pooja | TheJulia: Doesn't ironic calculate the root partition size though? I am only setting the disk_gb property of the node | 00:01 |
---|---|---|
pooja | Any other attributes that need to be set so that space is reserved for boot loader, swap and a config drive? | 00:02 |
TheJulia | Oh... hmmm | 00:03 |
TheJulia | pooja: I guess your manually triggering deployments? | 00:06 |
*** zhangfei has joined #openstack-ironic | 00:08 | |
*** thegreenhundred has joined #openstack-ironic | 00:14 | |
*** ijw has joined #openstack-ironic | 00:15 | |
*** sasha2 has quit IRC | 00:16 | |
*** ijw has quit IRC | 00:19 | |
*** sasha2 has joined #openstack-ironic | 00:20 | |
*** hw_wutianwei has joined #openstack-ironic | 00:35 | |
*** sasha2 has quit IRC | 00:37 | |
*** baoli has joined #openstack-ironic | 00:46 | |
*** baoli has quit IRC | 00:47 | |
*** baoli has joined #openstack-ironic | 00:48 | |
*** rama_y has quit IRC | 00:48 | |
pooja | TheJulia: Sorry, got pulled in a meeting earlier! | 00:52 |
pooja | No, I am triggered deploy using nova and have setup the flavor and baremetal node with disk_gb as 117Gb. | 00:52 |
pooja | * triggering baremetal node deploy | 00:52 |
*** yuanying has joined #openstack-ironic | 01:00 | |
*** phuongnh has joined #openstack-ironic | 01:02 | |
*** TuanLA has joined #openstack-ironic | 01:04 | |
*** yuanying_ has quit IRC | 01:04 | |
*** yuanying_ has joined #openstack-ironic | 01:04 | |
*** yuanying has quit IRC | 01:05 | |
*** mat128 has joined #openstack-ironic | 01:07 | |
*** ijw has joined #openstack-ironic | 01:07 | |
*** yangyapeng has joined #openstack-ironic | 01:14 | |
*** ijw has quit IRC | 01:14 | |
*** ijw has joined #openstack-ironic | 01:15 | |
wong | hi,may I ask a question? | 01:16 |
wong | may I configure tenant network by using horizon dashboard? | 01:17 |
wong | I want to make ironic nodes to serve tenant | 01:19 |
*** ijw has quit IRC | 01:20 | |
*** syqian has joined #openstack-ironic | 01:30 | |
TheJulia | pooja: if it is a partition image, Swap would also go into that if defined, at least if memory serves, so that needs to be accounted for potentially. | 01:33 |
*** yangyapeng has quit IRC | 01:34 | |
TheJulia | wong: with neutron networking, as in the network_interface set to "neutron" on the node, I don't see why not if the users are creating ports in neutron, and then attaching them to instances they are deploying via nova. | 01:35 |
*** yangyapeng has joined #openstack-ironic | 01:35 | |
pooja | Oh okay, i couldn't find documentation around how to specify swap_mb as such, but came across root device hints. It also says IPA by default selects smallest disk and not necessarily the first disk for partitioning? | 01:35 |
pooja | I am now trying to use add properties/root_device='{"size": ">= 110"} to see if that helps and also reducing MiB value by 1. | 01:36 |
TheJulia | yeah that might help, root_device hints can definitely help | 01:38 |
TheJulia | at least, then you can make sure if you have multiple devices in the mix | 01:39 |
*** mat128 has quit IRC | 01:41 | |
wong | TheJulia:you mean we can just enroll the nodes first,then user can create ports and attach them to instance? but the node can't access TFTP after they are deployed | 01:54 |
TheJulia | wong: and yes, that is because the conductor will move the node to a provisioning network for the deployment to take place | 01:55 |
*** yangyape_ has joined #openstack-ironic | 01:57 | |
*** yangyapeng has quit IRC | 01:57 | |
*** yuanying_ has quit IRC | 02:01 | |
*** litao__ has joined #openstack-ironic | 02:02 | |
*** yuanying has joined #openstack-ironic | 02:04 | |
*** sasha2 has joined #openstack-ironic | 02:18 | |
wong | tks | 02:20 |
*** syqian has quit IRC | 02:25 | |
*** syqian has joined #openstack-ironic | 02:25 | |
*** yangyape_ has quit IRC | 02:28 | |
*** yangyapeng has joined #openstack-ironic | 02:28 | |
pooja | TheJulia: Tried using "size": ">= 110", but nova complains about disk size not being an integer, possibly this (>=) was not supported in Newton? | 02:31 |
pooja | Anyway setting the Gib value to 110 worked for me. Thanks! | 02:31 |
wong | TheJulia:when I create an instance by using horizon dashboard,whther the ironic conductor can finish all things? include deploying and changing network ? | 02:33 |
pooja | 120Gb is 111.7 Gib, so setting local_gb to 110Gib worked. I guess it wasn't clear that local_gb should be set to Gib value - 1 (not Gb -1). | 02:34 |
*** namnh has joined #openstack-ironic | 02:35 | |
wong | and the network between deploying and tenant network is separated | 02:35 |
*** mat128 has joined #openstack-ironic | 02:40 | |
*** zhangfei has quit IRC | 02:41 | |
wong | pooja:once I set local_gb 500G and my disk is 500G,it's failed also. | 02:43 |
*** pooja has quit IRC | 02:50 | |
*** ijw has joined #openstack-ironic | 02:56 | |
wong | it seems that the disk must be larger than local_gb | 02:57 |
*** zhangfei has joined #openstack-ironic | 02:58 | |
openstackgerrit | Ruby Loo proposed openstack/python-ironicclient stable/ocata: DNM testing devstack change in ironic https://review.openstack.org/499391 | 03:00 |
*** zshi_ has joined #openstack-ironic | 03:01 | |
*** ijw has quit IRC | 03:01 | |
TheJulia | Pooja: I thought it was :(. | 03:02 |
TheJulia | Oh, they left :( | 03:02 |
TheJulia | wong: that is what the conductor does, orchestrate deployment including changing network in that process | 03:03 |
wong | TheJulia:I know it now, tks a lot | 03:08 |
*** tiendc has joined #openstack-ironic | 03:09 | |
wong | TheJulia;it's too late in your place,can't you have a sleep? | 03:09 |
*** zshi_ has quit IRC | 03:10 | |
TheJulia | wong: I'm actually heading to sleep now | 03:10 |
TheJulia | Have a wonderful day! | 03:10 |
wong | :) have a good night | 03:10 |
*** mat128 has quit IRC | 03:12 | |
*** zshi_ has joined #openstack-ironic | 03:23 | |
*** rama_y has joined #openstack-ironic | 03:28 | |
*** rama_y has quit IRC | 03:29 | |
*** rama_y has joined #openstack-ironic | 03:29 | |
*** rbudden has quit IRC | 03:34 | |
*** pooja has joined #openstack-ironic | 03:36 | |
pooja | wong: Yes, I think local_gb value should be set to (Gib-1) and not Gb. Maybe we should make that clear in the documentation. | 03:37 |
*** udesale has joined #openstack-ironic | 03:37 | |
*** links has joined #openstack-ironic | 03:39 | |
wong | pooja: you are right | 03:59 |
openstackgerrit | Shivanand Tendulker proposed openstack/ironic master: Boot from volume fails with 'iscsi' deploy interface https://review.openstack.org/499446 | 04:02 |
*** mat128 has joined #openstack-ironic | 04:09 | |
*** mat128 has quit IRC | 04:09 | |
*** jaianshu has joined #openstack-ironic | 04:11 | |
*** vmud213 has joined #openstack-ironic | 04:12 | |
*** ijw has joined #openstack-ironic | 04:16 | |
*** ijw has quit IRC | 04:21 | |
*** rama_y has quit IRC | 04:21 | |
*** baoli has quit IRC | 04:23 | |
*** baoli has joined #openstack-ironic | 04:37 | |
*** jlabarre has quit IRC | 04:41 | |
*** sasha2 has quit IRC | 04:44 | |
*** MattMan_1 has quit IRC | 04:49 | |
*** baoli has quit IRC | 04:49 | |
*** MattMan_1 has joined #openstack-ironic | 04:49 | |
openstackgerrit | Merged openstack/ironic master: Follow up Secure boot support for irmc-pxe driver. https://review.openstack.org/493394 | 04:50 |
*** phuongnh has quit IRC | 04:52 | |
*** syqian has quit IRC | 04:52 | |
*** phuongnh has joined #openstack-ironic | 04:53 | |
*** syqian has joined #openstack-ironic | 04:53 | |
*** rbartal has joined #openstack-ironic | 05:07 | |
*** ianychoi__ has joined #openstack-ironic | 05:12 | |
*** ianychoi_ has quit IRC | 05:16 | |
*** oomichi has quit IRC | 05:28 | |
*** oomichi has joined #openstack-ironic | 05:30 | |
*** rbartal has quit IRC | 05:31 | |
openstackgerrit | Shivanand Tendulker proposed openstack/ironic master: Fix ``agent`` deploy interface to call ``boot.prepare_instance`` https://review.openstack.org/499050 | 05:35 |
*** yangyapeng has quit IRC | 05:42 | |
*** yangyapeng has joined #openstack-ironic | 05:43 | |
openstackgerrit | Shivanand Tendulker proposed openstack/ironic master: Fix ``agent`` deploy interface to call ``boot.prepare_instance`` https://review.openstack.org/499050 | 05:52 |
*** marios has joined #openstack-ironic | 05:56 | |
*** mjura has joined #openstack-ironic | 06:00 | |
wong | hi,ironic as a hypervisor can be coexist with other hypervisor types? | 06:02 |
*** snswin has quit IRC | 06:03 | |
*** snswin has joined #openstack-ironic | 06:03 | |
*** pooja has quit IRC | 06:12 | |
*** links has quit IRC | 06:15 | |
*** parlos has joined #openstack-ironic | 06:15 | |
syqian | wong, what are you mean? other hypervisor such as kvm, xen...? | 06:17 |
wong | yeah | 06:20 |
*** pcaruana has joined #openstack-ironic | 06:22 | |
*** ntpttr_laptop has quit IRC | 06:22 | |
wong | if they can be coexist in single nova service,I don't need support another nova service | 06:22 |
*** ntpttr_laptop has joined #openstack-ironic | 06:22 | |
*** ntpttr_laptop__ has joined #openstack-ironic | 06:23 | |
*** ntpttr_laptop has quit IRC | 06:23 | |
syqian | wong, this may be cann't. you should support a single nova-compute service for ironic. | 06:24 |
wong | sure? | 06:24 |
*** marios has quit IRC | 06:25 | |
syqian | wong, i think it is. in the nova.conf you can config compute_driver=xxxx to select which driver you ues. | 06:27 |
wong | and only choose one | 06:27 |
syqian | wong, yes, it seems you can only choose one driver. | 06:28 |
wong | I know, tks a lot | 06:28 |
*** mjura has quit IRC | 06:28 | |
syqian | wong, no prob, by the way, it's my opinion, you can search it deeper. | 06:30 |
wong | OK | 06:31 |
