Wednesday, 2013-11-20

« Tuesday, 2013-11-19 Index Thursday, 2013-11-21 »
*** xmltok has joined #openstack-ironic00:01
*** tatyana has quit IRC00:03
*** xmltok has quit IRC00:06
*** epim has joined #openstack-ironic00:07
*** harlowja has quit IRC00:16
*** matsuhashi has joined #openstack-ironic00:29
*** coolsvap has quit IRC00:30
NobodyCamspinning disks are so slow :-p00:44
NobodyCamlol at least the one in my laptop00:44
*** matsuhashi has quit IRC00:48
*** matsuhashi has joined #openstack-ironic00:49
*** matsuhashi has quit IRC00:53
devanandaback00:56
NobodyCamwb00:56
NobodyCamok it dinner time for /me ... have a good night all01:07
*** matsuhashi has joined #openstack-ironic01:08
*** nosnos has joined #openstack-ironic01:09
devanandag'night!01:16
*** matsuhashi has quit IRC01:34
*** matsuhashi has joined #openstack-ironic01:35
*** matsuhashi has quit IRC01:40
*** matsuhashi has joined #openstack-ironic01:43
Haomenggood morning Ironic:)01:48
*** harlowja has joined #openstack-ironic01:51
*** sjing has joined #openstack-ironic01:58
*** sjing has quit IRC02:07
*** sjing has joined #openstack-ironic02:09
*** rloo has quit IRC02:09
openstackgerritA change was merged to openstack/ironic: Support building wheels (PEP-427)  https://review.openstack.org/5712202:48
*** xmltok has joined #openstack-ironic02:56
*** xmltok has quit IRC03:01
*** epim has quit IRC03:02
*** michchap has quit IRC03:21
*** michchap has joined #openstack-ironic03:22
*** matsuhashi has quit IRC03:31
*** matsuhashi has joined #openstack-ironic03:31
*** sjing has quit IRC03:32
*** sjing has joined #openstack-ironic03:33
*** matsuhashi has quit IRC03:36
*** matsuhashi has joined #openstack-ironic03:41
openstackgerritHaomeng,Wang proposed a change to openstack/python-ironicclient: Enable created_at/updated_at for port-show/chassis-show  https://review.openstack.org/5399103:53
*** coolsvap has joined #openstack-ironic04:16
*** prekarat has joined #openstack-ironic04:19
*** arata has joined #openstack-ironic04:46
*** arata has quit IRC05:14
*** datajerk has joined #openstack-ironic05:34
*** datajerk has quit IRC05:39
*** coolsvap has quit IRC05:39
Haomenggood morning/afternoon/evening Ironic:)05:40
Haomenglifeless:ping05:41
lifelessHaomeng: hi05:42
HaomengI have a new patch, which passed python27 and python33 gate, but it has pypy issue, can you help take a look? http://logs.openstack.org/91/53991/8/check/gate-python-ironicclient-pypy/68c2790/console.html  , thank you05:42
Haomenglooks my mock code does not work on pypy05:42
Haomengthank you lifeless:)05:43
lifelessHaomeng: I've no immediate ideas. Have you tried it yourself?05:52
*** harlowja has quit IRC05:52
Haomengsure, I will install pypy in my local and debug, anyway thank you lifeless:)05:52
Haomengjust think this is strange issue with pypy for mocking05:53
Haomengto discuss with you, so dont worry05:55
*** coolsvap has joined #openstack-ironic05:58
openstackgerritJenkins proposed a change to openstack/ironic: Imported Translations from Transifex  https://review.openstack.org/5709906:00
*** michchap has quit IRC06:04
*** michchap has joined #openstack-ironic06:05
*** datajerk has joined #openstack-ironic06:06
*** matsuhashi has quit IRC06:08
*** matsuhashi has joined #openstack-ironic06:08
lifelessHaomeng: yeah, it is odd :)06:10
*** datajerk has quit IRC06:10
*** matsuhashi has quit IRC06:13
*** nosnos_ has joined #openstack-ironic06:25
*** nosnos has quit IRC06:29
*** matsuhashi has joined #openstack-ironic06:29
*** nosnos_ has quit IRC07:05
*** sjing has quit IRC07:05
*** arata has joined #openstack-ironic07:05
*** nosnos has joined #openstack-ironic07:05
*** sjing has joined #openstack-ironic07:06
*** datajerk has joined #openstack-ironic07:06
*** datajerk has quit IRC07:10
*** matsuhashi has quit IRC07:22
*** matsuhashi has joined #openstack-ironic07:23
*** matsuhas_ has joined #openstack-ironic07:30
*** matsuhashi has quit IRC07:34
GheRiveromorning Ironic07:35
*** datajerk has joined #openstack-ironic07:37
*** datajerk has quit IRC07:42
*** tatyana has joined #openstack-ironic07:43
sandeeprGheRivero, morning07:47
*** prekarat has quit IRC08:04
*** romcheg has joined #openstack-ironic08:04
*** datajerk has joined #openstack-ironic08:08
*** ndipanov has joined #openstack-ironic08:12
*** datajerk has quit IRC08:13
*** tatyana has quit IRC08:14
*** matsuhas_ has quit IRC08:18
*** matsuhashi has joined #openstack-ironic08:19
*** matsuhashi has quit IRC08:24
yuriyzmorning GheRivero08:25
*** matsuhashi has joined #openstack-ironic08:31
*** arata has left #openstack-ironic08:38
*** datajerk has joined #openstack-ironic08:39
*** datajerk has quit IRC08:43
*** epim has joined #openstack-ironic08:47
*** derekh has joined #openstack-ironic08:54
*** romcheg has quit IRC08:55
openstackgerritYuriy Zveryanskyy proposed a change to openstack/ironic: Allow vendor passthru without Keystone authorization  https://review.openstack.org/5661208:58
*** jistr has joined #openstack-ironic08:58
*** lucasagomes has joined #openstack-ironic09:09
*** datajerk has joined #openstack-ironic09:10
*** arata has joined #openstack-ironic09:10
*** epim has quit IRC09:10
*** romcheg has joined #openstack-ironic09:11
*** datajerk has quit IRC09:14
*** sjing has quit IRC09:17
*** arata has quit IRC09:19
*** datajerk has joined #openstack-ironic09:41
*** datajerk has quit IRC09:45
*** bigjools has quit IRC09:59
*** bigjools has joined #openstack-ironic10:00
*** prekarat has joined #openstack-ironic10:08
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: API ports update for WSME 0.5b6 compliance  https://review.openstack.org/5668210:11
