[00:40] <negronjl> This officially sucks ..
[00:40] <negronjl> now that I have the MaaS WoL thing more or less worked out, juju is not working...
[00:40] <negronjl> juju bootstrap turns the machine up but, there is no IP address to that machine ... and I can't log in to the node either.
[00:41] <negronjl> Does anyone know which kickstart is used to load the machine up when it is in the "Ready" state ?  I need to add a password to the ubuntu user there.
[00:47] <negronjl> I need to take a break from this for a few minutes ... I'll be back in a few
[01:06] <bigjools> negronjl: don't do that, juju gives the ubuntu user your public ssh key
[01:06] <negronjl> bigjools: don't do which part ... I've done so many things that I am going crazy and losing track.
 Does anyone know which kickstart is used to load the machine up when it is in the "Ready" state ?  I need to add a password to the ubuntu user there.
[01:09] <negronjl> bigjools: I am having issues with juju now ( got past the WoL issues )
[01:10] <negronjl> bigjools: after the nodes are "Ready" I can't get juju bootstrap to finish
[01:10] <negronjl> bigjools: I am able to do "juju bootstrap" or "juju bootstrap --constraints maas-name=quantal"
[01:10] <bigjools> negronjl: how do you know it's not finished?
[01:11] <negronjl> bigjools: I have a monitor hooked up to the node
[01:11] <negronjl> bigjools: it WoL the system, installs everything that needs to be installed ( system, juju, etc. )
[01:13] <negronjl> bigjools: after i end up @ the login promt, I try "juju status" but, I get "no route to host"
[01:13] <negronjl> bigjools: i can't even ping the node.
[01:15] <negronjl> bigjools: I was wondering how to get inside of the node ( the juju bootstrap node ) to see if I can figure out what is happening
[01:17] <bigjools> are you relying on zeroconf for name lookup, or dns on the maas server?
[01:19] <negronjl> dns on maas server
[01:19] <bigjools> and is it correct?
[01:19] <negronjl> it is ... when I "host <insert node name>" it replies correctly
[01:20] <bigjools> can you ssh ubuntu@host?
[01:20] <negronjl> no ... "no route to <nodename>"
[01:20] <negronjl> it doesn't seem to be a DNS issue as I cannot get to the node by address
[01:21] <negronjl> I am working on 192.168.1.0/24 and I have checked every IP ...  the node doesn't appear to have an IP
[01:22] <bigjools> it doesn't sound like a juju or maas problem
[01:22] <bigjools> did the dhcpd assign an ip?
[01:22] <negronjl> bigjools: according to the DHCP ... no
[01:24] <negronjl> bigjools: The node doesn't appear to be asking for one
[01:27] <negronjl> bigjools: That's why I was asking for a way to get in the node so I can ( with the monitor/keyboard ) check things out.
[01:30] <lifeless> negronjl: are you running precise? is that what was used to do the demo ?
[01:30] <negronjl> lifeless: yes and yes
[01:30] <lifeless> ok cool
[01:30] <lifeless> uhm
[01:31] <lifeless> bigjools: ip allocation will be dhcp on each node right ?
[01:31] <lifeless> negronjl: have you got tcpdump / wireshark running on the dhcp server, see what comes in ?
[01:32] <negronjl> lifeless: yes but, when I do "juju bootstrap"  I don't see the node asking for an address
[01:32] <lifeless> ok, even though it powers up ?
[01:34] <negronjl> lifeless:  yes .. the node powers up and it appears to install correctly ... it then reboots and it leaves me at the login prompt.
[01:34] <lifeless> during the install, it must do networking - do you see dhcp requests during that period ?
[01:34] <negronjl> I then do "juju status" and I get a "no route to host"
[01:35] <lifeless> If you don't, it suggests that the wireshark config is wrong, if you do, but don't later, it tells us either that a) dhcp is being more persistent than usual :P[unlikely] or b) dhcp isn't trying on the reboot.
[01:36] <lifeless> negronjl: you could try using arp to locally associate the address used during the install with the node, and see if it responds to that via ping/ssh
[01:36] <negronjl> ahh ... I could try that lifeless
[01:36] <negronjl> lifeless:  give me a minute to try it out
[01:36] <lifeless> sure
[01:36] <lifeless> also check you see DHCP for the install phase.
[01:37] <negronjl> lifeless: i will
[01:39] <bigjools> I have had it up to HERE with ubuntu problems already today
[01:40] <lifeless> welcome back
[01:40] <negronjl> bigjools: amen to that :)
[01:40] <bigjools> three reboots
[01:40] <bigjools> and it's 11:40am
[01:54] <negronjl> just so we are all on the same page .. I am monitoring DHCP traffic at the maas server with tcpdump -i eth0 -n port 67 and port 68
[01:55] <negronjl> I just did a "juju bootstrap"  ( leaving maas to pick a node for me )
[01:55] <negronjl> I see the DHCP requests from the node and the responses from the DHCP server
[01:58] <negronjl> hah .. i found something
[01:59] <negronjl> i found that the IP that the node got differs from what's expected.
[02:00] <negronjl> ie:  the node's name is natty so, when i do "host natty"  I get "192.168.1.103" but, when I look at the tcpdump output, I see that the address is 192.168.1.168
[02:00] <negronjl> I cannot ping 192.168.1.103 but, I CAN ping 192.168.1.168
[02:01] <negronjl> I CANNOT ssh into 192.168.1.168 though ... it asks me for password but, I don't have one
[02:01] <bigjools> is cobbler set to manage DNS and DHCP?
[02:02] <bigjools> it is supposed to set the same IP in both
[02:02] <bigjools> if it doesn't then things will break like this
[02:02] <negronjl> cobbler is managing DNS but, not DHCP
[02:02] <bigjools> there's your problem
[02:02] <negronjl> there's one of my problems ...
[02:02] <bigjools> :)
[02:02] <lifeless> optimist :P
[02:02] <bigjools> ha
[02:02] <negronjl> lol
[02:02] <lifeless> why isn't cobbler managing DHCP ?
[02:02] <negronjl> this is the way the setup was sent to me
[02:03] <negronjl> but, even if I change that ... i still have the problem that I cannot log int
[02:03] <negronjl> s/int/in
[02:03] <lifeless> well
[02:03] <lifeless> so cloud-init may not have run
[02:03] <lifeless> because its not got the right ip now
[02:03] <negronjl> ok ... let's try to fix the DNS/DHCP issue first i guess....