*** vsaienk0 has joined #openstack-ironic | 06:33 | |
*** hshiina has joined #openstack-ironic | 06:35 | |
*** jtomasek has quit IRC | 06:36 | |
*** jtomasek has joined #openstack-ironic | 06:37 | |
*** jtomasek has quit IRC | 06:39 | |
*** jtomasek has joined #openstack-ironic | 06:40 | |
*** syqian has quit IRC | 06:40 | |
*** rcernin has joined #openstack-ironic | 06:40 | |
*** stendulker has joined #openstack-ironic | 06:40 | |
wong | another thing, Nova service can have multiple ironic-host-manager? | 06:41 |
*** mjura has joined #openstack-ironic | 06:41 | |
*** syqian has joined #openstack-ironic | 06:41 | |
*** livelace has joined #openstack-ironic | 06:45 | |
*** ntpttr_laptop__ has quit IRC | 06:51 | |
*** phuongnh has quit IRC | 06:57 | |
*** phuongnh has joined #openstack-ironic | 06:57 | |
*** hjensas has joined #openstack-ironic | 06:59 | |
*** hjensas has quit IRC | 06:59 | |
*** hjensas has joined #openstack-ironic | 06:59 | |
*** charcol has quit IRC | 07:00 | |
*** zhangfei has quit IRC | 07:03 | |
*** syqian_ has joined #openstack-ironic | 07:04 | |
*** hoonetorg has quit IRC | 07:04 | |
*** syqian has quit IRC | 07:07 | |
*** jaganathan has joined #openstack-ironic | 07:11 | |
*** zhangfei has joined #openstack-ironic | 07:11 | |
*** hoonetorg has joined #openstack-ironic | 07:17 | |
*** racedo has joined #openstack-ironic | 07:28 | |
*** links has joined #openstack-ironic | 07:29 | |
*** tesseract has joined #openstack-ironic | 07:32 | |
*** rbartal has joined #openstack-ironic | 07:38 | |
*** hjensas has quit IRC | 07:38 | |
*** syqian__ has joined #openstack-ironic | 07:41 | |
*** hjensas has joined #openstack-ironic | 07:43 | |
*** hjensas has quit IRC | 07:43 | |
*** hjensas has joined #openstack-ironic | 07:43 | |
*** oomichi has quit IRC | 07:44 | |
*** oomichi has joined #openstack-ironic | 07:44 | |
*** syqian_ has quit IRC | 07:45 | |
openstackgerrit | yolanda.robla proposed openstack/bifrost stable/pike: Fix bug in ironic policy https://review.openstack.org/499490 | 07:48 |
openstackgerrit | yolanda.robla proposed openstack/bifrost stable/ocata: Fix bug in ironic policy https://review.openstack.org/499491 | 07:49 |
vdrok | good morning ironic! | 08:01 |
vdrok | wong: yes, if you're using nodes without resource classes, better to use ironic_host_manager | 08:02 |
vdrok | or do you mean multiple host managers , like default one and ironic? | 08:02 |
vdrok | if so, than it's not possible. with mixed hypervisor case you should use the standard host manager | 08:02 |
openstackgerrit | yolanda.robla proposed openstack/bifrost master: DNM: testing centos https://review.openstack.org/499498 | 08:11 |
*** mgoddard has joined #openstack-ironic | 08:14 | |
etingof | morning again \o/ | 08:15 |
*** marios has joined #openstack-ironic | 08:16 | |
*** openstackgerrit has quit IRC | 08:17 | |
*** marios_ has joined #openstack-ironic | 08:19 | |
*** ralonsoh has joined #openstack-ironic | 08:23 | |
*** lucas-afk is now known as lucasagomes | 08:30 | |
wong | OK | 08:31 |
*** Nisha_Agarwal has joined #openstack-ironic | 08:35 | |
Nisha_Agarwal | Hey Ironic!!! | 08:35 |
Nisha_Agarwal | Good morning/afternoon | 08:35 |
vdrok | good morning etingof and Nisha_Agarwal | 08:37 |
Nisha_Agarwal | hey vdrok etingof | 08:37 |
Nisha_Agarwal | vdrok, i was looking ironic-inspector code. dtantsur|afk mentioned in past that inspector automates the creation/assigning of resource class name to the nodes. If you have any idea could you point me to the code where it does that. | 08:40 |
*** syqian__ has quit IRC | 08:41 | |
*** syqian__ has joined #openstack-ironic | 08:42 | |
vdrok | Nisha_Agarwal: I don't see resource_class in inspector code after a quick search, you sure it's already done? | 08:43 |
*** karimb has joined #openstack-ironic | 08:45 | |
Nisha_Agarwal | vdrok, thats what dtantsur|afk mentioned in his mail http://lists.openstack.org/pipermail/openstack-dev/2017-April/115278.html | 08:46 |
Nisha_Agarwal | vdrok, i was just trying to see how it does that and can we use that in ironic also... | 08:47 |
Nisha_Agarwal | vdrok, he says this "Also ironic-inspector is useful for automatically defining resource classes on | 08:47 |
Nisha_Agarwal | nodes, so I'm not sure about this purpose being defeated as well." | 08:47 |
vdrok | what's written there seem to indicate that it's possible, not that it's currently being done | 08:48 |
vdrok | aarefiev: ^^ | 08:48 |
*** openstackgerrit has joined #openstack-ironic | 08:49 | |
openstackgerrit | yolanda.robla proposed openstack/bifrost master: Fix the dhcp checks on centos https://review.openstack.org/499498 | 08:49 |
aarefiev | Nisha_Agarwal: there is no way to assign resource class automatically | 08:51 |
pas-ha | morning ironic :) | 08:51 |
aarefiev | Nisha_Agarwal: you could try inspector rules to assign it based on discovered properties | 08:52 |
*** dtantsur|travel has joined #openstack-ironic | 08:52 | |
dtantsur|travel | morning ironic | 08:52 |
Nisha_Agarwal | aarefiev, could you help me with some example | 08:52 |
Nisha_Agarwal | dtantsur|travel, hi | 08:52 |
dtantsur|travel | I'm unexpectedly traveling today, so I will be on and offline | 08:52 |
dtantsur|travel | will try to get a stable connection in the evening | 08:53 |
vdrok | good morning dtantsur|afk | 08:57 |
vdrok | or dtantsur|travel ? :) | 08:57 |
aarefiev | Nisha_Agarwal: let me check that we have appropriate `action`, we are going to discuss this on ptg more | 08:57 |
dtantsur|travel | the latter for now :) I'm outside of VPN, so not connecting to my bouncer | 08:57 |
aarefiev | morning dtantsur|travel :) | 08:57 |
*** mrch has joined #openstack-ironic | 09:02 | |
aarefiev | Nisha_Agarwal: something like http://paste.openstack.org/show/620002/ | 09:02 |
aarefiev | I'm not sure if it would work from the box | 09:02 |
*** openstackgerrit has quit IRC | 09:03 | |
dtantsur|travel | looks good | 09:03 |
Nisha_Agarwal | aarefiev, ohk. that means it cannot be used by ironic | 09:03 |
dtantsur|travel | actually, I was planning to provide a few examples in ironic-inspector docs | 09:03 |
dtantsur|travel | mmm, yeah. it's inspector. | 09:03 |
aarefiev | I guess it may need to hack ironic api version in inspector | 09:04 |
Nisha_Agarwal | aarefiev, yeah u can do it in inspector. but for OOB inspection the logic needs to sit inside ironic | 09:04 |
Nisha_Agarwal | aarefiev, thanks for the example. atleast i know what dtantsur|travel meant in past | 09:05 |
aarefiev | yeah, we need some logic for this in ironic | 09:05 |
aarefiev | like rules | 09:06 |
Nisha_Agarwal | aarefiev, what i feel is that it is similar to node-update | 09:06 |
Nisha_Agarwal | the value for resource class is still coming from user | 09:07 |
Nisha_Agarwal | i was thinking inspector automates the resource class name based on the inspected values | 09:07 |
Nisha_Agarwal | dtantsur|travel, i was looking at nova code for this. I feel if we allow resource class structure at ironic nodes, probably we can schedule based on standard resource class rather than skipping them | 09:09 |
aarefiev | yeah, the goal here is to make it automaticaly, so user may specify some resource classe based on inspected data | 09:09 |
*** e0ne has joined #openstack-ironic | 09:09 | |
Nisha_Agarwal | aarefiev, generally we dont know the resources before inspection | 09:10 |
Nisha_Agarwal | oh i got what u meant | 09:11 |
aarefiev | Nisha_Agarwal: why? | 09:11 |
sambetts|afk | Nisha_Agarwal: you should have the idea about the falvors you want to provide to your users | 09:11 |
*** sambetts|afk is now known as sambetts | 09:11 | |
Nisha_Agarwal | sambetts, i misunderstood aarefiev's example | 09:12 |
Nisha_Agarwal | sambetts, aarefiev now i know what he meant | 09:12 |
sambetts | Nisha_Agarwal: resource classes map directly to flavors, so a "large-baremetal" flavor should have a certain expected set of resources | 09:13 |
Nisha_Agarwal | sambetts, yes. | 09:13 |
sambetts | you can use inspector rules to group machines with ram >= certain amount, disk >= certain amount, and cpus >= certain amount, into that resource class | 09:14 |
Nisha_Agarwal | sambetts, aarefiev dtantsur|travel i still dont understand why ironic shouldnt be scheduled based on standard resources | 09:15 |
Nisha_Agarwal | sambetts, i understand that with the given example | 09:15 |
*** milan has joined #openstack-ironic | 09:18 | |
sambetts | Nisha_Agarwal: there are a lot of reasons, there are several specs covering the design, but the main reason is because an ironic node is indivisible, when you ask for a baremetal node you get the whole node | 09:18 |
milan | morning Ironic! :) | 09:18 |
sambetts | Nisha_Agarwal: so it made sense to have flavors that ask for 1 of a type of baremetal node | 09:18 |
sambetts | e.g. 1 large baremetal node | 09:19 |
sambetts | or 1 baremetal node with GPUs | 09:19 |
milan | #pixiesay -mflexing Yawn! | 09:19 |
PixieBoots | ᕙʕ⇀ᴥ⇀ʔᕗ: Yawn! | 09:19 |
dtantsur|travel | Nisha_Agarwal: please find a mail from me to the ML for explanation of why standard properties don't really work for ironic | 09:20 |
Nisha_Agarwal | sambetts, i have read the specs in nova already. | 09:20 |
* dtantsur|travel does not have time right now, sorry | 09:20 | |
Nisha_Agarwal | dtantsur|travel, i read that | 09:20 |
Nisha_Agarwal | sambetts, dtantsur|travel i am not able to put my point across , but i think ironic should have resource class structure at ironic nodes so as to allow scheduling based on standard resources | 09:23 |
sambetts | Nisha_Agarwal: so an ironic node can report multiple resource classes with values? | 09:23 |
Nisha_Agarwal | sambetts, yes | 09:23 |