*** datajerk has joined #openstack-ironic10:12
*** prekarat has quit IRC10:13
*** matsuhashi has quit IRC10:14
*** datajerk has quit IRC10:16
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Rework patch validation on Ports  https://review.openstack.org/5698410:20
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Rework patch validation on Ports  https://review.openstack.org/5698410:25
*** romcheg1 has joined #openstack-ironic10:35
*** romcheg has quit IRC10:39
*** datajerk has joined #openstack-ironic10:43
*** datajerk has quit IRC10:47
*** max_lobur_afk is now known as max_lobur10:49
*** coolsvap has quit IRC10:57
*** liteIRC has joined #openstack-ironic10:59
*** tatyana has joined #openstack-ironic11:06
*** liteIRC has quit IRC11:07
*** datajerk has joined #openstack-ironic11:14
*** romcheg has joined #openstack-ironic11:14
*** romcheg1 has quit IRC11:19
*** datajerk has quit IRC11:19
*** mattymo has joined #openstack-ironic11:20
openstackgerritA change was merged to openstack/python-ironicclient: Remove in-place try/except blocks from shell commands  https://review.openstack.org/5728511:24
*** datajerk has joined #openstack-ironic11:45
*** datajerk has quit IRC11:49
*** nosnos_ has joined #openstack-ironic12:09
*** lucasagomes is now known as lucas-hungry12:12
*** nosnos has quit IRC12:12
*** nosnos_ has quit IRC12:14
*** datajerk has joined #openstack-ironic12:16
*** datajerk has quit IRC12:20
*** michchap_ has joined #openstack-ironic12:25
*** michchap has quit IRC12:25
*** datajerk has joined #openstack-ironic12:47
Haomengmorning GheRivero yuriyz12:49
*** datajerk has quit IRC12:51
openstackgerritA change was merged to openstack/ironic: Imported Translations from Transifex  https://review.openstack.org/5709912:59
*** datajerk has joined #openstack-ironic13:08
*** datajerk has quit IRC13:13
HaomengIronic, who know how to setup pypy env for our unittest, I have a pypy issue, want to debug with pypy env - https://review.openstack.org/#/c/53991/13:23
Haomengmy code is working with py26/27, but does not work with pypy, any idea/comments?13:23
Haomengand try to install pypy unittest env, enuntered could not install deps  for tox setup command "tox -e pypy"13:25
*** tatyana has quit IRC13:27
*** max_lobur has quit IRC13:28
*** max_lobur has joined #openstack-ironic13:29
*** tatyana has joined #openstack-ironic13:32
max_loburHi Haomeng13:37
max_loburI've posted comments on 5399113:37
max_loburplease try, maybe you will get around pypy issue13:38
*** jdob has joined #openstack-ironic13:40
Haomengthank you13:45
Haomengmax_lobur:thank you13:46
*** jdob has quit IRC13:48
*** jdob has joined #openstack-ironic13:48
Haomengmax_lobur: I checked your comments, that is good idea, will create new shell test scripts: test_port_shell, test_node_shell and   test_chassis_shell for each object13:57
Haomengmax_lobur: nice day:)13:57
max_loburok =) I suggested to ask someone else.. I'm not sure we'll scale shell tests so far14:00
max_loburso maybe one file is fine14:00
max_loburneed to clarify14:00
Haomengyes, dont worry, let me try the unittest for _print_port_show method only14:04
Haomengmax_lobur: I agree with your comments, if we have separated files, that is clear and for future more unittest code, will discuss with others14:07
max_loburOk, cool14:08
Haomengit works, thank you max14:17
HaomengI want to commit the patch and add core reviewer to see if they have more concern/comments14:18
*** jdob has quit IRC14:18
*** jdob has joined #openstack-ironic14:19
*** jbjohnso has joined #openstack-ironic14:19
jbjohnsomorning all14:19
Haomenglet us create  new shell test scripts: test_port_shell, test_node_shell and   test_chassis_shell for each object and to review14:19
jbjohnsoI was wondering about holding a screen shared demonstration of some code that might be the best path for console access14:20
jbjohnsosadly, it would require participants to use java...14:20
jbjohnsoI see the ipmitool blueprint and think this would be a better path (less resource intensive, fewer shellinaboxes, etc)14:21
Haomengjbjohnso: morning14:22
jbjohnsoHaomeng, morning14:22
Haomengjbjohnso: are you talking about our bp - https://blueprints.launchpad.net/ironic/+spec/serial-console-access ?14:24
jbjohnsoHaomeng, yeah14:25
Haomengneed to set  IPMI Serial-over-LAN mode first from BMC14:31
Haomengjbjohnso: need to set  IPMI Serial-over-LAN mode first from BMC14:31
Haomengjbjohnso: but, by default, I think  IPMI Serial-over-LAN is not enabled14:32
Haomengjbjohnso: have to setup by manually from BMC side first14:32
Haomengmax_lobur: I checked Ceilometer client and Heat client, they have a lot of shell test unittest, but only one test_shell.py, so can we just use one single test_shell to cover our shell unittest ?14:41
Haomengmax_lobur: and I want to commit the code with single test_shell file first, to see the review comments from core reviewer14:42
max_loburHaomeng, I assume they also have a single file for shell code?14:42
max_loburam I right?14:42
Haomenglet me find out14:42
NobodyCamGood Morning Ironic14:45
GheRiveromorning NobodyCam14:45
NobodyCamhey hey GheRivero :)14:46
NobodyCami am just making coffee, but wanted to how full your plat is14:46
Haomenghttps://github.com/openstack/python-ceilometerclient/blob/master/ceilometerclient/tests/test_shell.py  and https://github.com/openstack/python-heatclient/blob/master/heatclient/tests/test_shell.py14:46
HaomengNobodyCam: morning14:46
NobodyCamgood morning Haomeng14:47
