johnsonshi_ | Hi upstream folks, I realized that cloud-init recently began using pycloudlib for integration tests. Azure support for integration tests was just added in https://github.com/canonical/cloud-init/pull/604 | 00:09 |
---|---|---|
johnsonshi_ | Do you know where the integration tests are defined in? I've looked around pycloudlib and cloud-init, but can't seem to find where the suite of integration tests are defined. | 00:10 |
vijayendra | rharper, smoser how can we set IPV6_FORCE_ACCEPT_RA to no from cloudinit, for any ipv6 static configuration. I see this set to yes by default, env: RHEL8, cloudinit-19.4, power hardware. | 08:47 |
vijayendra | we wanted to set IPV6_AUTOCONF=no in ipv6 interface in ifcfg-* files | 08:49 |
vijayendra | above is for static configuration of ipv6 | 09:18 |
meena | final failing tests fixed, now it's time to start removing more duplicates and wrappers of generate_fallback_config()! | 12:55 |
amansi26 | smoser: Hi it will be great help if you can review the changeset https://github.com/canonical/cloud-init/pull/584/files | 14:42 |
rharper | vijayendra: what does your network-config look like? reading cloudinit/net/sysconfig.py: IPV6_AUTOCONF defaults to no; except if the subnet is 'ipv6_slaac' or 'ipv6_dhcpv6-stateless'; the 'accept-ra' network config key translates to IPV6_FORCE_ACCEPT_RA on rhel; | 15:04 |
rharper | vijayendra: I suggest you work with cloud-init devel net-convert ; you can provide input config and it will render distro specific network config output to a directory so you can adjust the inputs to get your desired output | 15:05 |
blackboxsw | paride: minor merge conflict that will need sorting on https://github.com/canonical/cloud-init/pull/576 | 15:33 |
blackboxsw | paride: and since Dan is about to land another PR for Azure, might want to rebase off of tip of master once https://github.com/canonical/cloud-init/pull/590 lands | 15:34 |
blackboxsw | Odd_Bloke: is https://github.com/canonical/cloud-init/pull/590 ready? | 15:34 |
paride | blackboxsw, +1 | 15:34 |
paride | I'll rebase after #590 lands | 15:34 |
blackboxsw | CI passed on #590 now, and in approved state | 15:34 |
blackboxsw | BTW paride first time I've seen const used on argparse. Thanks for that | 15:35 |
paride | not that I knew it before using it there :P | 15:36 |
Odd_Bloke | Yep, #590 is the PR I referenced in stand-up as just waiting on CI; I've just merged it. | 15:54 |
Odd_Bloke | paride: ^ #590 has landed. | 15:55 |
paride | ty Odd_Bloke | 15:56 |
meena | what's const in python / argparse? | 16:49 |
vijayendra | rharper, Looks like cloudinit did not any entry related to IPV6_AUTOCONF and IPV6FORWARDING | 17:11 |
vijayendra | when no entry added IPV6_AUTOCONF seems to be yes and IPV6FORWARDING is no | 17:11 |
vijayendra | If IPV6FORWARDING=yes, then IPV6_AUTOCONF will default to no. | 17:11 |
vijayendra | If IPV6FORWARDING=no, then IPV6_AUTOCONF will default to yes | 17:11 |
vijayendra | configdrive metadata_json: https://paste.ubuntu.com/p/8HkQtzWFyh/ | 17:11 |
vijayendra | cloud.cfg: https://paste.ubuntu.com/p/KbNGjqHYYq/ | 17:11 |
vijayendra | ifcfg-env3: https://paste.ubuntu.com/p/CZvP3yGBGH/ | 17:11 |
vijayendra | rharper, ignore bootcmd in the cloud.cfg at the end | 17:13 |
vijayendra | was added that because I couldn't figure out other way | 17:13 |
rharper | vijayendra: I don't know if there is a bug or if you want a different configuration; but given what we have from openstack in your network config; it's not clear to me how cloud-init can make a choice about the variables you want set or unset; | 17:54 |
vijayendra | rharper, I did check code here cloudinit/net/sysconfig.py, it doesn't seems to have any key I can specify in network_data.json for the same | 17:56 |
rharper | the flow is network_data.json (parsed by cloudinit/sources/helpers/openstack.py:convert_net_json() into a network-v1 config -> (internal net state) -> sysconfig.py renders; now; looking at network_data.json input ... what in that JSON indicates to cloud-init that it should configure IPV6 the way you want it to be configured? | 17:58 |