sambetts | I agree, I think that would be useful, even if its not for schduling based on standard resources, but just for adding additonal information about what a baremetal node has e.g. number of GPUs etc | 09:25 |
vdrok | morning sambetts and milan | 09:25 |
milan | morning vdrok! :) | 09:26 |
milan | also morning dtantsur|travel Nisha_Agarwal and sambetts! :) | 09:26 |
vdrok | sambetts: if it's not needed for scheduling why do these additional resource classes? | 09:26 |
* milan 's coffee is gone but his brain still loading | 09:26 | |
sambetts | vdrok: it is for schdulaing, but not for standard resources (ram disk cpu) | 09:26 |
vdrok | ok, but still everything is consumed, so why not have resource class baremetal_large_with_5_gpus | 09:27 |
sambetts | vdrok: so you could have a flavor that asks for 1 large-baremetal, that has at least 2 gpus for example, and the standard ironic resource classes matches the 1 large-baremetal and consumes the whole node, but the GPUs will match as long as that node has at least 2 avaible | 09:28 |
dtantsur|travel | yep, the problem with several resource classes is that you cannot consume half of them | 09:28 |
stendulker | sambetts: Did you mean a single baremetal associated with multiple resource classes? | 09:28 |
dtantsur|travel | and we get back to the problem of double assignment and ugly hacks in the ironic virt driver | 09:28 |
Nisha_Agarwal | dtantsur|travel, yep we cannot consume half of them. but we can consume them fully in the ironic-virt-driver as we do for standrad resource classes today | 09:29 |
sambetts | Nisha_Agarwal: that is the exact hack we're trying to kill with resource classes | 09:29 |
vdrok | Nisha_Agarwal: it gives us lots of headache | 09:29 |
dtantsur|travel | Nisha_Agarwal: no, we cannot. | 09:31 |
Nisha_Agarwal | mmmm | 09:31 |
dtantsur|travel | ironic virt driver is not the place to track resources. what we do now is ugly and buggy, and is already half broken in Pike | 09:31 |
sambetts | Nisha_Agarwal: that hack is what makes us different and special when it comes to schdulaing in nova, when we switch to resource classes as proposed today all that horrible code goes away and we don't need custom host manager logic or custom filters to make ironic work | 09:31 |
dtantsur|travel | Nisha_Agarwal, sambetts, vdrok, I'd say this conversation is not overly helpful without nova folks involved, e.g. jaypipes or dansmith | 09:32 |
* dtantsur|travel runs to the bus, sorry | 09:32 | |
*** dtantsur|travel has quit IRC | 09:32 | |
* Nisha_Agarwal also needs to run right now. but it would help we can discuss this along with nova guys. may be 6hrs from now.(would be around 9 pm my time) | 09:33 | |
sambetts | personally I don't like resource classes as they are implmented today, but we're in a transition period, flavors as they exist today will be going away, and resource classes will become first class citizens, and I think at that point it'll make much more sense that we're doing it this way | 09:34 |
Nisha_Agarwal | sambetts, :) | 09:34 |
Nisha_Agarwal | sambetts, vdrok should we discuss this around 9 pm my time(~6 hrs from now) | 09:35 |
Nisha_Agarwal | with the nova guys also involved | 09:35 |
*** hoonetorg has quit IRC | 09:36 | |
vdrok | well, you can just come and ask the questions :) | 09:36 |
Nisha_Agarwal | vdrok, :) | 09:36 |
Nisha_Agarwal | thanks | 09:36 |
* Nisha_Agarwal need to log off for some time now. | 09:37 | |
*** Nisha_Agarwal has quit IRC | 09:41 | |
wong | hi,if the node is removed to the tenant network,wether the ironic manager can connect it? | 09:46 |
*** etingof has quit IRC | 09:46 | |
*** hw_wutianwei has quit IRC | 09:47 | |
sambetts | wong: connect to it how? regardless of which data network a node is connected to ironic should always be able to talk to the out of band managment (IPMI) for a node | 09:48 |
*** openstackgerrit has joined #openstack-ironic | 09:53 | |
openstackgerrit | Markos Chandras (hwoarang) proposed openstack/bifrost master: Test OPNFV CI verify job. Please ignore. https://review.openstack.org/367608 | 09:53 |
wong | I understand it | 09:53 |
*** hoonetorg has joined #openstack-ironic | 09:53 | |
*** ralonsoh has quit IRC | 10:03 | |
*** ralonsoh has joined #openstack-ironic | 10:04 | |
*** priteau has joined #openstack-ironic | 10:04 | |
*** yangyapeng has quit IRC | 10:05 | |
*** yangyapeng has joined #openstack-ironic | 10:05 | |
*** rajame has joined #openstack-ironic | 10:06 | |
openstackgerrit | Merged openstack/ironic master: [doc] add FAQ about updating release notes https://review.openstack.org/497708 | 10:08 |
*** yangyapeng has quit IRC | 10:10 | |
vdrok | sambetts: https://review.openstack.org/499545 here is why reschedule not happening in that test | 10:12 |
patchbot | patch 499545 - nova - Allow reschedules for ironic computes if one force... | 10:12 |
yolanda | TheJulia, or other cores, fix for bifrost in centos... https://review.openstack.org/499498 | 10:14 |
patchbot | patch 499498 - bifrost - Fix the dhcp checks on centos | 10:14 |
sambetts | vdrok: ewww... huge -1 to adding ironic hypervisor specific code outside of the ironic driver :/ | 10:15 |
vdrok | yeah I know | 10:15 |
vdrok | the problem is, it's scheduler, and this part does not seem to be customizable at all | 10:16 |
vdrok | s/scheduler/conductor | 10:17 |
*** etingof has joined #openstack-ironic | 10:20 | |
*** etingof has quit IRC | 10:25 | |
*** wong has quit IRC | 10:25 | |
*** etingof has joined #openstack-ironic | 10:26 | |
*** udesale has quit IRC | 10:27 | |
*** karimb has quit IRC | 10:28 | |
*** priteau has quit IRC | 10:28 | |
*** amarao has quit IRC | 10:31 | |
*** priteau has joined #openstack-ironic | 10:31 | |
vdrok | sambetts: this is not ironic specific https://review.openstack.org/#/c/499545/2/nova/scheduler/utils.py, but still ugly, and not sure if eg for vmware it's possible to run hypervisor on the same host as n-cpu. if it is, this logic is wrong too :( | 10:33 |
patchbot | patch 499545 - nova - Allow reschedules for ironic computes if one force... | 10:33 |
pas-ha | dtantsur|afk: vdrok some small fixes to staging drivers stable branches https://review.openstack.org/#/q/I537a731df702273d0184b02fa7532eaff8a167ba,n,z | 10:37 |
pas-ha | the pike one is actually quite needed for unit tests on gates (install ironic as of pike, not from master). not broken now, but may break any time in the future | 10:38 |
pas-ha | others are just some more convenience for developers | 10:38 |
*** amarao has joined #openstack-ironic | 10:39 | |
*** karimb has joined #openstack-ironic | 10:39 | |
*** jkilpatr has quit IRC | 10:41 | |
stendulker | sambetts: Hi | 10:44 |
sambetts | Hi | 10:44 |
stendulker | sambetts: regarding patch https://review.openstack.org/#/c/499050 | 10:44 |
patchbot | patch 499050 - ironic - Fix ``agent`` deploy interface to call ``boot.prep... | 10:44 |
stendulker | This is regarding your comment to move the boot device setting code from agent.py:reboot_to_instance() | 10:45 |
stendulker | It can break the 3rd party vendor drivers if they were relying on it | 10:46 |
stendulker | If we keep it, we would be setting the boot device twice, in worst case | 10:46 |
stendulker | yuriy also had similar opinion | 10:46 |
stendulker | sambetts: what you think? | 10:46 |
sambetts | stendulker: actually I think if we leave it in there we'll end up setting the boot device 3 times, once in reboot_to_instance, once in prepare_instance_to_boot, and once in prepare_instance | 10:48 |
sambetts | making calls to the baremetal can be slow | 10:48 |
stendulker | oh yes, it would thrice.. | 10:48 |
sambetts | so reducing unnessary requests to the BMC would be ideal | 10:48 |
sambetts | IMO the boot interface should be the interface that is handling setting the boot device, thats its purpose | 10:48 |
stendulker | i agree, it should be done only in the boot interface | 10:49 |
stendulker | only for legacy reasons this code is lying in agent.py | 10:49 |
stendulker | all our in-tree drivers do it properly | 10:49 |
stendulker | only question is about out of tree drivers | 10:50 |
*** rajame has quit IRC | 10:50 | |
stendulker | sambetts, i will keep it as it for now. | 10:51 |
stendulker | sambetts: thank you. | 10:51 |
sambetts | np | 10:51 |
*** zhangfei has quit IRC | 10:54 | |
openstackgerrit | Andrey Shestakov proposed openstack/ironic master: Change pxe dhcp options name to codes. https://review.openstack.org/499553 | 10:59 |
*** jkilpatr has joined #openstack-ironic | 11:01 | |
*** namnh has quit IRC | 11:01 | |
*** TuanLA has quit IRC | 11:02 | |
*** jaianshu has quit IRC | 11:04 | |
*** parlos has quit IRC | 11:06 | |
*** parlos has joined #openstack-ironic | 11:12 | |
*** jistr is now known as jistr|biab | 11:12 | |
*** phuongnh has quit IRC | 11:14 | |
*** stendulker_ has joined #openstack-ironic | 11:25 | |
*** stendulker has quit IRC | 11:26 | |
*** rbudden has joined #openstack-ironic | 11:28 | |
*** stendulker_ has quit IRC | 11:30 | |
*** vmud213 has quit IRC | 11:35 | |
*** mat128 has joined #openstack-ironic | 11:37 | |
*** lucasagomes is now known as lucas-hungry | 11:44 | |
*** bfournie has quit IRC | 11:46 | |
*** bfournie has joined #openstack-ironic | 11:47 | |
*** efried_off is now known as efried | 11:50 | |
*** bfournie has quit IRC | 11:51 | |
*** thegreenhundred has quit IRC | 11:55 | |
*** litao__ has quit IRC | 11:55 | |
*** jkilpatr has quit IRC | 12:03 | |
*** TuanLA has joined #openstack-ironic | 12:07 | |
*** karimb has quit IRC | 12:07 | |
*** jkilpatr has joined #openstack-ironic | 12:09 | |
*** jkilpatr has quit IRC | 12:12 | |
*** jkilpatr has joined #openstack-ironic | 12:12 | |
*** jlabarre has joined #openstack-ironic | 12:12 | |
*** jkilpatr has quit IRC | 12:17 | |
*** fellypefca has joined #openstack-ironic | 12:19 | |
*** bfournie has joined #openstack-ironic | 12:20 | |
*** jistr|biab is now known as jistr | 12:20 | |
fellypefca | Good morning :) | 12:26 |