HaomengNobodyCam: :)14:47
HaomengNobodyCam: I am evening:)14:48
HaomengNobodyCam: np:)14:48
NobodyCamheheheh oh thats right14:48
NobodyCamI don't have any coffee yet14:48
Haomeng:)14:49
max_loburHaomeng, take a look, https://github.com/openstack/python-ceilometerclient/blob/master/ceilometerclient/v1/shell.py they have single shell for all objects, therefore they heave single test file14:51
Haomengmax_lobur: yes, you are right - > https://github.com/openstack/python-ceilometerclient/blob/master/ceilometerclient/tests/test_shell.py  and https://github.com/openstack/python-heatclient/blob/master/heatclient/tests/test_shell.py14:51
*** ekarlso has quit IRC14:51
max_loburfor us, since we split files by objects, i thing we need to copy that structure in tests14:52
max_loburto make it easier to find corresponding tests14:52
*** ekarlso has joined #openstack-ironic14:52
Haomengmax_lobur: ok, good idea14:52
HaomengHaomeng: yes14:52
*** ekarlso has quit IRC14:52
Haomenglet me split files and commit14:52
Haomengmax_lobur:  let me split files and commit, thanks for your ideas14:52
max_lobursure, np :)14:53
Haomengmax_lobur: :)14:53
*** ekarlso has joined #openstack-ironic14:53
max_loburmorning NobodyCam =)14:53
max_loburand GheRivero14:53
HaomengGheRivero: morning:)14:54
jbjohnsoHaomeng, depends on the manufacturer14:54
Haomengjbjohnso: yes14:54
jbjohnsoHaomeng, in xCAT, that's handled by 'bmcsetup'14:54
jbjohnsointel was one vendor that disables everything by default14:54
Haomengjbjohnso: so we have to make sure to enable it first14:55
Haomengjbjohnso: how do you think if we can enable it by our code, maybe need BIOS interface14:55
GheRiveroNobodyCam: so, what is the final idea about having a dhcp server for pxe? are we going for it?14:57
jbjohnsoHaomeng, usually not14:58
jbjohnsoHaomeng, the knobs to kick SOL to life I have always seen14:58
jbjohnsoare industry standard14:58
jbjohnsowell, they are if you understand the situation well enough14:58
jbjohnsothe script is horribly messy and needs to be cleaned up, however...14:58
jbjohnsoHaomeng, http://sourceforge.net/p/xcat/xcat-core/ci/master/tree/xCAT-genesis-scripts/bin/bmcsetup15:00
Haomengjbjohnso: ok, let me check, thank you15:00
jbjohnsothat is part of the starting point that Sun Jing is coming from in pursuit of https://blueprints.launchpad.net/ironic/+spec/discovery-ramdisk15:00
Haomengjbjohnso: this is xcat code for bmcsetup15:00
Haomengjbjohnso: yes15:01
jbjohnsoHaomeng, right, the key takeaway is...15:01
jbjohnsofor sol enablement...15:01
jbjohnsoecho -n "Enabling SOL for channel $LANCHAN:"15:01
jbjohnsoso that weird 'raw 0xc 0x21 $LANCHAN 1 1'15:01
jbjohnsokicks it to life15:01
jbjohnsofor the channel15:02
Haomengjbjohnso: this line "ipmitool -d $idev raw 0xc 0x21 $LANCHAN 0x1 0x1 ", right?15:02
jbjohnsothen you see raw 6 0x4c $LANCHAN $USER 2 0 0 015:02
jbjohnsoyeah15:02
Haomengjbjohnso: this ipmitool command can enable sol15:02
Haomengjbjohnso: cool15:02
jbjohnsoalso, putting SOL on the channel15:02
jbjohnsothere are several there15:02
jbjohnsoIntel bmc is the only bmc that requires those, but it's been safe to do everywhere in our experience15:02
jbjohnsoit is the only one we saw15:03
Haomengjbjohnso: yes, that is xcat has done15:03
jbjohnsoenabling the lan channel15:03
jbjohnsois another thing we had to do15:03
Haomengjbjohnso: we can port these code to our Ironic15:03
jbjohnsoHaomeng, right, that is our intent15:04
Haomengjbjohnso: how do you think if that is working in our Ironic15:04
Haomengjbjohnso: great15:04
Haomengjbjohnso: I guss you are xcat expert, great:)15:04
jbjohnsoGheRivero, NobodyCam so interesting thing15:05
jbjohnsoabout dhcp serving15:05
jbjohnsowhat if the same process managed consoles and PXE configuration, no tight dhcp integration required?15:05
GheRiverohow will it be done?15:06
jbjohnsoGheRivero, 401115:08
jbjohnsoGheRivero, now things like rfc 4173 are out of reach15:08
jbjohnsoand *some* dhcp server is needed to provide ip15:09
jbjohnsobut 'filename' doesn't need to be served up by dhcpd15:09
jbjohnsoat least not to fully pxe compliant beasts (POWER is notably excluded from that party)15:09
Haomengmax_lobur: Max, can we define same class "class ShellTest(utils.BaseTestCase)"  in our new test_port_shell.py to cover port shell unittest?15:09
max_loburwhy not15:10
max_loburjust PortShellTest maybe15:10
NobodyCamGheRivero: I was about to have a chat with dkehn about just that, How your plate looking?15:11
Haomengmax_lobur: yes15:11
Haomengmax_lobur: PortShellTest is meanful15:11
Haomengmax_lobur: will do, and help me review, thank you Max15:12
GheRiveroNobodyCam: i-m always hungry :)15:12
NobodyCamGheRivero: want to join the skype call we are about to have?15:13
GheRiverojbjohnso: 4011?15:13
GheRiveroNobodyCam: sure..15:13
jbjohnsoGheRivero, proxydhcp15:13
jbjohnsohttp://en.wikipedia.org/wiki/Preboot_Execution_Environment15:14
jbjohnso"The Proxy DHCP service may also run on the same host as the standard DHCP service. Since two services cannot share port 67/UDP, the Proxy DHCP runs on port 4011/UDP and expects the extended DHCPDISCOVER packets from PXE Clients to be DHCPREQUESTs."15:14
NobodyCamdkehn: I'm going to try a sskype conf call... I not sure I have both of you on G+15:15
GheRiverowill add to the list of things to read this week :) but looks promising15:16
*** blamar has quit IRC15:16
*** coolsvap has joined #openstack-ironic15:17
jbjohnsoanyway, my 'console management' service15:21
jbjohnsois a bit more potent in architecture than a console management service, proxydhcp and slp are on the list of things to manage15:21