[02:04] <lifeless> definitely
[02:04] <negronjl> I have a dd-wrt router in the mix acting as DHCP ... let me pull that out of the equation...then I will need a bit of help fixing the DHCP in the maas server ...
[02:04] <negronjl> brb ...breaking things :)
[02:04] <bigjools> /etc/cobbler/settings
[02:05] <bigjools> manage_dhcp: 1
[02:05] <lifeless> negronjl: the whole thing should be its own standalone network, with (I believe) a local archive mirror and everything
[02:05] <negronjl> bigjools:  checking /etc/cobbler/settings now
[02:05] <negronjl> lifeless: i have the local mirror thing working.
[02:05] <lifeless> negronjl: or at worst, with the maas server itself attached on one side to the maas cluster LAN,a nd on the other to your network.
[02:06] <lifeless> again, AIUI, second hand knowledge ;)
[02:06] <negronjl> oh wow ... according to /etc/cobbler/settings .... the maas server is managing both dns and dhcp ( manage_dhcp: 1 manage_dns: 1 )
[02:06] <negronjl> this doesn't seem right ....
[02:09] <bigjools> check that your node has got an entry in the dhcpd config
[02:09] <bigjools> is it using dnsmasq or isc?
[02:10] <bigjools> it should map a mac to an IP
[02:15] <negronjl> bigjools: ok .. let me check ...
[02:15] <negronjl> let's start with this ... the maas server should have a static ip address
[02:15] <negronjl> right now it doesn't
[02:15] <negronjl> it is assumes that it will be 192.168.1.2 ( I am working with a 192.168.1.0/24 network )
[02:16] <negronjl> I should probably fix the maas server first
[02:16] <lifeless> yup
[02:16] <lifeless> if its dual attached, it can be dhcp on the external side
[02:17] <lifeless> but the cluster side has to be pretty fixed ;)
[02:18] <negronjl> lifeless: the maas server is a laptop with a broken screen :/
[02:20] <negronjl> and one interface only
[02:20] <lifeless> no wifi? or wifi only ?
[02:21] <lifeless> negronjl: it arrived broken?
[02:22] <negronjl> lifeless: it did although I have a vague recollection of a conversation with mrussell ( Mark Russell ) that it may have been broken before shipment to my house.
[02:22] <negronjl> maas server has a static ip now
[02:22] <negronjl> now checking dhcpd ( /etc/dhcp3 ) configuration
[02:23] <lifeless> cry for help sent
[02:24] <negronjl> bigjools: in MaaS ... what handles the DHCP server ( I see maas-dhcp package installed but, don't really know where the config resides )
[02:24] <lifeless> its lacking all technical info about where we are at; wrong audience for that I think.
[02:24] <negronjl> lifeless: thx ... well said .. cry for help
[02:24] <bigjools> negronjl: cobbler does everything
[02:24] <bigjools> maas is just driving it
[02:25] <negronjl> bigjools: ok .. so earlier you had mentioned for me to make sure that I had something in dhcp config ... where should I be looking ?
[02:25] <bigjools> dhcp config is written by cobbler
[02:25] <bigjools> so we need to see if it wrote it right
[02:25] <bigjools> so, just looking for that MAC to IP mapping so the node gets the IP address we expect
[02:28] <negronjl> bigjools: I'll start a node to see how it behaves now
[02:28] <bigjools> ok
[02:30] <negronjl> hah ... i have to laugh because I don't want to cry ... the maas server just froze ... rebooting it
[02:30] <bigjools> you are seriously unlucky
[02:34] <negronjl> bigjools: what handles dhcp and dns here ...
[02:34] <bigjools> negronjl: it's in settings
[02:35] <negronjl> I'm really not trying to ask stupid questions ... I am just trying not to assume anything at this point
[02:35] <bigjools> so you can tell cobbler to drive isc dhcp or dnsmasq
[02:35] <bigjools> it's dnsmasq out the box IIRC
[02:36] <negronjl> it appears to be dnsmasq
[02:38] <negronjl> bigjools: where is the config for dnsmasq ?
[02:39] <bigjools> /etc/dnsmasq.conf?
[02:41] <negronjl> i may have to rebuild this environment .. ( maas-flush and all ).  I think that having two dns/dhcp servers in the network ( cobbler and the router ) may have screwed the config on the nodes
[02:41] <bigjools> very likely
[02:44] <negronjl> rebuilding now ... it'll take a few minutes
[02:54] <negronjl> bigjools: now pxe booting is not working .... TFTP open timeout
[02:54] <bigjools> yay
[02:54] <negronjl> bigjools: i have cobbler using the system's tftp ... should I change it so cobbler uses it's own instead ?
[02:55] <negronjl> system's tfpt => tfptd-hpa ( /etc/init.d/tftpd-hpa )
[02:55] <bigjools> that should be ok
[02:55] <negronjl> the system's ?  it was working before so it seems to be a config issue somewhere in cobbler
[02:55] <bigjools> you must have been use pxe on the router before
[02:56] <negronjl> apparently
[02:56] <bigjools> using*
[02:56] <bigjools> I don't know what's best here, try flipping and see
[02:56] <negronjl> ok
[03:02] <negronjl> nope ... I think I was getting further with the router ...
[03:02] <negronjl> maybe try having the router pass the pxe and dhcp and have the maas server handle dns
[03:03] <negronjl> bigjools: ^^
[03:03] <bigjools> I don't know what's best, sorry
[03:04] <bigjools> you need roaksoax perhaps
[03:04] <bigjools> if he is up
[03:05] <lifeless> negronjl: its what, 8pm for you ?
[03:05] <negronjl> lifeless: it is
[03:06] <lifeless> negronjl: ok; I'm going to break for a late lunch in a minute
[03:06] <lifeless> negronjl: whats the network layout you have in use?
[03:06] <negronjl> lifeless: flat network ( 192.168.1.0/24 )
[03:06] <lifeless> negronjl: e.g. nodes + maas server on one switch, where does your router come into it, and so on.