openstackgerrit | Pavlo Shchelokovskyy proposed openstack/ironic master: Deny some driver_info updates when console is on https://review.openstack.org/330003 | 12:26 |
*** dprince has joined #openstack-ironic | 12:26 | |
*** mat128 has quit IRC | 12:27 | |
*** mat128 has joined #openstack-ironic | 12:30 | |
*** jkilpatr has joined #openstack-ironic | 12:30 | |
*** rama_y has joined #openstack-ironic | 12:34 | |
*** jaosorior has quit IRC | 12:37 | |
*** thegreenhundred has joined #openstack-ironic | 12:38 | |
*** jaosorior has joined #openstack-ironic | 12:38 | |
*** jaosorior has quit IRC | 12:44 | |
*** jaosorior has joined #openstack-ironic | 12:45 | |
*** jaosorior has quit IRC | 12:45 | |
*** jaosorior has joined #openstack-ironic | 12:46 | |
*** TuanLA has quit IRC | 12:47 | |
milan | morning fellypefca! :) | 12:50 |
*** lucas-hungry is now known as lucasagomes | 12:51 | |
milan | sambetts, https://review.openstack.org/#/c/487909/9 please ;) | 12:51 |
patchbot | patch 487909 - ironic-inspector - Allow periodics to terminate inspector | 12:51 |
sambetts | milan: taking a look now | 12:53 |
sambetts | :) | 12:53 |
milan | sambetts, thanks! :) | 12:53 |
milan | long road to the dnsmasq stuff but getting closer with every patch ;) | 12:54 |
fellypefca | milan o/ | 12:54 |
openstackgerrit | Fellype Cavalcante proposed openstack/ironic master: Introduce hpOneView and ilorest to OneView https://review.openstack.org/468129 | 12:57 |
*** trown|outtypewww is now known as trown | 12:57 | |
sambetts | milan: why an event and not a lock? | 12:58 |
milan | sambetts, yeah, looking at the same, a non-blocking lock should do the same job | 12:58 |
fellypefca | sambetts, yuriyz and TheJulia ^ thanks again for the reviews, and if you could take a look.. :) | 12:59 |
* milan tries to find a "smarter" answer or remember one O:-) | 12:59 | |
sambetts | milan: I just saw event and expected something to be checking whether the event had been set | 12:59 |
sambetts | in a loop somwhere | 12:59 |
sambetts | e.g. if shutting_down.set(): shutdown() | 13:00 |
sambetts | that kind of thing | 13:00 |
*** udesale has joined #openstack-ironic | 13:00 | |
sambetts | buts its being used as a lock/semaphore instead | 13:00 |
milan | sambetts, it has a check later but the other way: try self.shutting_down.set() except: | 13:00 |
milan | yeah | 13:00 |
*** tiendc has quit IRC | 13:01 | |
milan | if a semaphore was used, the shutting down would have to happen in one block | 13:01 |
milan | sambetts, ^ | 13:01 |
milan | I guess | 13:01 |
*** baoli has joined #openstack-ironic | 13:02 | |
*** sasha2 has joined #openstack-ironic | 13:02 | |
*** sasha2 has quit IRC | 13:02 | |
*** sasha2 has joined #openstack-ironic | 13:02 | |
*** Goneri has joined #openstack-ironic | 13:03 | |
sambetts | milan: with the a semaphone you would define it like shutting_down = Semaphore(1) and then in the shutdown function you would do, if not shutting_down.acqire(blocking=False): return | 13:04 |
milan | yup can be done that way | 13:04 |
* milan looks why he put it this way | 13:05 | |
sambetts | yeah the alternative is exactly the same but using a lock, shutting_down = Lock() and then in the shutdown function it would be indentical to the semaphore code | 13:05 |
milan | sambetts, dunno, most likely it didn't cross my mind... can update it though or follow up, thanks for the check | 13:09 |
TheJulia | good morning erveryone | 13:12 |
milan | morning TheJulia! :) | 13:13 |
*** udesale has quit IRC | 13:15 | |
*** mat128 has quit IRC | 13:17 | |
*** ianychoi__ is now known as ianychoi | 13:20 | |
milan | sambetts, the only thing in favour of the event is it's really a single-shot thing that isn't going to be released if anything happened | 13:22 |
milan | ... the process is shutting down | 13:22 |
sambetts | I wouldn't expect a lock to be released unless something explictly calls .release() on it | 13:23 |
milan | but wouldn't you expect releasing a lock eventually? | 13:24 |
sambetts | as far as I can tell an event is basically just a lock with an additional wait() method for sitting waiting until its set by something else | 13:25 |
milan | event is (implemented as) an atomic flag; yes you can wait on it | 13:26 |
milan | it's a signalling mechanism IMO | 13:26 |
sambetts | exactly, which is why it doesn't really fit in this use case | 13:27 |
sambetts | IMO | 13:27 |
sambetts | unless we had another thread that was the "shutdown thread" and we start it and then it just sits and does "shutting_down.wait()" until something sets the flag | 13:27 |
sambetts | that is the use case for an event | 13:28 |
*** zshi_ has quit IRC | 13:29 | |
milan | sambetts, OK, if you won't mind a semaphore that never is going to be released | 13:30 |
milan | I'll amend | 13:30 |
sambetts | milan: I think that is fine, it makes more sense to me anyway, because I saw Event and was looking to see what was checking it, but there isn't anywhere trying to check it, only set it | 13:31 |
milan | sambetts, fair enough, updating | 13:33 |
sambetts | :D thanks! | 13:33 |
*** jcoufal has joined #openstack-ironic | 13:34 | |
*** rama_y has quit IRC | 13:34 | |
*** mgoddard has quit IRC | 13:40 | |
*** jaganathan has quit IRC | 13:42 | |
*** mgoddard has joined #openstack-ironic | 13:44 | |
*** crushil has quit IRC | 13:45 | |
*** zhangfei has joined #openstack-ironic | 13:48 | |
*** zhangfei has quit IRC | 13:53 | |
*** crushil has joined #openstack-ironic | 13:55 | |
*** ChubYann has quit IRC | 13:56 | |
*** jaosorior has quit IRC | 13:57 | |
*** mrch has quit IRC | 13:58 | |
openstackgerrit | Milan Kováčik proposed openstack/ironic-inspector master: Refactoring the firewall https://review.openstack.org/471831 | 13:59 |
openstackgerrit | Milan Kováčik proposed openstack/ironic-inspector master: Allow periodics to terminate inspector https://review.openstack.org/487909 | 13:59 |
openstackgerrit | Milan Kováčik proposed openstack/ironic-inspector master: Introducing a dnsmasq PXE filter driver https://review.openstack.org/466448 | 13:59 |
openstackgerrit | Milan Kováčik proposed openstack/ironic-inspector master: TEST check dnsmasq PXE filter in devstack https://review.openstack.org/468712 | 13:59 |
milan | sambetts, thanks for the check! :) | 13:59 |
*** hjensas has quit IRC | 14:08 | |
*** ChubYann has joined #openstack-ironic | 14:10 | |
*** ijw has joined #openstack-ironic | 14:13 | |
*** hshiina has quit IRC | 14:18 | |
*** snswin has quit IRC | 14:21 | |
*** rivers has joined #openstack-ironic | 14:21 | |
*** ntpttr_laptop has joined #openstack-ironic | 14:22 | |
rivers | Hi, i needed help with tftp-boot, I'm trying to boot up Baremetal node using nova, and after "nova boot", i see the baremetal node getting a DHCP response back from controller, gets an IP, but fails at tftp, and times out | 14:24 |
rivers | I've noticed that tftp-server which is probably supposed to be started by neutron via xinetd isn't running, not sure which logs to look at to debug this | 14:25 |
sambetts | rivers: the tftp server isn't started by neutron, it should be started by the operator as part of the ironic install | 14:26 |
rivers | And also, the pxe.tftp-server IP address is not in the same CIDR as the flat network used to boot up the batemetal node | 14:26 |
sambetts | rivers: it doesn't have to be in ther same network as long as there is a route to it | 14:26 |
rivers | sambetts: I used these instructions to setup tftp, https://docs.openstack.org/ironic/latest/install/configure-pxe.html , I'm not sure if they are sufficient | 14:27 |
rivers | sambetts: tftp-server gets an IP address which is the same OAM ip for the controller node, which I'm trying to boot the baremetal host on a private flat network | 14:28 |
*** srobert has joined #openstack-ironic | 14:28 | |
sambetts | rivers: the flat network needs to have a route to the ironic services including the tftp server | 14:29 |
rivers | sambetts: so basically i did a neutron net-create, and a subnet on that, should i create a router too "in the neutron router-list"? Sorry not too well versed in networking. Thanks | 14:30 |
TheJulia | rivers: yes, you'll likely want to create a router | 14:31 |
TheJulia | or, directly plug your controllers into the provisioning/cleaning networks... but that is like 1000x easier said than done | 14:32 |
sambetts | rivers: if you are using ironic in flat network mode, its expected that your flat network is a provider network with a real world gateway, so that the ironic control plane services can talk directly to the baremetal servers on the IPs they are given | 14:33 |
rivers | TheJulia: in my ironic.conf, I have the cleaning/provisioning set to the uuid of the flat network, Do i set the "#tftp_server=$my_ip" too ? | 14:33 |
sambetts | if you use a neutron network you'll introduce a NAT which means that the conductor will not be able to talk to the baremetal servers | 14:34 |
sambetts | s/neutron network/neutron router/ | 14:34 |
TheJulia | rivers: I would first ensure that there is connectivity between the baremetal node, regardless of the network, and some way for it to reach your conductor. That will dictate what settings you may need to change in your conductor in terms of addresses provided | 14:36 |
rivers | sambetts: When i do a nova boot, i have to specify a "--nic net-id=$IRONIC_NETWORK_ID " , which is configured on a flat network, does that mean that I'm introducing a NAT ... sorry for asking these questions, just trying to clear the concept here | 14:44 |
