rharperrobjo: https://bugs.launchpad.net/suse/+source/cloud-init/+bug/1843502  ;   curious bug here around azure + suse image and network-config parsing15:19
ubot5Launchpad bug 1843502 in cloud-init "Network config is incorrectly parsed when nameservers are specified" [Undecided,Incomplete]15:19
robjorharper: I worked with Moustafa a little bit on this, as in, help to find one place in the code why resolv.conf doesn't have the proper value. Looks like he made much more progress since15:26
robjobased on the description this appears to be a generic issue with the "v1" syntax and a typo, "address" vs. "addresses"15:27
rharperazure renders v2, so I mentioned the 'nameservers' bits need to be indented underneath the interface name (eth0)15:28
robjoAlso added a comment15:33
rharperrobjo: the other confusing  part (to me) was where the nameserver values came from;  it's been a while since I've looked at the contents of azure IMDS15:34
rharperrobjo: thanks,  I suspect if there's no downstream changes, then it's got to be local image modifications15:35
rharperthere are some odd debug messages in the log as well,15:36
rharper2019-09-09 16:21:29,416 - handlers.py[DEBUG]: start: azure-ds/parse_network_config:15:36
rharper2019-09-09 16:21:29,416 - DataSourceAzure.py[DEBUG]: Azure: generating network configuration from IMDS15:36
rharper2019-09-09 16:21:29,416 - DataSourceAzure.py[DEBUG]: netconfig &s:15:36
rharperin which cloudinit has never emitted the string "netconfig &s"15:36
rharperso I'm thinking to close this out as invalid since someone is hacking their own changes into cloud-init15:36
robjoI just don't know where the nameserver information is supposed to be coming from15:36
robjoif Azure puts the info into the metadata then the data source should pick it up and make it availabe15:37
rharperso, Azure, AFAIK, only provides dns info via dhcp lease15:37
robjoif that is the case then how could network configuration via cloud-init ever work?15:38
rharperso we never explicitly generate network config with 'nameservers' values; azure is a dhcp on eth0 shop, you can dhcp on additional interfaces, and add static ips; but beyond that, I've never seen any dns info in IMDS data15:38
rharperwe generate dhcp on eth015:38
rharperand the OS's dhcp client does the rest15:38
robjoWell something triggers that resolv.conf gets written by cloud-init which prevents netconfig from writing the dns info from the dhcp server15:39
robjoThere was a bug where an empty resolv.conf was written tha triggered that condition but that has since been fixed15:40
robjoOK, so now I am even more confused15:40
smoserif anyone intereseted15:40
robjoI guess I have to fiddle with this on my own and see what happens15:41
rharperrobjo: do you have a recent boot of an unmodified sles image on azure that we can compare cloud-init.logs ?  I really suspect there's some local modifications due to that strange log output15:42
robjoCurrently I am not building images with cloud-init, it would be a matter of staring a SLES instance in Azure, then adding cloud init and running cloud-init local to see what happens with network configuration15:43
rharperoh interesting, so it's almost certainly a  custom image then15:44
robjoGiven the state a reboot to get a full log may not be advisable as one may not get back into the system15:44
robjoIt's certainly customized w.r.t. the initialization code, I am not certain Moustafa made any changes to the cloud-init package delivered in the SUSE repos15:46
rharperI'll ask explicitly,  would rpm -verify cloud-init tell us if anything has changed?15:47
robjoyes that would be expected15:54
smoserrharper: ^16:56
rharpersmoser: replied17:26
smoserrharper: you can approve https://code.launchpad.net/~smoser/cloud-init/+git/cloud-init/+merge/372622 if you approve my dropping of the bug url20:08
smoserthat was your only question) and Odd_Bloke already approved.20:08
rharpersmoser: sure;20:27
rharpersmoser: hrm, I was hoping for a new url to a bug indicating that we needed to narrow the definition , ie, the reason for the bug ;  was that unclear ? or do you feel it's unnecessary ?20:29
rharperthat is, if someone else were to add another look-a-like like zstack and brightbox, the comment could indicate that we want to make sure it doesn't widely match other domain names20:30
smoseri think its unnecessary.20:33
smosergit log / blame will indicate the bug20:33
smoserand (for zstack) there will be a doc/datasources/zstack20:33
smosercode does not need references like that.20:34
smoserwhen i went looking, i did a git log, didn't even see the in-file comment.20:34
blackboxswrharper: Goneri I've updated my review comments on https://code.launchpad.net/~goneri/cloud-init/+git/cloud-init/+merge/36850721:26
blackboxswone minor behavioral thing I'd like to have fixed. but otherwise looks good to me21:26
blackboxswbtw I'm also going to add Goneri to our CI group given the number of branches proposed (it'll automate CI runs on any new branch pushes)21:32
blackboxswwelcome to the CI team Goneri :)21:32
blackboxswyou'll get CI votes on your branches from here on out21:33
rharperblackboxsw: +121:37
Goneriblackboxsw, thanks for the review and the group inclusion :-)22:00

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!