[03:07] <negronjl> lifeless:  router is 192.168.1.1 ( gw and dhcp )
[03:07] <negronjl> lifeless: 192.168.1.2 -> maas-server
[03:07] <lifeless> negronjl: you haven't turned dhcp off ?
[03:07] <lifeless> on the router that is
[03:07] <bigjools> are all the necessary daemons running on the maas box?
[03:07] <negronjl> lifeless:  i did turn dhcp off on the router but, pxe stopped working
[03:07] <negronjl> bigjools: they are
[03:07] <lifeless> negronjl: ok, leave it off please :)
[03:08] <lifeless> negronjl:  you won't have your router at Structure :>
[03:08] <bigjools> dnsmasq, dhcpd, bind, tftpd
[03:08] <bigjools> and tgtd
[03:08] <negronjl> lifeless:  the router was shipped here with the rest of the stuff so, yes .. i could
[03:08] <negronjl> bigjools: no dhcpd nor bind ... dnmasq instead
[03:08] <lifeless> negronjl: oh, it was? See, bad assumptions R us :)
[03:08] <negronjl> bigjools: tftpd-hpa for tftp
[03:09] <bigjools> ok
[03:09] <bigjools> so we need to work out why the node cannot see a pxe server (ie dhcpd)
[03:09] <bigjools> firewalled?
[03:09] <negronjl> bigjools: straight cable
[03:09] <lifeless> negronjl: are the nodes plugged into the router, or a switch/hub ?
[03:09] <negronjl> lifeless: switch
[03:09] <negronjl> here is the setup
[03:10] <negronjl> one switch where all of the nodes and maas-server are connected
[03:10] <negronjl> maas server is a laptop running the above mentioned daemons
[03:10] <negronjl> that is all
[03:10] <lifeless> the router is plugged into the switch?
[03:10] <negronjl> 9 nodes ( HP microservers ) one switch and one laptop.
[03:10] <bigjools> can you get a pxe test client?
[03:11] <negronjl> when I was using the router ... it was also plugged in the switch to provide dhcp/dns
[03:11] <lifeless> ok, cool.
[03:11] <lifeless> so, default switch environment is to forward everything, pxe should work fine there.
[03:11] <lifeless> negronjl: got wireshark up? can you get the DCHP OFFER contents, that should have the pxe metadata in it.
[03:11] <negronjl> lifeless: pxe was working fine with the router but, I believe that the nodes were getting their IPs from the DHCP but not updating the DNS.
[03:12] <negronjl> I believe the DHCP was on the router but, DNS was on router AND maas server ...
[03:12] <negronjl> not sure but, I really believe that there was a conflict between the router's dns/dhcp and the maas-server's dns/dhcp
[03:13] <lifeless> sure, so we now have dhcp and dns both set to managed in cobbler
[03:13] <lifeless> and wol is firing up the node
[03:13] <lifeless> and its failing to PXE boot ?
[03:13] <rick1> execuse me, does anyone know the MaaSslave's default username and password ?
[03:14] <lifeless> rick1: your ssh key should be copied into the ubuntu user for you by cloud-init.
[03:14] <negronjl> lifeless:  with all the wires running across my dining room table, I may have found something ... let me try again .. give me just a second
[03:14] <lifeless> rick1: If I remember correctly.
[03:14] <rick1> lifeless: so, I can just login to it?
[03:14] <lifeless> ssh yes
[03:15] <lifeless> it won't have a password at all
[03:15] <rick1> lifeless: Thanks, I'll try
[03:15] <negronjl> lifeless:  pxe booting is now working off of the maas server ( no router involved )
[03:15] <negronjl> I am going to continue deploying to see if things are better now
[03:15] <lifeless> negronjl: what was wrong?
[03:16] <roaksoax> negronjl: that machine has been broken for ages and that's the same one used at the ODS demo
[03:16] <bigjools> rick1: the user is "ubuntu"
[03:16] <negronjl> network loop
[03:16] <rick1> bigjools: oh, Okay
[03:16] <roaksoax> negronjl: use the dd-wrt router for DNS/DHCP
[03:16] <roaksoax> negronjl: don't use maas
[03:17] <lifeless> roaksoax: ohhai!; I'll let you steer :)
[03:17] <negronjl> roaksoax: i see now that by using the maas-server for dns/dhcp i can't spoof the archives ...
[03:17] <negronjl> roaksoax: I'll switch it back ... give me just a second
[03:17] <negronjl> roaksoax: do I need to turn dns/dhcp off in cobbler now that the router is back in the picture ??
[03:18] <roaksoax> negronjl: yeah don't use maas-server as DHCP/DNS because you'll have to manually assign what IP addresses the clients will get throught DHCP
[03:18] <roaksoax> negronjl: yes you do
[03:18] <roaksoax> negronjl: the dd-wrt should already e confogured for DNS/DHCP so you shouldn't really have any issues there
[03:18] <rick1> lifeless: It shows Connection reset by peer, It must be something wrong with the PXE installation.... always shows commissioning on the WebUI
[03:19] <negronjl> roaksoax: just to be sure .... manage_tftpd and manage_dns should both be set to 0 ( and not 1 ) on /etc/cobbler/settings ??
[03:19] <roaksoax> negronjl: manage_dns manage_dhcp
[03:19] <roaksoax> negronjl: we *do* manage_tfptd
[03:19] <negronjl> roaksoax: sorry yes .. those two
[03:20] <negronjl> roaksoax: manage_dns and manage_dhcp
[03:20] <negronjl> roaksoax: they are both set to 1 on /etc/cobbler/settings
[03:21] <roaksoax> negronjl: I can't recall (i'm zombie already), but you should be able to dpkg-reconfigure maas-dhcp and disable DNS/DHCP
[03:21] <lifeless> roaksoax: negronjl: I'm going to take that lunch break. Will check in in ~40
[03:21] <negronjl> lifeless: thx
[03:21] <negronjl> roaksoax: I'll do that ..
[03:21] <roaksoax> lifeless: provecho :)
[03:22] <negronjl> roaksoax: did dpkg-reconfigure but, I don't get the option to disable it.
[03:26] <negronjl> roaksoax: trying again ( i disabled it directly in /etc/cobbler/settings ) at this point, i am just trying things out
[03:26]  * roaksoax checks
[03:26] <roaksoax> negronjl: try purging maas-dhcp
[03:27] <negronjl> roaksoax: will do
[03:54] <negronjl> roaksoax: not sure if you are still around but, I have a question....