*** rama_y has joined #openstack-ironic | 14:45 | |
sambetts | rivers: sorry I meant if you use a neutron router you;ll introduce a NAT | 14:45 |
sambetts | rivers: doing nova boot with the network is correct | 14:45 |
*** rama_y has quit IRC | 14:46 | |
rivers | sambetts: So I should not create a neutron router to connect to the neutron net that i created... How do i verify if there is connectivity between the baremetal node and the tftp-server... I'm all over the place, not sure what to do at this point | 14:47 |
sambetts | rivers: the flat network should have a real gateway router and subnet accaible by the subnet that your conductor is in | 14:48 |
rivers | sambetts: ok got it | 14:48 |
*** osdev_ has joined #openstack-ironic | 14:52 | |
openstackgerrit | Merged openstack/ironic master: Boot from volume fails with 'iscsi' deploy interface https://review.openstack.org/499446 | 14:54 |
*** ntpttr_laptop has quit IRC | 14:57 | |
*** rpioso has joined #openstack-ironic | 15:00 | |
*** trungnv has quit IRC | 15:03 | |
*** trungnv has joined #openstack-ironic | 15:04 | |
openstackgerrit | Julia Kreger proposed openstack/bifrost master: VM networking for Neutron testing https://review.openstack.org/498271 | 15:06 |
openstackgerrit | Julia Kreger proposed openstack/bifrost master: DNM: Execute tests for neutron integration https://review.openstack.org/498268 | 15:06 |
openstackgerrit | Julia Kreger proposed openstack/bifrost master: WIP: Initial neutron task changes https://review.openstack.org/452515 | 15:06 |
*** dprince has quit IRC | 15:11 | |
*** parlos has quit IRC | 15:12 | |
openstackgerrit | Peter Piela proposed openstack/ironic-ui master: Cleaup ironic-ui portgroup REST interface https://review.openstack.org/499636 | 15:13 |
*** amarao has quit IRC | 15:18 | |
*** ralonsoh has quit IRC | 15:23 | |
rivers | sambetts: According to these instructions, https://docs.openstack.org/ironic/latest/install/configure-pxe.html , I only start the xinetd process, but never the "systemctl start tftp " , Am i supposed to start the tftp process manually ? | 15:29 |
*** mjura has quit IRC | 15:32 | |
*** pcaruana has quit IRC | 15:32 | |
*** srobert has quit IRC | 15:34 | |
*** ntpttr_laptop has joined #openstack-ironic | 15:37 | |
*** yangyapeng has joined #openstack-ironic | 15:43 | |
*** e0ne has quit IRC | 15:47 | |
sambetts | xinetd should manage the tftp service | 15:48 |
NobodyCam | Good Morning Ironic'ers | 15:54 |
jlvillal | Good morning Ironic | 15:58 |
openstackgerrit | Markos Chandras (hwoarang) proposed openstack/bifrost master: playbooks: roles: bifrost-ironic-install: Set defaults for /etc/hosts https://review.openstack.org/499692 | 16:03 |
vdrok | good morning TheJulia NobodyCam and jlvillal | 16:04 |
NobodyCam | Good morning vdrok and jlvillal :) | 16:04 |
jlvillal | :) | 16:04 |
NobodyCam | :) | 16:04 |
*** lucasagomes is now known as lucas-hungry | 16:04 | |
jlvillal | eat lucas-hungry eat! | 16:04 |
*** marios has quit IRC | 16:06 | |
*** srobert has joined #openstack-ironic | 16:08 | |
*** rama_y has joined #openstack-ironic | 16:11 | |
openstackgerrit | Milan Kováčik proposed openstack/ironic-inspector master: Introducing RPC https://review.openstack.org/497689 | 16:12 |
*** mat128 has joined #openstack-ironic | 16:15 | |
*** hoonetorg has quit IRC | 16:18 | |
anupn | Hi sambetts: after recheck, some of the jobs that were failing due to slow node are not failing. But, tinycore jobs are failing | 16:18 |
anupn | sambetts, http://logs.openstack.org/85/495385/2/check/gate-tempest-dsvm-ironic-ipa-wholedisk-bios-agent_ipmitool-tinyipa-src-ubuntu-xenial/bfaed46/console.html#_2017-08-30_18_39_12_486611 | 16:19 |
anupn | sambetts, I checked that image gets created, and there is no problem is building tinycore image http://logs.openstack.org/85/495385/2/check/gate-tempest-dsvm-ironic-ipa-wholedisk-bios-agent_ipmitool-tinyipa-src-ubuntu-xenial/bfaed46/logs/devstacklog.txt.gz#_2017-08-30_18_20_07_811 | 16:19 |
sambetts | anupn: the problem is this: http://logs.openstack.org/85/495385/2/check/gate-tempest-dsvm-ironic-ipa-wholedisk-bios-agent_ipmitool-tinyipa-src-ubuntu-xenial/bfaed46/logs/devstacklog.txt.gz#_2017-08-30_18_39_08_156 | 16:20 |
sambetts | anupn: http://logs.openstack.org/85/495385/2/check/gate-tempest-dsvm-ironic-ipa-wholedisk-bios-agent_ipmitool-tinyipa-src-ubuntu-xenial/bfaed46/logs/ironic-bm-logs/node-0_console.txt.gz | 16:21 |
sambetts | that is the log of the server that tried to boot | 16:21 |
sambetts | it appears it was unable to mount the root fs of the image | 16:21 |
anupn | sambetts, yes cleaning, i noticed that. Yes i checked bmc logs and it says unable to mount something | 16:21 |
anupn | yeah exactly | 16:21 |
anupn | seems again something not related to the patch, so was wondering what caused it | 16:22 |
sambetts | anupn: I can see that compiling of the python 3 libs is failing here http://logs.openstack.org/85/495385/2/check/gate-tempest-dsvm-ironic-ipa-wholedisk-bios-agent_ipmitool-tinyipa-src-ubuntu-xenial/bfaed46/logs/devstacklog.txt.gz#_2017-08-30_18_19_38_394 | 16:22 |
sambetts | the tinyipa images without that are significantly bigger basically x2 the size of the originals | 16:24 |
sambetts | 34mb vs 61mb | 16:24 |
sambetts | and therefore they can't boot because there isn't enough ram | 16:24 |
sambetts | in the VM | 16:24 |
anupn | hmm, ^ is interesting. | 16:24 |
anupn | sambetts, what do you mean "without that"? | 16:24 |
anupn | "without that are bigger" | 16:25 |
sambetts | without compiling the python3 libs and deleteing all the source | 16:25 |
anupn | sambetts, but i am actually compiling them with python3 here https://review.openstack.org/#/c/495385/2/imagebuild/tinyipa/finalise-tinyipa.sh@171 | 16:26 |
patchbot | patch 495385 - ironic-python-agent - Build tinyipa with python3 | 16:26 |
anupn | and PYOPTIMIZE_TINYIPA is set to true | 16:26 |
anupn | so that will be executed | 16:26 |
sambetts | anupn: http://logs.openstack.org/85/495385/2/check/gate-tempest-dsvm-ironic-ipa-wholedisk-bios-agent_ipmitool-tinyipa-src-ubuntu-xenial/bfaed46/logs/devstacklog.txt.gz#_2017-08-30_18_19_38_394 <---- it is failing | 16:27 |
anupn | sambetts, hmm yes | 16:28 |
anupn | it is unable to execute that command due to not finding /usr/local/lib | 16:28 |
anupn | python3 | 16:28 |
sambetts | its probably /usr/local/lib/python3.4 or something like that | 16:28 |
sambetts | the same as it was python2.7 for python3 | 16:29 |
sambetts | python2* | 16:29 |
anupn | but without that, still image got created just that it got created bigger in size | 16:29 |
*** dprince has joined #openstack-ironic | 16:29 | |
sambetts | anupn: a bigger image can't boot because it doesn't fit into the amount of RAM the VMs have | 16:29 |
anupn | sambetts, yes agree. I see IRONIC_VM_RAM=1024 | 16:30 |
*** jistr is now known as jistr|afk | 16:31 | |
anupn | so yes. it should precompile in that case. But just wondering now, it will never work inside job if we set PYOPTIMIZE_TINYIPA=false | 16:31 |
sambetts | anupn: we know we created and set pyoptimize to true specifically so we could reduce the ram in the job | 16:32 |
*** yangyapeng has quit IRC | 16:32 | |
sambetts | anupn: so if we were to create a job with it turned off, we know we'd need to increase the amount of ram | 16:33 |
* sambetts ---> home | 16:34 | |
anupn | sambetts, hmm wise, agree | 16:34 |
*** nicodemos has left #openstack-ironic | 16:34 | |
*** sambetts is now known as sambetts|afk | 16:34 | |
anupn | sambetts, have a good rest of the evening | 16:34 |
sambetts|afk | o/ cya | 16:34 |
fellypefca | nigh sambetts|afk :) | 16:34 |
anupn | sambetts|afk, cya | 16:34 |
fellypefca | night* :D | 16:34 |
TheJulia | goodnight | 16:34 |
*** trown is now known as trown|lunch | 16:35 | |
*** crushil has quit IRC | 16:41 | |
*** mat128 has quit IRC | 16:44 | |
*** milan has quit IRC | 16:47 | |
*** namnh has joined #openstack-ironic | 16:48 | |
*** marios_ has quit IRC | 16:52 | |
*** crushil has joined #openstack-ironic | 16:53 | |
*** baoli has quit IRC | 17:04 | |
*** mgoddard has quit IRC | 17:04 | |
*** baoli has joined #openstack-ironic | 17:04 | |
*** baoli has quit IRC | 17:06 | |
rivers | Hi, the neutron port thats created for the Baremetal host always seems to remain DOWN, even though I see the baremetal host getting a DHCP response... Is that supposed to be this way ? | 17:12 |
*** rcernin has quit IRC | 17:12 | |
jlvillal | rivers: I think that is a known issue. I don't think it hurts anything. | 17:12 |
jlvillal | rama_y: Do you know for sure? ^^^^ | 17:12 |
rama_y | jlvillal, rivers: it shows DOWN; it is a known issue | 17:13 |
jlvillal | rama_y: Is there a bug on that? | 17:13 |
rama_y | Possibly there is a bug filed; | 17:13 |
rama_y | Let me check. | 17:14 |
*** tesseract has quit IRC | 17:14 | |
rama_y | Here it is: https://bugs.launchpad.net/nova/+bug/1599836 | 17:14 |
openstack | Launchpad bug 1599836 in OpenStack Compute (nova) "Booting Ironic instance, neutron port remains in DOWN state" [Medium,Confirmed] | 17:14 |
*** baoli has joined #openstack-ironic | 17:15 | |
*** srobert_ has joined #openstack-ironic | 17:15 | |
*** srobert has quit IRC | 17:15 | |
rivers | Ok, thanks for pointing that out, My baremetal host cannot connect to the tftp-server to PXE-boot after getting a DHCP response, thought maybe neutron port status "DOWN' may have something to do with this | 17:15 |
*** jkilpatr has quit IRC | 17:16 | |
*** namnh has quit IRC | 17:16 | |
rama_y | rivers: I too observed the same and thought so too. However, it was not the port status that was the issue. | 17:17 |
