was I believe ongoing network refactor discussion https://cloudinit.readthedocs.io/en/latest/topics/hacking.html#ongoing-refactors | 19:37 |
akutz | falcojr: ack | 19:37 |
falcojr | vSphere discussion: https://meet.google.com/acr-vrnr-rab | 19:40 |
ananke | yep, /etc/resolv.conf was being obliterated, due to apt-get --upgrade's pulling of resolv.conf | 20:33 |
ananke | I mean pulling of resolvconf | 20:34 |
Odd_Bloke | rharper: I just put a bottle of the 2017 release of https://www.beeradvocate.com/beer/profile/32763/322625/ in the fridge, so I'm at least replicating part of the in-person summit experience we'd have had. :p | 21:05 |
rharper | oooo | 21:08 |
rharper | Odd_Bloke: nice! | 21:09 |
meena | mock me, why don't you https://github.com/canonical-web-and-design/cloud-init.io/pull/145#issuecomment-698538416 | 21:39 |
blackboxsw | haha "not yet" :) just you wait webteam-app I'll show you who is a collaborator | 21:45 |
blackboxsw | will ping the internal web team tomorrow if they have questions on the PR meena thanks for filing | 21:47 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!