[03:54] <roaksoax> negronjl: shoot
[03:54] <negronjl> roaksoax: After the node is in "Ready" state ... can I then run "juju bootstrap" ?
[03:55] <negronjl> roaksoax: or is there anything that I have to do between Ready and juju bootstrap
[03:55] <roaksoax> negronjl: ready state are ready to be used
[03:55] <negronjl> roaksoax: so, juju bootstrap would be ok at that point ?
[03:56] <roaksoax> negronjl: yes,
[03:57] <negronjl> roaksoax: thx ...  I think I should be good for now ... thank you VERY much for all you patience and help during this.
[03:57] <roaksoax> no worries
[03:57] <roaksoax> negronjl: we've been there
[03:57] <roaksoax> :)
[04:04] <lifeless> hows it looking guys?
[04:09] <rick1> Still stuck on "Commissioning" state
[04:09] <lifeless> rick1: I don't know the states all *that* well, but I think that means 'trying to turn it on and install it'
[04:10] <rick1> lifeless: yes
[04:12] <lifeless> rick1: are you using wake on lan or a power board or something? or just reaching down and pushing the power button ?
[04:12] <rick1> after maas slave PXE booting, scrolling of many screen of messages, stay on "cloud-init-nonet killed by TERM signal" for a while, then jump to login screen
[04:13] <rick1> lifeless: yes, I set to wol
[04:13] <lifeless> bigjools: / roaksoax: ^ any thoughts?
[04:13] <bigjools> no idea why that TERM happens, I think we went through all the usual suspects
[04:14] <bigjools> so I'd have to defer to roaksoax or smoser who have more knowledge of cloud-init than I
[04:16] <negronjl> I have had so many issues today that I may be able to help with this.
[04:16] <negronjl> rick1:  would you pastebin the contents of /etc/maas/maas_local_settings.py please ?
[04:17] <rick1> ok, wait a moment
[04:17] <negronjl> rick1:  I am looking for the value of: DEFAULT_MAAS_URL
[04:18] <negronjl> rick1:  In my case ( out of many ), cloud-init was not working correctly because I had an issue where DEFAULT_MAAS_URL was not properly set to the IP address ( not the hostname ) of the maas-server
[04:18] <rick1> negronjl: DEFAULT_MAAS_URL = "http://10.20.0.1/"
[04:19] <negronjl> rick1:  also make sure that the time on your nodes ( all of them ) as well as the maas-server is the same .. weird .. I know but, it helps
[04:19] <bigjools> oauth FTL
[04:23] <negronjl> bigjools:  do I have to upload my SSH key to the maas server before i do anything with it ?
[04:23] <rick1> negronjl: oh, you are right, the time is different
[04:23] <negronjl> bigjools: juju status ( after a successful juju bootstrap ) gives me Invalid SSH key
[04:23] <bigjools> negronjl: no, unless you plan on starting nodes manually in the UI
[04:23] <negronjl> rick1:  put them all on the same time
[04:24] <negronjl> bigjools: I get Invalid SSH key ....
[04:24] <bigjools> negronjl: yeah that happens if it's still booting before its run the preseed that adds the user data
[04:24] <bigjools> if the console is at the login prompt then it's not booted properly
[04:25] <bigjools> wrong image or failure to get metadata
[04:25] <negronjl> bigjools: hmmm... how do I avoid that ?  Did I do something wrong here ?
[04:26] <bigjools> negronjl: the clock problem?
[04:26] <bigjools> perhaps
[04:26] <negronjl> bigjools: nope ... all nodes are time synced
[04:26] <negronjl> bigjools: maas-server as well
[04:27] <bigjools> is it at login prompt?
[04:27] <negronjl> bigjools: yup .. .and I can't log in to figure what may be wrong ...
[04:27] <bigjools> ok is there anything in maas-server syslog from the node?
[04:29] <negronjl> Test WP failed, assume Write Enabled
[04:29] <negronjl> Asking for cache data failed
[04:29] <negronjl> Assuming drive cache: write through
[04:29] <rick1> negronjl:now time is synced, but still the same problem.
[04:29] <negronjl> Test WP failed .....
[04:30] <negronjl> rick1:  .... 10.20.0.1 is your maas server right ?
[04:30] <rick1> negronjl: yes
[04:30] <lifeless> negronjl: is juju bootstrap running for you now ?
[04:31] <negronjl> lifeless: it does but, I cannot juju status ( invalid ssh key )
[04:31] <lifeless> check your environments config is right and so forth ?
[04:31] <negronjl> rick1:  dpkg-reconfigure maas ... make sure your maas server ip is correct ...
[04:31] <negronjl> rick1: sudo cobbler sync
[04:31] <negronjl> rick1:  try again
[04:32] <bigjools> negronjl: watch the console during boot, see if anything else looks bad
[04:32] <rick1> negronjl: ok
[04:39] <rick1> negronjl: still the same. and I found a bug report about this. https://bugs.launchpad.net/ubuntu/+source/maas/+bug/992075
[04:40] <negronjl> rick1:  for me, it was a simple case of /etc/maas/maas_local_settings.py having the wrong DEFAULT_MAAS_URL
[04:40]  * bigjools adds that to the FAQ
[04:40] <rick1> negronjl: :), I'm go out and take a break, thank you VERY much
[04:40] <negronjl> rick1:  no worries .. just paying it forward ... I have been helped a LOT
[04:43] <bigjools> please consider amending or adding to the https://answers.launchpad.net/maas/+faqs
[05:06] <negronjl> bigjools, lifeless: good news:  I am now able to ssh ubuntu@<node_ip>
[05:06] <negronjl> bigjools, lifeless: bad news:  the dhcp address associated with the name does NOT match what MaaS thinks it should be
[05:07] <negronjl> bigjools, lifeless: MaaS and Juju think that my node is 192.168.1.101 but, my node got an address of 192.168.1.127 ... I have to check the DHCP server to see what may be going on but, I will have to do that tomorrow ... I am calling it a night ...
[05:17] <bigjools> negronjl: what is dhcp serving?