*** mat128 has joined #openstack-ironic | 17:21 | |
rivers | rama_y: Yeah I'm lost at the moment, not sure how to debug tftp-server, I can see a DHCP request coming back, I have a flat network configured, and a subnet on it, used for nova boot, and I'm using PXE (not iPXE, or uefi) | 17:21 |
rivers | ramay_y: Any pointers , what to look for to debug tftp-server ? Thanks | 17:22 |
rama_y | rivers: in my case I had to configure the network bridge | 17:26 |
*** ijw has quit IRC | 17:28 | |
*** jkilpatr has joined #openstack-ironic | 17:29 | |
*** ijw has joined #openstack-ironic | 17:29 | |
openstackgerrit | Markos Chandras (hwoarang) proposed openstack/bifrost master: playbooks: roles: bifrost-ironic-install: Set defaults for /etc/hosts https://review.openstack.org/499692 | 17:30 |
rivers | rama_y: Well i think my network bridge is fine, as I can get a dhcp response, are there any logs that tftp-server flushes out, | 17:31 |
rivers | Whats the purpose of the map-file , that these instructions recommend " https://docs.openstack.org/ironic/latest/install/configure-pxe.html " | 17:31 |
*** trown|lunch is now known as trown | 17:33 | |
*** ijw has quit IRC | 17:33 | |
*** yangyapeng has joined #openstack-ironic | 17:33 | |
openstackgerrit | Peter Piela proposed openstack/ironic-ui master: Cleaup ironic-ui portgroup REST interface https://review.openstack.org/499636 | 17:33 |
jlvillal | TheJulia: Any ideas? | 17:34 |
jlvillal | openstack server create --flavor baremetal --volume 5b18aba4-613b-407b-bdc8-a5c262e336dc --key-name default testing | 17:34 |
jlvillal | Block Device 5b18aba4-613b-407b-bdc8-a5c262e336dc is not bootable. (HTTP 400) (Request-ID: req-18e5c5c4-8254-4936-88b8-950365fcaf87) | 17:34 |
rama_y | rivers: you will find a map file in your tftpboot directory | 17:34 |
* TheJulia reads | 17:35 | |
rama_y | Did you try ipxe_enabled = false ? | 17:35 |
TheJulia | jlvillal: how did you get the block device? | 17:36 |
jlvillal | TheJulia: I followed these instructions near the bottom: https://docs.openstack.org/ironic/latest/contributor/ironic-boot-from-volume.html | 17:36 |
rama_y | For me, it was my network config that gave me tftp ARP timeout | 17:36 |
rivers | rama_y: ipxe_enabled = false, its set to false in my ironic.conf | 17:36 |
jlvillal | # create volume | 17:36 |
jlvillal | volume=$(openstack volume create --image $image --size 1 my-volume -f value -c id) | 17:36 |
TheJulia | what kind of image did you use? | 17:37 |
jlvillal | TheJulia: I created an Ubuntu wholedisk image | 17:37 |
jlvillal | TheJulia: $ disk-image-create $DISTRO_NAME vm dhcp-all-interfaces -o baremetal-${DISTRO_NAME}-${DIB_RELEASE} | 17:37 |
rama_y | Ok; do you see a folder getting created under /data/ironic/tftpboot with the node UUID? | 17:37 |
jlvillal | export DIB_RELEASE=xenial | 17:37 |
jlvillal | export DISTRO_NAME=ubuntu | 17:37 |
rivers | rama_y: Yes i see a folder getting created under /tftpboot/UUID/ | 17:38 |
rama_y | rivers: if you see this folder appearing with 'config' file and deploy_ramdisk and deploy_kernel files, then your tftp boot is on the right track | 17:38 |
*** yangyapeng has quit IRC | 17:38 | |
rama_y | So, at what point your tftpboot fails? | 17:39 |
TheJulia | jlvillal: out of curiosity, did you set the volume as bootable? | 17:39 |
jlvillal | TheJulia: Uh, I don't know. I was just following the instructions. | 17:39 |
openstackgerrit | Peter Piela proposed openstack/ironic-ui master: Cleanup ironic-ui portgroup REST interface https://review.openstack.org/499636 | 17:40 |
jlvillal | TheJulia: Just this: volume=$(openstack volume create --image $image --size 1 my-volume -f value -c id) | 17:40 |
rivers | rama_y: Well when the baremetal host, gets to the tftp > stage, it fails immediately with PXE-E11: ARP timeout | 17:40 |
TheJulia | jlvillal: I guess it needs a cinder set-bootable... blah blah blah command | 17:40 |
rivers | rama_y: It most likely fails to copy the images from the conductor host to the newly launched baremetal host | 17:40 |
rama_y | rivers: in my case ARP timeout was due to the ovs bridge not configured with an IP address. | 17:41 |
jlvillal | TheJulia: Hmmm. I don't see that in our devstack code. I'll look some more | 17:41 |
rivers | rama_y: Do you need to configure the ovs-bridge with an IP ? I just configured the port, with an IP { the same port thats added to the ovs-bridge over a flat-network } | 17:42 |
rama_y | rivers: in /etc/neutron/plugins/ml2/ml2....ini I had to make changes. | 17:42 |
*** mgoddard has joined #openstack-ironic | 17:43 | |
rama_y | Yes; rivers: the ovs-bridge needs to be configured with an IP | 17:43 |
* TheJulia goes back to shade code | 17:43 | |
*** crushil has quit IRC | 17:43 | |
rama_y | rivers: not sure if you have something like this in ml2 conf file: bridge_mappings = public:br-ex,physnet1:br-eno4 | 17:44 |
rivers | rama_y: I'm assuming you just did that with " ifconfig <br-name> " ? Or did you set that in the "/etc/neutron/plugins/ml2/m" somewhere ? | 17:44 |
* mordred hands TheJulia a bunny rabbit | 17:44 | |
rama_y | rivers: yes; ifconfig <br-name> .... | 17:44 |
TheJulia | mordred: is it an alien bunny rabbit? | 17:45 |
rama_y | rivers: You may already have this in ml2 conf : tenant_network_types = flat,vxlan | 17:45 |
rama_y | rivers: and add your flat network to 'flat_networks = ...' | 17:46 |
* etingof is wondering if anyone has a cycle or two for reviewing https://review.openstack.org/#/c/498489/ | 17:47 | |
patchbot | patch 498489 - ironic-python-agent - Report /dev/disk/by-path on inspection | 17:47 |
rivers | rama_y ; yeah i do have these.. I also noticed that the "tftp-server' kept getting assigned the Host-ip of the conductor node, so I manually set the tftp-server IP on the same subnet as the "br-flat" , which is within the ramge of the netron-nic used for nova booting | 17:48 |
rivers | rama_y; i set the tftp-server =$ip, in the ironic.conf, Not sure if thats the right thing to do, because otherwise, the server-ip was being set to the same as the External OAM ip of the conductor node | 17:48 |
rama_y | rivers: I did not change tftp-server IP | 17:48 |
*** baoli has quit IRC | 17:49 | |
*** yangyapeng has joined #openstack-ironic | 17:49 | |
rivers | rama_y : was your tftp-server IP being set on the same subnet as your bridge ? | 17:49 |
jlvillal | TheJulia: Strange. The volume is in error state after I create it :( | 17:49 |
*** baoli has joined #openstack-ironic | 17:49 | |
rama_y | rivers: tftp-server is set to the same IP as the conductor node in ironic.conf | 17:50 |
rama_y | I left it as is | 17:50 |
*** jistr|afk is now known as jistr | 17:50 | |
rivers | rama_y: No idea where to go from here, Major roadblock for me :( | 17:51 |
rama_y | rivers: I understand completely | 17:52 |
*** links has quit IRC | 17:52 | |
rama_y | rivers: another thing you might want to check is ip routes | 17:53 |
*** etingof has quit IRC | 17:53 | |
*** yangyapeng has quit IRC | 17:53 | |
mordred | TheJulia: of course! | 17:53 |
rivers | rama_y: How do i do that ? | 17:55 |
*** crushil has joined #openstack-ironic | 17:55 | |
rama_y | rivers: #ip route | 17:56 |
jlvillal | TheJulia: This might be it :) 2017-08-31 17:50:41.396 23121 ERROR oslo_messaging.rpc.server ImageUnacceptable: Image 4dfeb2b5-0e7e-4c47-9c78-309ebf68c7d0 is unacceptable: Image virtual size is 2GB and doesn't fit in a volume of size 1GB. | 17:56 |
rama_y | rivers: #route -n | 17:56 |
TheJulia | jlvillal: doh! | 17:56 |
jlvillal | TheJulia: doh! | 17:56 |
*** yangyapeng has joined #openstack-ironic | 17:58 | |
jlvillal | TheJulia: :( ImageCopyFailure: Failed to copy image to volume: qemu-img: Could not open '/opt/stack/data/cinder/conversion/tmp6ILa1Rubuntu-xenial@lvmdriver-1': Unknown driver 'aki' | 17:58 |
rivers | rama_y: I also noticed another thing, when i check "neutron agent-list", i see that the "Metering Agent" , and "Metadata agent" is not alive.. Should they be ? | 17:58 |
rama_y | rivers: I do see metadata agent alive | 17:59 |
rama_y | I do not see metering agent. | 18:00 |
*** rwsu has quit IRC | 18:00 | |
* jlvillal tries to upload the image a different way | 18:00 | |
rama_y | rivers: I don't know much about neutron; my guess is metadata agent is needed. Not sure about metering agent. | 18:01 |
*** yangyapeng has quit IRC | 18:02 | |
*** ijw has joined #openstack-ironic | 18:03 | |
rama_y | jlvillal: Here is a paste for reference: http://paste.openstack.org/show/620121/ | 18:07 |
rama_y | jlvillal: I did this for BFV testing on VM node | 18:07 |
jlvillal | rama_y: Thanks | 18:11 |
rivers | rama_y: Been reading up on this, Did you have to do something with your iptables or rules , as part of your tftp-server bring up process ? Thanks | 18:12 |
rama_y | rivers: yes; I deleted a rule | 18:13 |
rama_y | for br-ex | 18:13 |
rama_y | rivers: ip route del x.x.x.x dev br-ex | 18:14 |
rivers | rama_y: did you have to add any rules, like for port 69 used for tftp "iptables -A INPUT -p tcp --dport 69 -j ACCEPT" , something like this | 18:16 |
rama_y | rivers: I don't recall adding it | 18:16 |
*** rwsu has joined #openstack-ironic | 18:16 | |
rivers | rama_y: Btw what does deleting the br-ex do for you ? | 18:17 |
rama_y | rivers: I was able to pxe boot the node | 18:18 |
rama_y | rivers: also take a look at the tftp config here: /etc/xinetd.d/tftp | 18:20 |
rama_y | you would see something like this: server_args = -v -v -v -v -v --map-file /opt/stack/data/ironic/tftpboot/map-file /opt/stack/data/ironic/tftpboot | 18:21 |
*** jistr is now known as jistr|off | 18:21 | |
rama_y | rivers: you were asking about the map file earlier ^^^ | 18:21 |