jbjohnsobut the console works today15:21
*** blamar has joined #openstack-ironic15:23
*** ben_duyujie has joined #openstack-ironic15:27
*** ben_duyujie has quit IRC15:30
openstackgerritHaomeng,Wang proposed a change to openstack/python-ironicclient: Enable created_at/updated_at for port-show/chassis-show  https://review.openstack.org/5399115:30
Haomengmax_lobur: I committed new patch, if you have time, help me take a look again, thank you, nice day15:31
Haomengmax_lobur: I am tired, will sleep:) see you:)15:31
*** ben_duyujie has joined #openstack-ironic15:31
max_loburbye. I'll review15:33
dkehnhttps://wiki.openstack.org/wiki/Icehouse_Release_Schedule15:33
dkehnfor those on call15:33
*** ben_duyujie1 has joined #openstack-ironic15:34
*** ben_duyujie has quit IRC15:38
NobodyCamdkehn: GheRivero https://review.openstack.org/#/c/51328/2/nova/virt/ironic/driver.py15:40
*** lucas-hungry is now known as lucasagomes15:46
*** yuriyz has quit IRC15:56
*** datajerk has joined #openstack-ironic16:03
*** prekarat has joined #openstack-ironic16:23
devanandag'morning, all16:26
NobodyCamgood morning devananda16:27
*** romcheg has quit IRC16:28
*** coolsvap has quit IRC16:29
devanandalucasagomes: i've got API docs (almost) auto generating: https://review.openstack.org/#/c/57346/116:33
NobodyCamhttp://falkvinge.net/2013/11/17/nsa-asked-linus-torvalds-to-install-backdoors-into-gnulinux/#!16:33
lucasagomesdevananda, w00ts! lemme take a look16:33
*** prekarat has quit IRC16:33
lucasagomesI'm almost finshing the complex type validations for nodes :)16:34
lucasagomesdevananda, awesome!16:36
lucasagomesjust to think that we don't need to write docs for the API anymore16:36
lucasagomesthat's great16:36
NobodyCam+++++++16:36
*** idknick has joined #openstack-ironic16:37
devanandalucasagomes: ya - all we will need is a small .rst file which references each API type & controller16:37
devanandabasically just the layout of the doc16:37
devanandaall the class details, samples, etc, get generated from code16:38
lucasagomesdevananda, I see, yea I have a bullet point on my todo list to write some sample() methods on our controllers16:38
lucasagomesjust like ceilometer does16:39
lucasagomespretty handy :)16:39
devanandalucasagomes: or will. when it isn't generating "list index out of error" and crashing ;)16:39
lucasagomes:P16:39
lucasagomesbut at least such errors we fix straight away, problem in updating documentation is that it's quite boring to keep it updated16:40
lucasagomesso takes a docs are almost always outdated, nobody fix them16:40
devanandaright16:41
*** datajerk has quit IRC16:41
*** prekarat has joined #openstack-ironic16:41
devanandaahhh, coffee at last!16:41
lucasagomes:D enjoy16:43
*** coolsvap has joined #openstack-ironic16:48
*** ndipanov is now known as ndipanov_gone16:49
*** coolsvap has quit IRC16:49
NobodyCambrb16:52
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Rework patch validation on Ports  https://review.openstack.org/5698416:56
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Reworks Node validations  https://review.openstack.org/5749517:04
devanandaah! figured out why it wasnt generating the HTTP samples. needed sphinxcnotrib.httpdomain17:09
NobodyCamand back17:12
*** jistr has quit IRC17:14
openstackgerritDevananda van der Veen proposed a change to openstack/ironic: Add hooks to auto-generate REST API docs  https://review.openstack.org/5734617:26
*** derekh has quit IRC17:27
max_loburping devananda17:36
devanandamax_lobur: pong17:41
max_loburHi deva, I wanted to ask about your comment to https://bugs.launchpad.net/ironic/+bug/124474717:41
max_loburyou said that Ceilometer has taken same approach17:42
devanandaya17:42
max_loburI wanted to use their implementation as example but can't find those hook17:42
max_loburhttps://github.com/openstack/ceilometer/blob/master/ceilometer/api/hooks.py17:42
max_loburhttps://github.com/openstack/ceilometer/blob/master/ceilometer/api/app.py#L58-L6417:43
devanandamax_lobur: IMBW, but look at ceilometer/api/middleware.py17:43
devanandai think that's where the magic happens17:43
max_loburoh, I think we have something similar17:44
max_loburhttps://github.com/openstack/ironic/blob/master/ironic/api/middleware/parsable_error.py17:44
*** idknick has quit IRC17:45
max_loburyea, I see the difference17:46
devanandajbjohnso: ping17:49
*** hemna_ has joined #openstack-ironic17:54
max_loburnow looking to the diff with their version of ParsableErrorMiddleware and seems those new code doesn't resovle our problem. probably they have something else17:55
*** anniec has joined #openstack-ironic18:01
*** harlowja has joined #openstack-ironic18:02
max_loburdevananda, anyway thanks, tomorrow I'll try to dig into their ParsableErrorMiddleware to see what it does18:03
max_loburleaving now18:03
*** max_lobur is now known as max_lobur_afk18:05
*** xmltok has joined #openstack-ironic18:07
lucasagomesmax_lobur_afk, have a g'night :)18:07
lucasagomesI'm also off now18:07
lucasagomessee yall tomorrow! have a good night everyone18:07
dkehnNobodyCam: u still around18:08
NobodyCammaybe18:08
*** lucasagomes has quit IRC18:08
dkehnlol18:08
NobodyCam:)18:08
dkehnNobodyCam: ok, assuming that the dnsmasq structure is there, where in the ironic would be something like nova/compute/manager.py:_build_instance where is would call something to the effect of allocate_networks?18:09
dkehnNobodyCam: I assuming this is going to spawn a conversation18:09
NobodyCamchecking18:11