[05:17] <negronjl> bigjools: checking that now ... 'cause as far as I know ...the router should be the _only_ DHCP  running and it has static dhcp clients
[05:18] <bigjools> negronjl: mac address mismatch then
[05:18] <bigjools> check static config vs what maas has
[05:18] <negronjl> bigjools: checking that too but, gotta be honest ... I'm pretty tired ... I'll be checking that and then calling it a night
[05:19] <negronjl> what do you mean ... what maas has ?
[05:19] <bigjools> maas stores macs for the nodes
[05:19] <bigjools> and tells cobbler what to write
[05:19] <bigjools> so if you're now doing your own static config then it's probably different
[05:20] <negronjl> bigjools: do tell ... the router's DHCP/DNS seems to have all of the right MAC addresses ...
[05:20] <bigjools> how do you know?
[05:20] <negronjl> bigjools: can you tell me where to find what maas-server has ?
[05:20] <bigjools> look in its UI
[05:21] <negronjl> bigjools: that seems to be right as well
[05:22] <bigjools> are you sure the DHCP server on the router is assigning the IP then?
[05:25] <negronjl> bigjools: checking again
[05:26] <bigjools> also, in some dhcp setups you need to explicitly say that pxe boots can have the same IPs as regular boots
[05:39] <negronjl> done for tonight ... I'll fight some more with this tomorrow
[05:39] <negronjl> thank you very much for the help all
[05:44] <lifeless> sleep well
[05:47] <bigjools> nn
[06:03] <SpamapS> life	o/
[06:03] <SpamapS> lifeless: o/
[06:04]  * SpamapS wonders what happened to irssi there
[06:04] <lifeless> o/
[06:04] <lifeless> you hit tab and typed too fast
[06:04] <lifeless> and triggered its 'paste detection'
[06:05] <lifeless> which, FTW, bypasses all character evaluation.
[06:06] <SpamapS> ahhh
[06:07] <SpamapS> probably the bursty ssh
[06:07] <SpamapS> actually I'd swear its only bursty when I'm in tmux
[07:50]  * czajkowski hugs jtv 
[08:26] <jtv> czajkowski: all for that little email?  *blush*
[08:27] <jtv> And hi to you too :)
[08:27] <czajkowski> jtv: yes you at least replied!
[08:27] <czajkowski> simple things in life realy
[08:27] <jtv> :)
[13:19] <flacoste> roaksoax: have you seen the question from Brian G. Peterson on the maas-devel list? he's having problems with the avahi boot story
[13:19] <flacoste> is the image supposed to be working?
[13:20] <flacoste> is the error he's encountering the same than you did when you tested with matsubara?
[13:27] <roaksoax> i havent
[13:27] <roaksoax> i will address it in a bit
[13:27] <flacoste> roaksoax: thanks!
[15:05] <jtv> Daviey: where on the TFTP path do the enlistment kernels live?
[15:05] <jtv> Or do we use the same as install/commissioning?
[15:12] <Daviey> jtv: wherever makes sense to you.
[15:12] <Daviey> enlistment kenel == install kernel
[15:12] <Daviey> comissioning kernel is sperate
[15:12] <jtv> OK, I'll use the install kernel then.  Thanks.
[15:13] <jtv> Daviey: do we know yet exactly what kernel options we need to append for enlistment?
[15:15] <rvba> jtv: I think it's the options of maas-enlist in /var/lib/tftpboot/pxelinux.cfg/default
[15:15] <rvba> jtv: am I right?
[15:15] <Daviey> roaksoax: do you have those to hand?
[15:16] <jtv> rvba: not sure I understand you correctly.  Are you referring to the maas-enlist profile in Cobbler?
[15:17] <rvba> jtv: no, I'm referring to the line " append initrd=..." next to the "LABEL maas-enlist" bit in the file /var/lib/tftpboot/pxelinux.cfg/default
[15:18] <jtv> Ah, thanks.
[15:18] <jtv> Unfortunately I don't seem to have that bit there.
[15:18] <rvba> jtv: http://paste.ubuntu.com/1049268/
[15:19] <jtv> Whoa that's a lot bigger than what I've got
[15:19] <roaksoax> Daviey: the kernel options
[15:19] <roaksoax> Daviey: yeah those are appended automatically by cobbler
[15:20] <roaksoax> err
[15:20] <jtv> Weird stuff in there...  multiple locale options
[15:20] <jtv> Yeah, talking to me, I know.  :)
[15:20] <roaksoax> some of them are appended automatically
[15:20] <roaksoax> jtv: some of the ones appended in the default file come from maas-import-isos and some from the commissioning
[15:21] <jtv> From the commissioning?  But this is for a node that isn't even ready for commissioning yet.
[15:21] <roaksoax> jtv: the default ones are: append initrd=/images/precise-i386/initrd.gz ksdevice=bootif lang=  text  hostname=precise-i386 domain=local.lan suite=precise
[15:21] <roaksoax> jtv: when you import an ISO it creates the file that appends a few kernel args
[15:21] <jtv> We need a way to organize generation of this stuff I guess.
[15:21] <roaksoax> jtv: when you run maas-import-isos, it appends other required kernel args
[15:22] <roaksoax> jtv: and when you run maas-import-ephemerals, it adds different kernel args
[15:22] <roaksoax> jtv: so we just need a default, and we need to be able to add and modify them as needed
[15:22] <jtv> That's all plumbing we don't have yet, I think.
[15:23] <roaksoax> jtv: the addition of new custom kernel args is done by both, maas-import-isos and maas-import-ephemerals. So we should be able to do similar way to add more kernel args. However, we should be able to also add kernel args per node and as needed
[15:25] <jtv> But we're driving this from the inside out, rather from the outside in as Cobbler does.
[15:26] <jtv> In Cobbler, the scripts make this stuff up because they need to create profiles and that's where this stuff gets stored.  We're driving it from the database, independently from the import scripts.
[15:28] <roaksoax> jtv: right, so import a netboot or server ISO into cobbler and check what args are added by default :).
[15:28] <roaksoax> jtv: should be similar to http://pastebin.ubuntu.com/1049288/
[15:29] <jtv> Have to organize this in some way that makes it reusable.