rivers | rama_y; the ip address to delete br-ex, is that the one you see from if-config or route ? I just tried and got an error "No such process", even though i see br-ex | 18:21 |
rivers | rama_y: yeah i have something similar in my /etc/xinetd/tftp " -v -v -v -v -v --map-file /tftpboot/map-file /tftpboot " | 18:22 |
rama_y | rivers: I did 'ip route' | 18:23 |
rama_y | rivers: I deleted an ip route; not 'br-ex' or the 'ip address' of br-ex | 18:24 |
*** livelace has quit IRC | 18:24 | |
rama_y | rivers: btw, did you use devstack in your environment? | 18:26 |
*** mat128_ has joined #openstack-ironic | 18:34 | |
*** mat128 has quit IRC | 18:37 | |
*** jtomasek has quit IRC | 18:37 | |
*** fellypefca has quit IRC | 18:38 | |
*** mat128_ has quit IRC | 18:43 | |
*** mat128 has joined #openstack-ironic | 18:43 | |
rivers | rama_y: i used packstack to install Vanilla openstack, and I'm on the Ocata release | 18:46 |
rama_y | rivers: ok; I used devstack, Pike release | 18:47 |
rama_y | rivers: good luck! | 18:49 |
rivers | Thanks | 18:53 |
*** eglute has left #openstack-ironic | 18:55 | |
*** eglute has joined #openstack-ironic | 18:56 | |
*** eglute has left #openstack-ironic | 18:56 | |
*** jkilpatr has quit IRC | 18:57 | |
*** yangyapeng has joined #openstack-ironic | 18:57 | |
*** jkilpatr has joined #openstack-ironic | 18:57 | |
*** rcernin has joined #openstack-ironic | 19:01 | |
*** yangyapeng has quit IRC | 19:01 | |
*** ijw has quit IRC | 19:08 | |
*** pooja has joined #openstack-ironic | 19:11 | |
pooja | Hi all.. Apart from the disk-image-builder, is there another way to generate user images (not deploy images) for baremetal provisioning? | 19:12 |
pooja | Is a qcow2 image used for VMs on KVM same as a whole disk image for Ironic? I have a windows qcow2 with cloudbase-init which works on KVM and would like to deploy it on baremetal | 19:13 |
*** hoonetorg has joined #openstack-ironic | 19:14 | |
*** krtaylor has quit IRC | 19:15 | |
*** mayela29 has joined #openstack-ironic | 19:15 | |
*** mayela29 has quit IRC | 19:15 | |
*** ijw has joined #openstack-ironic | 19:16 | |
*** ijw has quit IRC | 19:16 | |
*** ijw has joined #openstack-ironic | 19:16 | |
TheJulia | pooja: there are other ways, and libvirt does grok qcow2s for disks. You should be able to deploy that windows image | 19:17 |
pooja | TheJulia: thanks! Good to know that the same images can be used for VMs and baremetal :) | 19:18 |
TheJulia | pooja: so... | 19:18 |
TheJulia | the images must be what are called whole disk images in those cases | 19:19 |
pooja | Sure! | 19:21 |
pooja | Curious to know - what are the other ways of generating new images for ironic? | 19:21 |
*** etingof has joined #openstack-ironic | 19:29 | |
*** mrtenio has joined #openstack-ironic | 19:35 | |
*** ijw has quit IRC | 19:39 | |
*** mrtenio has quit IRC | 19:40 | |
*** jkilpatr has quit IRC | 19:40 | |
openstackgerrit | Ramamani Yeleswarapu proposed openstack/ironic master: [TEST] [DO NOT MERGE] Checking gate jobs https://review.openstack.org/499768 | 19:44 |
TheJulia | pooja: manually :) | 19:50 |
TheJulia | s/manually/manually by hand using loopbacks and eventually creating a raw or qcow2 file that represents the disk | 19:51 |
*** e0ne has joined #openstack-ironic | 19:53 | |
*** yangyapeng has joined #openstack-ironic | 19:58 | |
*** yangyapeng has quit IRC | 20:02 | |
*** jkilpatr has joined #openstack-ironic | 20:03 | |
pooja | TheJulia: This is the guide I followed for manually creating the Windows qcow2 image - https://docs.openstack.org/image-guide/windows-image.html | 20:08 |
*** yee379 has quit IRC | 20:09 | |
pooja | It worked on KVM, but for baremetal the node fails on boot and keeps restarting while trying to apply some configuration. Doesn't start in safe mode either. | 20:09 |
*** yee379 has joined #openstack-ironic | 20:09 | |
pooja | Are these image creation steps not enough for Ironic? | 20:09 |
TheJulia | network adapter would naturally be different | 20:12 |
TheJulia | does it even start booting? | 20:12 |
*** rpioso has quit IRC | 20:14 | |
pooja | Yes, it goes to the boot screen and then I see a popup for 'restarting due to error' | 20:16 |
*** krtaylor has joined #openstack-ironic | 20:16 | |
rivers | Hi, I managed to bootup a baremetal node, but it seemed to bootup just once, and took me to a ubuntu login screen , and on the controller host itself, I saw the provision_state of node go from "deploying" to "wait-call-back" | 20:31 |
rivers | I thought isnt the deploy image supposed to load the User Image, and reboot itself, I have no clue why it even took me to a login screen for a Ubuntu image , and never rebooted | 20:31 |
*** e0ne has quit IRC | 20:36 | |
TheJulia | rivers: could it be that you have an ubuntu based deployment ramdisk... and that you don't have network connectivity enabling the wait callback state to be exited? Have you checked your ironic-api and ironic-conductor logs? | 20:37 |
jlvillal | TheJulia: rama_y: Making progress on BFV testing. I got to an error since I am not using iPXE. Recreating my devstack VM...again | 20:38 |
TheJulia | heh | 20:39 |
TheJulia | yeah, that is kind of a super hard requirement | 20:39 |
rama_y | jlvillal, thanks. | 20:40 |
jlvillal | TheJulia: I had disabled it because I was having issues where it wouldn't chain boot :( But usually I can manually enter the iPXE command line and get it to work. | 20:40 |
*** ijw has joined #openstack-ironic | 20:40 | |
jlvillal | TheJulia: I think something strange with Vagrant/Virtualbox networking | 20:40 |
rivers | TheJulia: The only useful thing i see in ironic-conductor logs is this "Deploy boot mode is None for 764a5c62-21ee-494f-ae6b-044b6d02fadc. get_boot_mode_for_deploy /usr/lib/python2.7/site-packages/ironic/drivers/modules/deploy_utils.py:782 | 20:40 |
rivers | " | 20:40 |
TheJulia | jlvillal: I'd <3 a screenshot of that | 20:40 |
jlvillal | TheJulia: I can try :) | 20:41 |
jlvillal | TheJulia: The failure? | 20:41 |
jlvillal | I have a little video :) | 20:41 |
TheJulia | a video of it would work | 20:41 |
TheJulia | just to understand what state it is actually getting stuck in | 20:41 |
TheJulia | since the step immediately following should be dhcp'ing, and then that dhcp reply should have a url in it from the dhcp server | 20:42 |
jlvillal | TheJulia: https://goo.gl/photos/7S6Pn57cBkyC5P4A9 | 20:42 |
TheJulia | hmmmmm | 20:43 |
jlvillal | TheJulia: It will automatically loop the video. I found it easier to download it and watch in VLC | 20:43 |
TheJulia | does it never actually get to ipxe on it's own | 20:43 |
TheJulia | ? | 20:43 |
rivers | TheJulia: After deploying athe deploy image, in an ideal scenario, does the baremetal node itself reboot to load user image, after doing a successful Lookup ? | 20:44 |
jlvillal | TheJulia: Not really. Seems strange. PXE and tftp work. It downloads the iPXE program | 20:44 |
jlvillal | TheJulia: But then iPXE fails. Very strange I think | 20:44 |
*** ijw has quit IRC | 20:45 | |
*** yangyapeng has joined #openstack-ironic | 20:53 | |
TheJulia | rivers: lookup occurs to identify it's self. deploying node then gets instructions on what to do, image is written to disk, and then the node is rebooted. | 20:54 |
TheJulia | jlvillal: that is really weird | 20:54 |
jlvillal | TheJulia: Yeah :( | 20:55 |
*** mat128 has quit IRC | 20:55 | |
TheJulia | maybe the wrong ipxe binary? | 20:55 |
jlvillal | TheJulia: What do you mean? The one that is running that times out? Or the one it is trying to fetch (boot.ipxe) ? | 20:57 |
*** yangyapeng has quit IRC | 20:57 | |
*** crushil has quit IRC | 20:58 | |
rivers | TheJulia: I just did this again, and noticed that immediately after getting the pxe-boot image, the ironic node provisioning state, moves from "deploying" to "wait call-back" , and this is even before the node bootsup | 20:58 |
rivers | TheJulia: and upon the first boot, the node just goes into a login prompt, and doesnt seem to accept default ubuntu login credentials | 20:58 |
TheJulia | jlvillal: hmmmmmmmmm | 20:59 |
jlvillal | TheJulia: And my use case of my Virtualbox being the iPXE server seems like a rare case. Everything I search for seems to be the VM as the client for iPXE | 21:00 |
TheJulia | jlvillal: because it is not on the first device, and for some reason it is not going anywhere. is spanning tree enabled? | 21:00 |
TheJulia | rivers: yes, wait call-back is an intermediate state while we are waiting for the agent to check-in | 21:00 |
jlvillal | TheJulia: I tried this with an ethernet cable connected between my two boxes and it failed also. So no switch. | 21:00 |
jlvillal | TheJulia: Now I'm back to using a 10Gb fiber switch between the boxes | 21:01 |
TheJulia | make it use the first interface perhaps? since your pee'ing off of the second interface it looks like | 21:01 |
TheJulia | pxe'ing | 21:01 |
jlvillal | heh | 21:01 |
* TheJulia actually goes to figure out how to disable autocorrect | 21:01 | |
jlvillal | It made me laugh! | 21:01 |
rivers | TheJulia: oh, according to this https://docs.openstack.org/ironic/latest/_images/states.svg , it should have gone from "deploying" to "Active" , it only goes to wait-call-back if its failing or deleting | 21:02 |
jlvillal | I was like, what? | 21:02 |
TheJulia | \o/ no more autocorrect | 21:02 |
srart | Don't know if the rest of the country is panicking yet, but in TX we have gas lines, like in the 70s lines. | 21:02 |
TheJulia | srart: 70s vehicles deep? | 21:03 |
TheJulia | vehicles from the 70s? | 21:03 |
* TheJulia is otherwise confused | 21:03 | |
srart | like the lines in the 70s gas shortage. | 21:03 |
TheJulia | oh, ahh | 21:03 |
*** trown is now known as trown|outtypewww | 21:04 | |
TheJulia | people are not freaking out quite yet, at least in New Mexico. | 21:04 |