NobodyCamif I am answering the question you asked it would be some where arounf line 200 of https://review.openstack.org/#/c/51328/2/nova/virt/ironic/driver.py18:15
NobodyCamdevananda: thoughts about neutron and ironic18:16
NobodyCamwhile pxe will need dhcp, other deploy methods may / will require networking plumbing18:17
NobodyCamironic current has no consept of networking18:18
*** ben_duyujie1 has quit IRC18:18
dkehnNobodyCam: in spawn where's the network_info coming from, just curious18:19
NobodyCamhttps://github.com/openstack/nova/blob/master/nova/compute/manager.py#L100618:20
dkehnNobodyCam: ok, got that, so the manager will be calling the ironic from this point18:21
NobodyCamline 1027 of same file calls spawn18:22
dkehnso everything up to that point is going to be as it works today in nova18:22
NobodyCamthen from that we go to ironic nova driver18:22
dkehngotit so essentually the network is provisioned at this point18:23
dkehn#L100618:23
NobodyCamwhich will call ironic functions18:23
NobodyCamactulally you right18:24
dkehnhrm, so when the network's going to get provisioned in the manager the dhcp_options #L1004, that is call ironic pxe.py or  .....18:27
NobodyCamI need to add _allocate_network to the ironic nova driver18:27
dkehnI assuming that drvier is ironic18:27
NobodyCamit's acting on the ironic client18:28
NobodyCamwhich makes calls to to the ironic api18:28
dkehnironic client in this case would be the conductor?18:29
dkehnany pictures out there18:29
NobodyCamhttps://github.com/openstack/python-ironicclient18:29
dkehnI'm a image guy18:30
NobodyCamdevananda: do we have any thing on a slide that showes ironic client to api to conductor flow18:30
devanandaNobodyCam: not yet. want to create one?18:33
NobodyCamlol18:33
dkehncue crickets18:33
NobodyCami can.. /me note that he is not good with image stuff18:34
* devananda raeds up on the conversation18:34
devanandadkehn: i'll draw something real quick18:36
dkehndevananda: thx18:36
NobodyCam:)18:37
dkehnchkconfig --list18:37
dkehnwrong wind sorry18:37
NobodyCamlol18:37
devanandahttps://docs.google.com/drawings/d/16c3g8Rx4y24mTDA76A0gw0ZxrHugeCN1r0zSzohTVKw/edit?usp=sharing18:37
dkehnlove it realtime18:38
*** tatyana has quit IRC18:40
devanandadkehn: see the chat on the doc?18:45
dkehndevananda: I see your drawing if that's what you mean18:46
devanandadkehn: there's a chat there too18:46
dkehnhmm, that I don't see18:47
NobodyCamdkehn: you need to sign in18:47
NobodyCamahh there you go18:47
dkehnok, now I got it18:47
devanandaregarding https://github.com/openstack/nova/blob/master/nova/virt/baremetal/driver.py#L50018:54
NobodyCamso many windows18:54
dkehnthankd u18:54
devanandathe equivalent thing in Ironic's driver is going to be more interesting18:54
devanandaa) nova/virt/ironic/driver.py will need to use the ironic client library to call Ironic's API service18:55
devanandato get some information about the node to which that instance is being associated18:55
devanandathe actual information will be generated by ...18:56
jbjohnsodevananda, ICMP ECHO REPLY18:56
devanandahttps://github.com/openstack/ironic/blob/master/ironic/drivers/modules/pxe.py18:57
dkehnyes18:57
NobodyCamjbjohnso: icmp?18:57
jbjohnsoNobodyCam, he said 'ping'18:57
devanandadkehn: so your current patch should be ported to that file18:57
NobodyCamoh18:57
NobodyCamsorry18:57
devanandadkehn: 50749, i mean18:57
jbjohnsoI am very high latency18:57
dkehnyes, pxe.py -->pxe.py18:57
devanandadkehn: now the plumbing -- how we get the result of ironic.drivers.modules.pxe:dhcp_opts_for_instance() back to Nova, that's going to be a bit more work18:58
dkehnhttps://review.openstack.org/#/c/50749/3/nova/virt/baremetal/pxe.py to the ironic modelues pxe.py, to be clear18:58
jbjohnsoI was actually not paying attention to context of my very poor attempt at humor18:58
jbjohnsobut the context actually kind of makes it funny... kind of...18:59
dkehnwould that be previously set when the driver is initialized18:59
devanandadkehn: we need to somehow expose that information via Ironic's REST API18:59
dkehnouch18:59
devanandadkehn: it looks like none of this information is actually related to the instance itself19:00
devanandadkehn: it will just be the conductor's IP address and the config option 'pxe_bootfile_name'19:00
devanandaromcheg isn't here, but it looks like we need to solve the "what conductor owns this instance" problem NOW()19:01
jbjohnsodevananda, well, it's not that easy right19:01
NobodyCamdevananda: yea19:01
jbjohnsodevananda, well, it could be more fancy than that19:01
jbjohnsofor example:19:01
devanandajbjohnso: I PINGed you regarding: http://www.mail-archive.com/openstack-dev@lists.openstack.org/msg09279.html19:01
devanandajbjohnso: seems like smoething you are better suited to answer than I19:02
dkehnis the conductor the tftp server as well, as least in the simple case19:02
jbjohnsodevananda, ok, let me dig that out of my mailer19:02
devanandadkehn: for the ironic PXE driver, the TFTP server is run on the same host(s) as the ir-cond process, yes19:02
NobodyCamdkehn: yes (for now)19:02
devanandadkehn: the trick is that ironic will have many other drivers -- many of which may not be even remotely PXE based :)19:03
jbjohnsodevananda, for reference19:03
devanandaso our REST API abstracts a lot of these things19:03
jbjohnsohttp://paste.openstack.org/show/53691/19:03
jbjohnsodevananda, that is an example of an xCAT setting up boot entry for xNBA and uefi or bios to start an os19:03
jbjohnsodevananda, that's the output into dhcpd19:03
devanandaawesome19:04
jbjohnsonow we have amibitions in xCAT land19:04