[15:29] <rvba> roaksoax: another question for you: right now, a node, once the install is done, calls "wget "http://$http_server:$http_port/cblr/svc/op/nopxe/system/$system_name" -O /dev/null".  We've already created the equivalent method (to disable PXE boot) on the metadata API.  The trouble is that the call to this method needs to be oauth-authenticated.  I'm afraid I'm going to have to expose that method anonymously.  Do
[15:29] <rvba> you have a better idea?
[15:31] <rvba> roaksoax: I'm asking you because I'm not sure what the limitations are exactly… could we use a tool with oauth support instead of wget?
[15:33] <jtv> We have some code that we were hoping to extract into a library at some point…
[15:33] <jtv> Based on the juju code.
[15:33] <jtv> src/apiclient/maas_client.py
[15:33] <roaksoax> rvba: can't we just do similar way to how cloud init does it?
[15:34] <rvba> roaksoax: what would that mean exactly?
[15:34] <roaksoax> rvba: hold on, have an idea
[15:35] <roaksoax> rvba: ok, so for cloud-init, MAAS gives cobbler the MAAS_PRESEED which is a base64 blob right?
[15:36] <rvba> roaksoax: yes
[15:37] <rvba> roaksoax: in the cobbler-less world, this is not base64 encoded any more, but obviously the same data is present in the preseed context when it gets rendered.
[15:38] <roaksoax> rvba: ok, so even better then, in the template, we can just get the auth credentials and hand them to curl
[15:39] <roaksoax> rvba: unless ytou want to store them in base64. So in the template, we just get the variable where that blod is stored, decode it, separate the credentials and give them out to curl
[15:40] <rvba> roaksoax: are you sure that curl supports oauth?
[15:40] <roaksoax> rvba: weren't we using it at the beginning for enlistment? before we decided not to?
[15:41] <rvba> roaksoax: I /think/ we were using a python script.
[15:41] <roaksoax> rvba: if it does not, then we need it to be unauthenticated
[15:41] <roaksoax> rvba: unless we get a python script that runs the credentials
[15:41] <roaksoax> rvba: which might be a PITA
[15:43] <rvba> roaksoax: why would it be a PITA?
[15:44] <roaksoax> rvba: downloading a script of late_command and executing it
[15:44] <roaksoax> rvba: another thing we could do would be do something we used to do with juju/orchestra
[15:44] <roaksoax> rvba: which was basically have a base64 blob that decoded into a .py file which was run
[15:45] <roaksoax> rvba: it should allow us to import python-oauth
[15:45] <roaksoax> rvba: but it is even more prone to errors
[15:45] <rvba> Indeed.
[15:45] <rvba> And less testable.
[15:46] <roaksoax> indeed
[15:46] <roaksoax> that's the same issue with wgetting a python script and having to do the same thing
[15:46] <roaksoax> rvba: so I think we'd have to have it unauthenticated for ease
[15:47] <rvba> Yep, I'm afraid that's the only simple solution.
[15:47] <rvba> Note that we will have kinda the same problem when exposing the preseed files.  The preseed contains authentication creadentials… and I'm afraid we will have to expose it in the open.
[15:49] <rvba> roaksoax: because that preseed url will be passed as a kernel argument (url=...) and I suppose we don't have a way to use some kind of authentication when fetching that file do we?
[15:49] <roaksoax> rvba: right, but the preseed should only be obtainable by the PXE client machine and not by simply browsing the server http
[15:51] <roaksoax> rvba: or only exposed on the network which PXE is happening
[15:51] <roaksoax> rvba: and not that I know of.
[15:51] <rvba> roaksoax: how can we prevent someone browsing the server to fetch that file?
[15:52] <rvba> roaksoax: I mean how can we tell the request comes from the legitimate client machine?  Using the IP would be a start but kinda fragile.
[15:54] <roaksoax> rvba: i have seen websites on which for instances, if you access to http://www.test.com/files/ it does not list the files, however, if you access test.com/files/file1.xyz it is downloadable
[15:54] <roaksoax> rvba: it is probably just hidding
[15:54] <roaksoax> it
[15:55] <rvba> Preventing directory listing is not a proper protection.
[15:55] <rvba> Besides, this is an API so there is no directory listing anyway.
[15:56] <roaksoax> yeah, never mind me then
[15:56] <roaksoax> err don't mind me :)
[16:00] <rvba> roaksoax: anyway, I think we will concentrate on having full feature parity with cobbler without making things worse w.r.t. security.  And then we will iterate on that.  Thanks for your help.
[16:00] <roaksoax> welcome
[16:02] <negronjl> ... and back again with this
[16:04] <negronjl> roaksoax: 'morning.  have another question for ya
[16:05] <negronjl> roaksoax: yesterday you had mentioned that we should be using the router as DHCP/DNS
[16:05] <negronjl> roaksoax: If we are using the router as DHCP/DNS should we turn dnsmasq off on the maas server ?
[16:06] <negronjl> roaksoax: or is the router acting as a relay ( DHCP/DNS relay that is ) ?
[16:08] <roaksoax> negronjl: when you remove maas-dhcp dnsmasq will no longer be running for cobbler
[16:08] <roaksoax> sudo cobbler sync should update the settings to not run it
[16:08]  * roaksoax brb
[16:08] <negronjl> roaksoax: ok ...thx
[16:17] <jtv> Daviey, roaksoax: actually… should the kernel options for enlistment be based on the “ephemeral” kopts or on the “install” kopts?
[16:22] <negronjl> I am now getting a Bad archive mirror error
[16:32] <negronjl> roaksoax, Daviey: i need help with the mirror ... it doesn't seem to be working.  I get "Bad archive error" on the screen.  On syslog ( on the node ), I get mirror does not have any suite symlinks and mirror does not support the specified release (precise) but, when I run the wget command manually ( wget -q http://archive.ubuntu.com/ubuntu/dists/precise/Release -O - | grep -E '^(Suite|Codename):' I do get Suite: precise .... this st
[16:32] <negronjl> opped working when I stopped dnsmasq BTW to allow the router to answer DHCP requests.
[16:43] <negronjl> Daviey, roaksoax:  nm ... i figured it out
[16:50] <Daviey> spoofing archive.ubuntu.com to the wrong ip address?
[17:21] <negronjl> Daviey: not really sure but, I just edited /etc/dnsmasq.conf and commented out all of the dhcp parts and it seems to be working now
[17:53] <Daviey> negronjl: wait, on the router?!