rharper | I don't believe network_data.json from openstack is *specific* enough to let cloud-init know that IPV6 should enable/disable RAs for static addresses; | 17:58 |
rharper | or anything at all about forwarding | 17:58 |
rharper | so, what can cloud-init do here? | 17:59 |
vijayendra | rharper, True, I did work with few changes in network_data.json like specifying accept_ra: no and mode: link-local etc... no success , I failed to construct a right network_data.json here to tell cloudinit to configure IPV6_AUTOCONF to no and IPV6_FORCE_ACCEPT_RA to no | 18:02 |
rharper | are you creating a custom config-drive payload? or is openstack generating this for you ? | 18:03 |
vijayendra | openstack generated one(with minor changes to it), I am working with cloud-init devel convert as shown here https://paste.ubuntu.com/p/Tb5vk6SgT4/ | 18:07 |
vijayendra | rharper, currently trying to construct network_data.json to so that I can have sysconfig something like shown here https://paste.ubuntu.com/p/kPVP8JrZqH/ | 18:10 |
rharper | do none of the typical modes work? it seems like you'd want one of the DHCP modes; otherwise if you disable RA; how will you get next neighbor hop ? you have no gateway provided, no ? | 18:13 |
rharper | typically we see, SLAAC, or DHCP6 (stateful or stateless); and if you;re purely static I'd think you'd want to accept router advertisements for getting packets to/from the instance; ? | 18:14 |
vijayendra | currently no ipv6 router, ipv6 interface is used for local lan communication b/w vms | 18:15 |
vijayendra | its a static configuration | 18:15 |
rharper | ok, but what's wrong with enabling autoconf ? | 18:16 |
rharper | if you have no routers doing advertisements, then it just sends out the request every so often; could be disabled separately from cloud-init (or in a runcmd ) | 18:16 |
vijayendra | when you enable autoconf ra is enabled too. meaning without ipv6 router, ipv6 interface has few issues on RHEL | 18:17 |
rharper | right, I know it enables RA; what issues? | 18:18 |
rharper | it sends neighbor requests every so often and backs off ; you can also disable RA via sysctl commands | 18:18 |
vijayendra | ipv6 interface ip address fluctuates every now and then | 18:19 |
rharper | that doesn't sound right; you have a static ip; it does neighbor discovery to determine if the address is local or global ... but I wouldn't think the address actually changes; that sounds like a serious bug | 18:19 |
rharper | in any case; network_data.json doesn't have a "ipv6_static_no_ra" network type yet; so I don;t think cloud-init is ever going to render a config the way you want; | 18:20 |
vijayendra | yes true. Let me share you ip a output for you which we were monitoring using some script | 18:22 |
rharper | fixing up the rendered config later in boot with runcmd program might not help if you networking is unreliable until you can disable RA; maybe bake a sysctl setting into the image? | 18:22 |
rharper | vijayendra: I don't need to see it; it sounds like a RHEL bug; but maybe it is expected for ipv6 in some scenarios; I know a bit but not a ton on ipv6; | 18:22 |
vijayendra | rharper, I did try sysctl settings like below but no success | 18:25 |
vijayendra | net.ipv6.conf.all.autoconf=0 | 18:25 |
vijayendra | net.ipv6.conf.all.accept_ra=0 | 18:25 |
rharper | sounds like a bug or the issue isn't related to ra; or that doesn't turn off RA the way its needed; | 18:29 |
vijayendra | rharper, ok. but not sure of any other way right now. if you check nmcli -o connection show System\ env5 ---> ipv6.method auto even though we have static ip address. this has to be manual or link-local | 18:30 |
rharper | we don't configure nm directly | 18:30 |
rharper | so that could be an nm bug as it's parsing sysconfig second hand | 18:31 |
aswinr | Hi. Looking to get this into the October SRU. Has been reviewed by AnhVoMSFT and signed off. Would appreciate review and help in getting this merged. https://github.com/canonical/cloud-init/pull/613 | 20:41 |
meena | we have another SRU coming up?? | 21:18 |
AnhVoMSFT | I believe there's a tentative plan for an SRU end of October | 21:26 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!