jbjohnsoof ceasing to bother dhcp with that stuff19:04
NobodyCamjbjohnso: where do you ${NEXT-SERVER}19:04
devanandadkehn: eventually we may need to pass things like that to Neutron :)19:04
devanandadkehn: fun stuff, eh19:04
jbjohnsoNobodyCam, that is actually an ipxe variable19:04
dkehndevananda: oh yea19:04
NobodyCam:)19:04
jbjohnsoNobodyCam, in this particuar case, it was inherited by the network definition19:04
jbjohnsoNobodyCam, the intent being that entries without fixed-address can actually float network to network19:05
jbjohnsoNobodyCam, but in practice, no one does that in xcat land19:05
jbjohnsoNobodyCam, actually, the more important aspect is that things can float boot server to boot server19:05
devanandayes19:05
jbjohnsoall the dhcp servers always send 'me' as the value that meanders into ${next-server}19:05
jbjohnsoactually19:06
jbjohnsoneutron only cares about addressing, right?19:06
jbjohnsoit doesn't care about this stuff unless ironic makes it...19:06
devanandaneutron has a dhcp agent19:06
jbjohnsohow much do you know about proxy dhcp?19:06
devanandaand we (ironic) want to be able to reconfigure the dhcp opts on demand19:06
devanandanot enough19:07
jbjohnsoso for at least some things, you can get by without messing with dhcpd19:07
* NobodyCam jumps into devananda's boat19:07
dkehndevananda: when you say passing it to neutron, your talking about the creation as its done today from nova? Its just that is doesn't exist in ironic presently?19:07
jbjohnsoI'm not sure that, for example, rfc 4173 could be done without dhcpd19:07
jbjohnsobut filename and some other stuff can be handled without butting into dhcpd's business19:08
jbjohnsoand as strange as this may sound19:09
dkehndevananda: like https://github.com/openstack/nova/blob/master/nova/network/__init__.py19:09
jbjohnsothe daemon that I can show for console serving I'm toying with making a proxy dhcp server as well19:10
devanandadkehn: for example, if conductor X fails, and ironic needs to migrate control of all the instances formerly  managed by X over to conductor Y19:10
devanandadkehn: then ironic needs to (directly, not via Nova) inform neutron that the dhcp opts for all those instances need to be updated19:11
dkehndevananda: gotit so you will need in ironic a handle to the neutron that is managing the net, currently understood by nova19:11
devanandadkehn: the goal is that one or more ironic-conductor instances will take over for the failed one. without having to re-deploy any nova instances that were managed by it.19:12
devanandaright19:12
dkehndevananda: and the API interface to talk directly19:12
jbjohnsodevananda, so are you thinking RFC4173 type usage, or just pxe boot?19:12
dkehnto neutron19:12
jbjohnsodevananda, or other things delivered by way of dhcp instead of other ways19:13
devanandajbjohnso: boot-from-network-volume is a thing a lot of folks want19:13
devanandajbjohnso: it's not part of the present discussion, but it's definitely on the roadmap, if someone comes along who wants to implement it19:14
jbjohnsodevananda, ok, in that case that would be in the bucket of 'stuff you can do only if you *really* trust the tenant19:14
jbjohnsosince rfc 4173 doesn't play so well with authentication19:14
devanandajbjohnso: and using dhcp to facilitate that is one possible solution. another means is the OOB mgmt interfaces19:14
devanandajbjohnso: many of those support mounting a network volume and booting from it,e ven when no local disks are present19:15
jbjohnsodevananda, right, that is very familiar to me19:15
devanandajbjohnso: eg, folks doing high density clustering want hoards of diskless CPUs booting from a single image19:15
devanandaso Ironic should, ideally, abstract the different means of doing that into a common API19:16
devanandawhether it's rfc4173 or iLO virtual media or what ever19:16
jbjohnsoright, but note that iscsi with authentication isn't something that can safely play into dhcp19:16
devanandasure. but lots of folks are (validly) not worried about security, because they will deploy this only inside oftheir firewall, for a specific workload.19:17
devananda(and ofc other folks are validly VERY worried abotu security)19:17
jbjohnsodiskless cpus booting from a single image is a peculiar one except the form of like a xcat stateless image or particular nfs root solution19:17
devanandaironic needs to communicate clearly which methods are/not secure19:17
devanandaanyhow. we've side tracked a lot19:18
jbjohnsook19:18
devanandadkehn: i feel like we didn't finish that discussion about ironic & dhcp opts in nova. did we?19:18
jbjohnsoI'm realizing you probably meant network block device with clones instead of strictly one image19:18
NobodyCamdevananda: from my point of view ironic really needs to know what conductor (in this case tftp) server is goig to handle the request19:20
NobodyCamtrying not to get side tracked19:21
*** prekarat has quit IRC19:21
devanandaNobodyCam: yea. which isnt represented in our data model at all right now19:21
jbjohnsoNobodyCam, now this is an interesting question... does it really19:21
jbjohnsowell.... that one can get complicated19:21
jbjohnsobut we have a concept of 'whoever offers first gets to handle that boot transaction' in xCAT19:22
NobodyCamjbjohnso: we have until jan 23rd to make it work19:22
jbjohnsoand the relationship with that boot server evaporates post pxe19:22
devanandajbjohnso: we're not setting the boot flag. nodes always PXE boot, even on reboot.19:22
jbjohnsodevananda, right, but19:23