[17:53] <negronjl> Daviey: the router is the DHCP/DNS
[17:54] <negronjl> Daviey: I disabled dhcp on dnsmasq on the maas server
[17:54] <Daviey> negronjl: hmm, who installed dnsmasq on the maas server?!
[17:55] <negronjl> Daviey:  no idea .. it was like that when I got it.
[17:55] <negronjl> Daviey: I still need dnsmasq running on maas server.  If I don't, the nodes complain about a bad archive mirror ...
[17:55] <Daviey> hmm
[17:56] <negronjl> Daviey: so I just commented the dhcp parts of /etc/dnsmasq.conf and it seems to work better.
[17:56] <Daviey> negronjl: the router should handle all dhcp/dns..
[17:56] <negronjl> Daviey: now I am to the point where juju bootstrap works ( still without WoL ) but, when I do juju status, the command hangs
[17:56] <Daviey> or at least it did, when i left it
[17:56] <negronjl> Daviey: the router does
[17:57] <Daviey> juju status should hang, until the boostrap node is ready, no?
[17:57] <Daviey> negronjl: WOL should work, only when a node is cleanly shutdown
[17:58] <Daviey> negronjl: i did suggest attempting to switch out the power control 'card' in the APC to the other one.
[17:58] <Daviey> The one which you can plug into the wall, has a duff management card.
[17:58] <Daviey> I hoped you could just swap it out.
[18:01] <negronjl> Daviey: I could try that.  I should open both PDUs and see about that
[18:01] <negronjl> Daviey: that would still leave me with the juju status problem.
[18:04] <Daviey> negronjl: juju status should hang until the ootstrap node is ready
[18:04] <Daviey> is it ready?
[18:05] <negronjl> Daviey:  it is at the login prompt ( has been like that for 10 minutes already ). Nothing is happening.
[18:05] <negronjl> Daviey: juju status is still hanging
[18:06] <Daviey> negronjl: okay, then this wouldn't seem to be a maas problem as such.. but a juju problem
[18:06] <Daviey> have you logged in to see what the heck is going on?
[18:06] <Daviey> zookeeper poorly?
[18:06] <lifeless> negronjl: Daviey: how is it looking?
[18:07] <negronjl> lifeless, Daviey:  I am trying again.
[18:08] <negronjl> lifeless:  I have the MaaS part working ( no WoL yet ).  juju bootstrap is working but, juju status hangs
[18:08] <negronjl> lifeless: at the same time, Daviey recommends i open the PDUs and take a part out of one to put it into the other to see if the PDU works.
[18:13] <Daviey> I haven't seen the new one, but if they are the same model.. that makes sense.. the old model had a duff management computer
[18:13] <Daviey> WoL is fragile at best, but handled with care.. and a dry run on stage to prepare them.. should leave them in an adequate state
[18:14] <negronjl> Daviey:  I'm trying to open the PDUs right now ... re: WoL, I am trying to see if I can reproduce it to where I can consistently have WoL working ... so far .. no luck.
[18:15] <Daviey> (WOL only works if cleanly shutdown)
[18:16] <negronjl> Daviey: within the maas process, how do you go about cleanly shutting them down when maas controls the entire process.
[18:17] <Daviey> negronjl: doing a dry run, turning them on by hand.. then briefly pressing the power button to shut them down.
[18:17] <Daviey> (brief press)
[18:21] <negronjl> Daviey: WoL is now working ....
[18:21] <negronjl> Daviey: waiting on bootstrap node ( connecting monitor to it )
[18:23] <negronjl> lol ... now I get Invalid SSH key ...
[18:26] <Daviey> :S
[18:29] <lifeless> Daviey: seen that before ?
[18:35] <Daviey> lifeless: no
[18:35] <lifeless> SpamapS: roaksoax: ^ ?
[18:35] <Daviey> It's not clear to me if it is the juju ssh key, or the maas ssh key failing
[18:35] <Daviey> if there is no maas ssh key, then it could be a hint that juju/zookepper has failed.
[18:35] <Daviey> (or not yet online)
[18:35] <SpamapS> lifeless: sorry I missed the context
[18:36] <Daviey> negronjl: can you confirm that there is an accurate ssh key in MAAS ui?
[18:36] <Daviey> attacking the problem by allowing maas to inject the ssh key AND juju inserting the same key makes much sense.
[18:37] <lifeless> SpamapS: we've now gotten everything seeming to run up, but juju status is whinging about the ssh key being invalid
[18:38] <SpamapS> raw ssh works?
[18:38] <Daviey> juju would be using the same ssh key as the laptop, so i can't imagine raw ssh would work
[18:39]  * Daviey wonders if negronjl is using the laptop, rather than his own machine.
[18:39] <SpamapS> worth a try with ssh -v
[18:40] <Daviey> i'd rather negronjl validates the ssh key in MAAS ui first.
[18:40] <SpamapS> ok I still dont' know what I'm being asked. :)
[18:40]  * Daviey suspects it is null
[18:40] <negronjl> Daviey: I am using the ssh key in the maas server
[18:40] <negronjl> i have NOT put any ssh key in the maas server UI
[18:41] <Daviey> if the database has been reset since i last had my mits on it, it will be empty.
[18:41] <Daviey> you need to copy and paste the public key into the maas ui, under Pref's
[18:42] <negronjl> Daviey:  ok ... let me try that ... brb
[18:42] <Daviey> ... my hunch is that juju/zookeeper is borked.. and you can't get into it, as no key has been injected.
[18:43] <Daviey> negronjl: on a side note, did you update the local apt mirror?
[18:43] <negronjl> Daviey: no ... this is an isolated network
[18:43]  * lifeless hopes thats good :)
[18:44]  * negronjl too .  It has the same archive as it did when it was "working"
[18:44] <Daviey> well, if it were me.. i'd have updated tto precise release.. curently it's pre-release archive
[18:44] <Daviey> negronjl: well, juju seems fragile in that snapshot.
[18:45] <Daviey> i'd hope that precise-release is more stable :)
[18:45] <lifeless> negronjl: so, you're adding the ssh key to maas and trying again I guess?
[18:46] <negronjl> Daviey: this experience has me second-guessing everything that i do fearing i may break more things ...
