mwhudson | bigjools: hello yet again | 02:50 |
---|---|---|
mwhudson | bigjools: do you know why juju might thing that maas nodes have a public-address like "calxeda01-18-00.localdomain"? | 02:51 |
mwhudson | the .localdomain doesn't work in our set up | 02:51 |
bigjools | mwhudson: because you probably left the default setting on the cluster name | 02:52 |
bigjools | you have to de-enlist everything, rename the cluster and start again | 02:52 |
mwhudson | oh yay! | 02:53 |
mwhudson | where do i rename it? | 02:53 |
bigjools | you might get away with just releasing nodes | 02:53 |
bigjools | but you'd have to edit all the nodes' hostnames too | 02:53 |
bigjools | edit the cluster, there's a domain or name or something | 02:53 |
mwhudson | the fqdn in the nodes' pages is what i want, no localdomain there | 02:54 |
bigjools | should be blindingly obvious, I hope | 02:54 |
bigjools | you need to edit the nodes then | 02:54 |
mwhudson | and there is "Default domain for new nodes" on /settings/ but that's set to "local" | 02:54 |
bigjools | right | 02:54 |
bigjools | so just destroy-env and rename the nodes | 02:54 |
mwhudson | afaics the node names are already correct though | 02:55 |
bigjools | no need to add a domain IIRC | 02:55 |
mwhudson | i don't understand what i would rename them too | 02:55 |
bigjools | if I recall correctly (sorry this has always just worked for me), the nodes should have a name without a dot | 02:55 |
bigjools | then it uses the cluster's name as a domain | 02:55 |
mwhudson | hmm | 02:56 |
mwhudson | this doesn't appear to be what is happening | 02:57 |
* mwhudson rummages through the source | 02:59 | |
mwhudson | hah, the cluster name is "Cluster 74184da7-7791-4a3b-a057-95d6a5b4bd60" | 03:00 |
mwhudson | i don't think that's making an appearance | 03:00 |
mwhudson | bigjools: is there a way of running a 'manage.py shell' sort of thing for maas? | 03:05 |
bigjools | mwhudson: "maas shell" | 03:05 |
bigjools | I'm interested in why it's gone a bit Pete Tong for you, since this stuff really does just work for me with no messing | 03:06 |
mwhudson | >>> n.fqdn | 03:06 |
mwhudson | u'calxeda01-17-00' | 03:06 |
mwhudson | that's correct | 03:06 |
mwhudson | can i see the serialized form of this node that's being sent to juju somehow? | 03:07 |
bigjools | I think one of the maas-cli ops will return it | 03:09 |
bigjools | there's a query or something | 03:09 |
mwhudson | ah yeah, nodes list prints stuff | 03:10 |
mwhudson | and it looks right there too | 03:10 |
mwhudson | "hostname": "calxeda01-21-00", | 03:10 |
mwhudson | so i kinda think it's juju's fault maybe | 03:11 |
bigjools | do you not get a domain name in there then? | 03:13 |
mwhudson | no | 03:13 |
bigjools | a domain of .local is Avahi | 03:13 |
bigjools | I always make sure it's something else | 03:14 |
mwhudson | this makes no effing sense :( | 03:14 |
bigjools | heh | 03:14 |
bigjools | I suggest destroying your env and starting again | 03:15 |
bigjools | and keep an eye on what maas has vs what juju thinks is going on | 03:15 |
mwhudson | it's pretty clear that the machine in juju is constructed with both dns_name and private_dns_name set to the 'hostname' from the serialization of the node | 03:15 |
mwhudson | (this is pyju, fwiw) | 03:15 |
bigjools | ok | 03:15 |
bigjools | is "localdomain" mentioned *anywhere* ? | 03:16 |
mwhudson | it's not in the juju source | 03:16 |
mwhudson | or the maas source | 03:16 |
mwhudson | it's possible our dns is stupidly misconfigured somehow i guess | 03:17 |
bigjools | mmmm | 03:17 |
bigjools | does juju status show the right thing | 03:17 |
bigjools | ? | 03:17 |
mwhudson | no | 03:17 |
mwhudson | oh wait | 03:18 |
mwhudson | the dns names for the machines are right | 03:18 |
mwhudson | but the public adresses for the units are not | 03:18 |
mwhudson | ARGH | 03:19 |
mwhudson | ubuntu@juju:~$ ssh calxeda01-17-00 hostname -f | 03:19 |
mwhudson | calxeda01-17-00.localdomain | 03:19 |
bigjools | ! | 03:19 |
mwhudson | ARGH | 03:19 |
mwhudson | so where does that come from? | 03:20 |
bigjools | cloudinit | 03:20 |
bigjools | it could be an artifact of not setting an explicit domain of your own | 03:20 |
mwhudson | # Added by cloud-init | 03:21 |
mwhudson | 127.0.1.1 calxeda01-17-00.localdomain calxeda01-17-00 | 03:21 |
mwhudson | (from /etc/hosts on the node) | 03:21 |
bigjools | yeah | 03:21 |
mwhudson | ok | 03:29 |
mwhudson | so who makes the cloud init config? | 03:29 |
mwhudson | maas? | 03:29 |
bigjools | mwhudson: no, juju | 03:29 |
mwhudson | ok | 03:29 |
bigjools | well let me clarify | 03:29 |
bigjools | juju makes user-data | 03:29 |
bigjools | the cloud-init config itself is fairly static, there's a template with it in | 03:30 |
mwhudson | it seems cloud-init really wants to have a fqdn with a dot in it | 03:35 |
bigjools | yeah | 03:35 |
bigjools | if you want maas to manage the domain you need to set something | 03:36 |
mwhudson | i wonder if i can man dnsmasq resolve .localdomain | 03:37 |
mwhudson | *make | 03:37 |
mwhudson | haha, i wonder how relevant this is http://askubuntu.com/questions/281628/juju-dnsmasq-and-localdomain | 03:38 |
mwhudson | "I added domain=localdomain to dnsmasq.conf file which sorts the problem out. | 03:38 |
mwhudson | " | 03:38 |
mwhudson | pff | 03:41 |
bigjools | ha | 03:41 |
bigjools | mwhudson: btw your tweet "like a boss"... really? :) | 03:41 |
mwhudson | heh | 03:41 |
mwhudson | oh look, creche pick up time! | 03:41 |
mwhudson | bigjools: thanks for the help | 03:43 |
bigjools | mwhudson: no worries, hope it works eventually | 03:43 |
=== kentb-out is now known as kentb | ||
=== matsubara is now known as matsubara-lunch | ||
=== matsubara-lunch is now known as matsubara | ||
=== chuck_ is now known as zul | ||
=== robbiew is now known as robbiew-afk | ||
=== kentb is now known as kentb-ou | ||
=== kentb-ou is now known as kentb-out |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!