openstackgerritA change was merged to openstack/python-ironicclient: Support building wheels (PEP-427)  https://review.openstack.org/5714219:23
jbjohnsoI meant the relationship is reset every pxe boot19:23
jbjohnsodevananda, if serverA responded one boot19:23
jbjohnsoserverB can respond just as easily19:23
jbjohnsoand basically in a pooled deployment server configuration, it's just racing to see who wins19:23
devanandajbjohnso: you're assuming every server has the same set of tftp images19:23
devanandawhich we are not19:23
jbjohnsodevananda, fyi, the only tftp image we assume are static for every xCAT instance in the world19:24
jbjohnsodevananda, but move that to http space19:24
devanandajbjohnso: also, as NobodyCam just pointed out, we have an actual deadline for the project, and need to focus on implementing it now19:24
jbjohnsoso the protocol changes, but the concept does not.19:24
NobodyCamwe rebuild if the tftp env if a conductor fails19:24
jbjohnsojust wanted to point it pount, might be a non starter if you can't define a one-to-one dhcp to netwboot server anyway19:26
devanandaNobodyCam: so, rebuild the tftp env means we preserve that-there-is-an-association between conductor and instance. we just update that association and trigger a rebuild.19:26
jbjohnsobut it might be rendered an optional thing to worry about eventually19:26
devanandaNobodyCam: and notify neutron19:26
jbjohnsoincidently, if the boot configuration came down throught http, would that alter your design point any?19:27
jbjohnsoe.g. generate the config on demand rather than precreation of a static file?19:27
*** anniec has quit IRC19:32
*** anniec has joined #openstack-ironic19:36
jbjohnsoshort of the mailing list thing, I need to pay attention to pyghmi again and get the sensors, eventlog, and inventory modules done19:38
*** blamar has quit IRC19:41
*** blamar has joined #openstack-ironic19:44
*** xmltok has left #openstack-ironic19:44
*** jdob has quit IRC19:54
*** jdob has joined #openstack-ironic19:54
devanandajbjohnso: there's work that needs to be done on getting sensor data into ironic and fromthere, into ceilometer. if you feel like writing python :)19:56
jbjohnsodevananda, there was one question I had, not having touched an openstack driver19:57
jbjohnsohow long lived can sensors object be?19:57
jbjohnsotrying to figure out if the sdr caching in memory on connection makes sense for the usage here19:57
jbjohnsoalternatively.....19:58
jbjohnsoif the console server also had api to get sensor and inventory data19:58
*** anniec has quit IRC19:59
*** derekh has joined #openstack-ironic20:11
* NobodyCam wanders afk for a few min20:12
*** jdob has quit IRC20:12
*** jdob has joined #openstack-ironic20:13
*** anniec has joined #openstack-ironic20:13
devanandaNobodyCam: https://blueprints.launchpad.net/ironic/+spec/distributed-tftp-for-pxe20:22
devanandalifeless: ^20:22
* NobodyCam looks20:23
*** datajerk has joined #openstack-ironic20:27
NobodyCamGheRivero: also note ^^^^ things may differ from our call this morning20:27
*** anniec has quit IRC20:30
devanandaexample of auto-built API docs: http://docs-draft.openstack.org/46/57346/3/check/gate-ironic-docs/26b188b/doc/build/html/webapi/v1.html20:30
devanandawant lucas to see but he's not here :)20:30
devanandaformatting isnt quite right for the parameters & returns :(20:31
NobodyCambut looking really good20:31
NobodyCam:)20:32
*** datajerk has quit IRC20:32
*** datajerk has joined #openstack-ironic20:35
*** anniec has joined #openstack-ironic20:40
*** epim has joined #openstack-ironic20:41
*** blamar has quit IRC20:44
*** blamar has joined #openstack-ironic20:50
* NobodyCam runs to t-mo store to look at the nexus 5 ... /me really needs to update his sidekick running 2.220:50
*** datajerk has quit IRC21:02
lifelessdevananda: is there an etherpad where we can tune that?21:08
lifelessdevananda: I don't think caching all on all makes sense21:08
devanandalifeless: if we dont cache all-on-all, then we need to model an association between conductors and instances21:10
lifelessdevananda: right. I thought you were happy with 'do nothing for now and focus on nova-bm replacement' ?21:11
lifelessdevananda: I am confused!21:11
devanandalifeless: yes. if we limit the icehouse release of ironic to only support a single (active) conductor21:12
devanandalifeless: then we can take some short cuts, but we still need to model certain things, like the IP of the conductor that "owns" this instance -- even if there's only one conductor, the API still needs to expose that IP21:12
lifelessdevananda: nova bm can run multiple compute processes today.21:12
lifelessdevananda: so that wouldn't be parity.21:12
devanandalifeless: right. well. thus we need this :)21:12
lifelessdevananda: but 'this' is a different thing21:13
devanandanova-bm has a sense of ownership21:13
devanandathis n-cpu "owns" this node21:13
lifelessdevananda: I'm worried that running a large DC of e.g. 10K machines will overwhelm any one conductor21:13
devanandaironic does not have a sense of ownership, and I am trying to avoid adding that21:14
lifelessand we're looking at densities of 900 machines in a single rack21:14
lifelessso 10K is going to be 'medium' size deployment soon.21:14
devanandai know21:14
devanandamodern file systems can handle 10K sym links in a directory21:14
devanandaso having each conductor maintain 10k symlinks to the same K & R pair is not a substantial burden21:15
lifelessfor the case where all the images are identical, yes.21:15
lifelesswhich is an expected common case.21:15