[18:46] <negronjl> Daviey: Do you think I should update the mirror ?  If so,  how ?
[18:46] <Daviey> negronjl: i don't blame you at all.
[18:47] <Daviey> negronjl: Lets see the current status.. juju destroy-enviroment .. add you key to the MAAS ui
[18:47] <Daviey> (your key == the laptop key)
[18:47] <Daviey> do a juju bootstrap.. so at least you can sanely get into the box
[18:48] <Daviey> then, if we hit the same problem again.. SpamapS will be a better person to help debug why juju has failed.
[18:48] <Daviey> (sorry SpamapS!)
[18:48] <negronjl> Daviey: did that already .. waiting on the node to come to life
[18:48] <Daviey> negronjl: cool
[18:49] <SpamapS> Daviey: no worries, I'm good at ignoring other peoples' problems
[18:50] <Daviey> hah
[18:52] <Daviey> negronjl: if i am not mistaken, the public key should be http://pb.daviey.com/ITE8/
[18:53] <negronjl> Daviey: it is
[18:53] <Daviey> negronjl: for /what to do/, get inspiration from /home/ubuntu/charms/timed-deploy.sh
[18:54] <Daviey> utils.py is really helpful
[18:54] <negronjl> Daviey: i have.  It is useful indeed.
[18:55] <Daviey> negronjl: timed-deploy.sh was an effort to leave it in an install loop, bootstrap, deploy, teardrop
[18:55] <Daviey> teardown
[18:57] <SpamapS> hah
[18:57] <SpamapS> freudian slip?
[18:57] <SpamapS> the deploy does indeed bring tears to some peoples' eyes :)
[18:58] <Daviey> SpamapS: one of my servers is called teardrop, so automagic hangs :)
[18:58] <Daviey> s/hangs/hands
[18:58] <Daviey> damn, i need to EOD
[18:58] <negronjl> Daviey: thx for all of your help :)
[18:59] <Daviey> negronjl: I won't go just yet, but my typing suggests i should :)
[18:59] <negronjl> Daviey: ah .. ok  ... well thanks for your help anyway ... and stick around to help some more :)
[19:09] <negronjl> The bootstrap machine is done.... juju status still hangs
[19:09] <negronjl> I was able to ssh into the box
[19:09] <SpamapS> negronjl: status may be hanging on zookeeper starting up
[19:09] <Daviey> with the ssh key in MAAS aswell?
[19:09] <SpamapS> sshd comes up before zk
[19:10] <negronjl> i see that /var/log/cloud-init-output.log has something about juju-admin:  error: unrecognized arguments: --constraints-data=......
[19:10] <Daviey> right, but if the ssh public key came from the metadataservice,he should be able to directly ssh now.
[19:10] <Daviey> of crap
[19:10] <negronjl> SpamapS: zookeeper is running
[19:10] <SpamapS> negronjl: I bet you have an old juju
[19:10] <SpamapS> negronjl: dpkg -l juju
[19:11] <negronjl> node:  504
[19:11] <negronjl> maas server: 539
[19:11] <negronjl> that is not right :/
[19:13] <SpamapS> yeah but that should support --constraints-data
[19:13] <SpamapS> oh wait
[19:13] <SpamapS> node 504
[19:13] <SpamapS> yeah thats broken
[19:13] <SpamapS> negronjl: apt-cache policy juju
[19:13] <SpamapS> where did that come from?!
[19:13]  * SpamapS saw 504 and thought 540
[19:14] <Daviey> SpamapS: hapen to know off the top of your head what precise juju paa is?
[19:14] <Daviey> i want to rule out that it isn't usin that
[19:16] <SpamapS> juju should have 531
[19:16] <SpamapS> precise rather
[19:16] <SpamapS>      0.5+bzr531-0ubuntu1.1 0
[19:16] <SpamapS>         400 http://archive.ubuntu.com/ubuntu/ precise-proposed/universe amd64 Packages
[19:16] <SpamapS>      0.5+bzr531-0ubuntu1 0
[19:16] <SpamapS>         500 http://mirrors.kernel.org/ubuntu/ precise/universe amd64 Packages
[19:18] <negronjl> SpamapS: this does appear to now be a juju issue
[19:19] <negronjl> SpamapS: I see that "juju status" is trying to connect to the node but, it just hangs there .
[19:19] <negronjl> SpamapS: on the node, I do see juju.agents.machine and juju.agents.provision running
[19:20] <negronjl> SpamapS: I am also able to ssh into the node ( ssh ubuntu@<node> )
[19:20] <negronjl> SpamapS: any thoughts?
[19:21] <SpamapS> negronjl: apt-cache policy juju
[19:21] <SpamapS> negronjl: wherever that 504 came from, thats your problem
[19:21] <SpamapS> negronjl: its well before constraints landed, and your client is incompatible with that version entirely
[19:22] <negronjl> SpamapS: damn ... that means that I have to refresh my mirror
[19:22] <SpamapS> perhaps a very out of date mirror on the node?
[19:22] <negronjl> SpamapS: I need help doing that
[19:22] <SpamapS> negronjl: yeah get that mirror up to date, *OR* point your machine at it, and use the same client version
[19:23] <negronjl> SpamapS: got it. apt-mirror correct ?
[19:24] <SpamapS> no idea
[19:24] <SpamapS> I mean, thats likely
[19:24] <negronjl> SpamapS: no worries ... I found a script that does it for me ....
[19:24] <SpamapS> Daviey: ^^ can you help negronjl get the mirror updated?
[19:26] <Daviey> cool
[19:26] <negronjl> Daviey: I found a create-mirror.sh script in the maas-server.
[19:26] <negronjl> it seems to create the mirror
[19:30] <Daviey> negronjl: that is the one we used, yes.
[19:33] <negronjl> Daviey:  running it now ... it needs to downlod 5.5G of data so, it will take a while ...
[19:34] <Daviey> cool
[19:51] <SpamapS> 5.5G? ugh
[19:51] <SpamapS> negronjl: hopefully you're not doing that on conference wifi ;)
[19:53] <negronjl> SpamapS: nope ... this is an isolated network
[19:54] <negronjl> SpamapS: I am downloading everything that I need into the maas server .
[19:54] <negronjl> SpamapS: and spoofing archive.ubuntu.com so I can server it all locally.