devanandaeven in cases with some heterogeneity, the # of k & r will be $small21:16
lifelessThe worst case is 1:1 with nodes21:16
devanandai'm avoiding all the coplexit of scheduling "this instance goes on taht conductor" -- and the corrolary of "rebalance all these instances off of this conductor"21:17
devanandacomplexity, even.21:17
devanandamy fingers aren't working. i need lunch soon :)21:17
lifelesshow about you get lunc21:18
lifelessthen ping me21:18
lifelessmaybe we can do smoe moderate/high bw discussion21:18
devanandai dont want to introduce an actual scheduling service in ironic. and nova shouldn't know about the ironic-condcutor services or which ones own which instances21:18
lifelessI have a call 2h from now21:18
devanandagreat. i should hopefully be back when you're off that call21:18
devananda4pm my time21:18
lifelessmy call *starts* in 2h21:19
lifeless1230 my time21:19
lifelessshouldn't be a long one.21:19
lifelessanyhow, shoo. Eat. Ping on return:)21:19
devanandalifeless: cal invite sent. bbl :)21:20
*** anniec has quit IRC21:25
*** datajerk has joined #openstack-ironic21:35
NobodyCamnexus 5 looks like a very nice phone. I did not get it..21:37
*** jdob has quit IRC21:45
*** anniec has joined #openstack-ironic21:50
hemna_so I've been trying to get up to speed on Ironic.  looking at the drivers and drivers/modules22:03
hemna_it looks like we do some iscsi operations to login, attach, partition and format the target LUN, then detach22:04
hemna_There is code in Brick (from Cinder) that does iscsi initiator work, device discovery, attaching22:05
hemna_as well as Fibre Channel devices22:05
hemna_I'm going to refactor Nova's low level iscsi work to use Brick22:05
hemna_so it might be worth while to use Brick in Ironic as well22:06
NobodyCamhemna_: that is used to deploy (DD) the user requested image on to the baremetal node22:06
NobodyCamhemna_: replace this? https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/deploy_utils.py#L38-L7022:10
*** jbjohnso has quit IRC22:13
*** ekarlso has quit IRC22:15
*** ekarlso has joined #openstack-ironic22:17
NobodyCamwalkies time22:21
*** datajerk has quit IRC22:23
*** datajerk has joined #openstack-ironic22:32
hemna_NobodyCam, yah exactly22:45
hemna_we are working on extracting Brick from Cinder into it's own library22:46
hemna_to help eliminate the duplicate code in Nova and Cinder for iscsi/fc attaches22:46
openstackgerritlifeless proposed a change to openstack/ironic: PEP8-ify imports in test_ipmitool.  https://review.openstack.org/5755022:56
*** datajerk has quit IRC23:00
*** epim has quit IRC23:04
*** datajerk has joined #openstack-ironic23:06
openstackgerritHaomeng,Wang proposed a change to openstack/python-ironicclient: Enable created_at/updated_at for port-show/chassis-show  https://review.openstack.org/5399123:12
*** michchap_ has quit IRC23:15
NobodyCamhemna_: that would be cool23:16
NobodyCamgah just found a flaw in the ironic yaml in triple-heat-templates...23:17
*** michchap has joined #openstack-ironic23:18
hemna_I'm basically trying to learn how Ironic works and see how we can integrate with Cinder.23:21
*** datajerk has quit IRC23:22
hemna_see how Ironic can coordinate with Cinder for volume attaches for boot from volume.  I'd like to see baremetal running on root volumes provisioned by Cinder23:22
hemna_a la diskless servers23:22
NobodyCamboot from vol will require a driver other then the pxe driver23:23
NobodyCamthis has yet to be written23:23
hemna_looks like we'll need to somehow tell the bios on the server that it needs to boot from a SAN volume iscsi/FC23:24
hemna_still trying to wrap my head around it all23:24
hemna_learning PXE in the process23:24
*** datajerk has joined #openstack-ironic23:25
NobodyCampxe is just one way to remote boot.23:25
*** anniec has quit IRC23:34
*** anniec has joined #openstack-ironic23:41
Haomenghemna_: I understand PXE uses dhcp and tftp to download PXE bootloader into ram and start sys from bootloader, then enable networking and do other things whatever23:45
hemna_Haomeng, that's what I'm beginning to understand.  What I'm trying to figure out is how an iSCSI/FC/SAN volume can be presented as the root volume for the install of the image23:47
hemna_and then how to coordinate that with an existing non attached Cinder volume.23:47
hemna_There needs to be some coordination from Ironic to Cinder to tell Cinder to attach that volume to the host.23:47
Haomengcool23:48
hemna_Cinder will need to call the backend driver's initialize_connection, to present the LUN to the host.23:48
Haomenghemna_: we want to do boot from storage volume, right?23:48
hemna_and the host will have to do what it can do iscsi/fc attach23:48
hemna_Haomeng, yup23:48
hemna_that's the idea23:49
*** datajerk has quit IRC23:49
hemna_I'm trying to climb the ironic learning curve for now to see how/what we need to do in Cinder to enable that23:49
hemna_I wrote the Brick Initiator code in Cinder and am going to refactor Nova to use Brick, once Brick is a standalone lib.23:49
Haomenghemna_: but I think we depends on deferent hardware system, because the sys in volumn have to match the hardware, such arch23:49
hemna_so isn't the PXE image the OS that will be installed onto the root volume ?23:50
Haomenghemna_: maybe we can do for x86 arch first23:50
Haomenghemna_: will be back...23:51
*** datajerk has joined #openstack-ironic23:53
*** datajerk has quit IRC23:56
devanandalifeless: i'll be around in ~10 min23:58
« Tuesday, 2013-11-19 Index Thursday, 2013-11-21 »

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!