*** yusef has joined #openstack-ironic | 21:04 | |
srart | may want to beat your local panic and get it before prices finish skyrocketing. | 21:04 |
* TheJulia has a full gas tank and a full diesel tank | 21:04 | |
* TheJulia also has a full propane tank, but that is only used for warmth and cooking | 21:04 | |
* srart has half electric car | 21:05 | |
*** ijw has joined #openstack-ironic | 21:05 | |
srart | so I get to be extra smug right now. ;-) | 21:05 |
* TheJulia wishes her RV was all electric..... | 21:05 | |
srart | biodiesel it and chop up the bodies. | 21:05 |
TheJulia | rivers: so the state machine doesn't let it go from deploying to active until it is done deploying | 21:06 |
TheJulia | rivers: think of wait callback as a state in between | 21:06 |
TheJulia | srart: Biodiesel blends, sadly gets me less milage when towing my car | 21:07 |
rivers | TheJulia: ok, this is the furtheset ive gotten in a week, and now deployment fails, because I'm taken to a Ubuntu login page, on first boot, and then failing... the ironic-conductor.log dont show anything useful, looks like its not even getting to a lookup phase, should i suspect the deploy image created using "disk-image-builder" ? | 21:08 |
*** penick has joined #openstack-ironic | 21:10 | |
rivers | Also, do both the Deploy Image and the User Image need to be the same OS ? Can i have deploy as Ubuntu, and User image as Fedora etc ? | 21:10 |
TheJulia | rivers: that is exactly what you should suspect | 21:10 |
TheJulia | they do not need to be the same distribution or os | 21:10 |
TheJulia | so, I'd recommend coreos or maybe building a fedora image. Also, there are DIB elements that you can use to burn a user into the image for troubleshooting | 21:11 |
TheJulia | Most likely, with the ramdisk booting, it is hardware support, like it doesn't know how to use the network cards or something silly like that | 21:12 |
rivers | TheJulia: I used this page, and just used the disk-image-builder (second half on this page) https://docs.openstack.org/ironic/latest/install/deploy-ramdisk.html | 21:12 |
rivers | TheJulia: Do you recommend using ironic python agent instead of disk image builder ? | 21:13 |
*** Goneri has quit IRC | 21:13 | |
TheJulia | https://docs.openstack.org/diskimage-builder/latest/elements/devuser/ | 21:14 |
*** crushil has joined #openstack-ironic | 21:15 | |
TheJulia | the diskimage-builder ironic-agent image creation just installs ironic python agent inside of the image. You may want to try our pre-built CoreOS images | 21:15 |
*** mgoddard has quit IRC | 21:15 | |
jlvillal | rivers: Do you have a serial connection to the system? It can be very illuminating compared to not having it. | 21:15 |
TheJulia | o/ crushil | 21:15 |
jlvillal | rivers: My minicom status line: CTRL-A Z for help | 9600 8N1 | NOR | Minicom 2.7 | VT102 | Offline | ttyS0 | 21:16 |
*** ijw has quit IRC | 21:16 | |
jlvillal | TheJulia: Yeah, failed again. But I hit CTRL-B. Did 'dhcp' and then a few moments later 'chain http://10.0.2.15:3928/boot.ipxe' and it worked. Weird! | 21:26 |
TheJulia | So, a Ethernet interface hold down it is then.... | 21:27 |
rivers | TheJulia: I wasn't able to ssh into the baremetal node, CTRL-A-Z at which point? Sorry didnt full understand what you meant by that | 21:27 |
TheJulia | But... Just dhcping should be enough | 21:27 |
TheJulia | Since the reply from the server should tell you what to do........ | 21:28 |
*** etingof has quit IRC | 21:32 | |
*** etingof has joined #openstack-ironic | 21:32 | |
*** racedo has quit IRC | 21:33 | |
rivers | TheJulia: Where do i find the prebuild CoreOs deployment images ? | 21:34 |
TheJulia | rivers: ignore the bit about control-a-z | 21:35 |
rivers | ok | 21:35 |
TheJulia | https://tarballs.openstack.org/ironic-python-agent/coreos/files/ | 21:35 |
*** karimb has joined #openstack-ironic | 21:35 | |
TheJulia | you'll need a ramdisk and kernel file that match the version of openstack ironic your using | 21:36 |
TheJulia | if you look at our troubleshooting docs, it tells you how to set it up so you can ssh into coreos | 21:36 |
TheJulia | you can also just login on the console | 21:36 |
*** srobert_ has quit IRC | 21:36 | |
*** penick has quit IRC | 21:41 | |
openstackgerrit | John L. Villalovos proposed openstack/python-ironicclient stable/pike: Allow OS_BAREMETAL_API_VERSION=latest to work https://review.openstack.org/499819 | 21:41 |
*** baoli has quit IRC | 21:43 | |
rivers | TheJulia: Found a download link to prebuild coreos deploy images here too https://docs.openstack.org/ironic/latest/install/deploy-ramdisk.html .... Btw with this, I can still use the Ubuntu User images right ? The User image can be any OS right, regardless of what the Deploy image OS is ? | 21:45 |
rivers | thanks | 21:45 |
*** baoli has joined #openstack-ironic | 21:45 | |
TheJulia | yes, you can. A user image can be any os when it is a whole disk image | 21:46 |
*** priteau has quit IRC | 21:49 | |
*** bfournie has quit IRC | 21:49 | |
*** baoli has quit IRC | 21:50 | |
rivers | TheJulia: Does it have to be a whole disk image ? I created parition images using "disk-image-create ubuntu baremetal dhcp-all-interfaces grub2 -o my-image " | 21:52 |
jlvillal | TheJulia: I think I just booted from volume on real bare-metal :) | 21:53 |
TheJulia | rivers: I think that actually makes you a whole disk image | 21:54 |
*** yangyapeng has joined #openstack-ironic | 21:54 | |
TheJulia | jlvillal: woot! | 21:54 |
jlvillal | TheJulia: Probably old news for you. But cool for me to see it in real-life | 21:54 |
*** yangyapeng has quit IRC | 21:58 | |
rama_y | jlvillal, awesome! | 21:58 |
jlvillal | But then I can't login. Since no DHCP | 21:59 |
jlvillal | And both consoles are dead. Serial and KVM | 21:59 |
rivers | TheJulia: No the whole disk image is created with this "disk-image-create ubuntu vm dhcp-all-interfaces -o my-image " or so the openstack page says... .. but to ask the question again, if its not a whole disk image, then a CoreOS deploy image will not support a Ubuntu user image ? is that the case , thanks | 22:00 |
jlvillal | But that was happening with cirros before. Since I have 4 interfaces and it only seems to DHCP the first interface. | 22:00 |
*** [1]rpioso has joined #openstack-ironic | 22:00 | |
* jlvillal finds it confusing that dib element 'vm' makes a wholedisk image and dib element 'baremetal' makes a non-wholedisk image. | 22:01 | |
TheJulia | it doesn't really afaik | 22:01 |
TheJulia | anyway | 22:01 |
jlvillal | Not exactly intuitive names to me. | 22:02 |
TheJulia | Things have become confused | 22:02 |
TheJulia | If it is not linux, and does not use grub, then you must use a whole disk image. So, coreos really doesn't care about what it deloys | 22:02 |
rivers | TheJulia: There is this parameter for ironic node "driver_internal_info | {u'is_whole_disk_image': False} ", does this get automatically set, or am i supposed to modify that based on the type of image ? | 22:11 |
pooja | is_whole_disk_image was set automatically in my case when used with nova-compute | 22:13 |
TheJulia | it is set based upon the parameters supplied for the deployment | 22:13 |
TheJulia | so, automatically | 22:13 |
*** jcoufal has quit IRC | 22:16 | |
*** ijw has joined #openstack-ironic | 22:17 | |
rivers | TheJulia: So i just tried the prebuild coreos image you pointed me to, and even this on first boot takes me to "login prompt " | 22:17 |
TheJulia | https://docs.openstack.org/developer/ironic-python-agent/troubleshooting.html | 22:18 |
rivers | TheJulia: Am i supposed to do something at this point, or i thought its supposed to reboot itself with the user-image ... and i have the "local boot" option also set for node and flavor | 22:18 |
*** dprince has quit IRC | 22:20 | |
TheJulia | rivers: Well, deployment is not instantanious. You'll want to look at the node status. In the mean time, you might want to look at the webpage I just linked | 22:20 |
*** bfournie has joined #openstack-ironic | 22:23 | |
*** ijw has quit IRC | 22:23 | |
*** bfournie has quit IRC | 22:24 | |
*** bfournie has joined #openstack-ironic | 22:25 | |
*** ijw has joined #openstack-ironic | 22:27 | |
*** Nisha_Agarwal has joined #openstack-ironic | 22:28 | |
*** krtaylor has quit IRC | 22:28 | |
*** rcernin has quit IRC | 22:30 | |
*** Nisha_Agarwal has quit IRC | 22:38 | |
*** leyal has quit IRC | 22:38 | |
*** thegreenhundred has quit IRC | 22:38 | |
*** leyal has joined #openstack-ironic | 22:38 | |
*** [1]rpioso has quit IRC | 22:40 | |
*** Nisha_Agarwal has joined #openstack-ironic | 22:51 | |
*** yangyapeng has joined #openstack-ironic | 22:55 | |
*** pooja has quit IRC | 22:55 | |
*** yangyapeng has quit IRC | 22:59 | |
*** rbudden has quit IRC | 23:06 | |
*** ijw has quit IRC | 23:11 | |
*** stendulker has joined #openstack-ironic | 23:11 | |
*** rbudden has joined #openstack-ironic | 23:11 | |
*** ijw has joined #openstack-ironic | 23:12 | |
*** stendulker_ has joined #openstack-ironic | 23:13 | |
*** stendulker has quit IRC | 23:15 | |
*** ijw has quit IRC | 23:16 | |
crushil | TheJulia, Sorry was afk | 23:29 |
*** charcol has joined #openstack-ironic | 23:36 | |
*** Nisha_Agarwal has quit IRC | 23:37 | |
*** ntpttr_laptop has quit IRC | 23:40 | |
*** tcrowe has joined #openstack-ironic | 23:41 | |
*** ijw has joined #openstack-ironic | 23:45 | |
*** stendulker_ has quit IRC | 23:46 | |
*** rama_y has quit IRC | 23:48 | |
*** priteau has joined #openstack-ironic | 23:49 | |
*** ijw has quit IRC | 23:51 | |
*** yangyapeng has joined #openstack-ironic | 23:52 | |
*** priteau has quit IRC | 23:54 | |
*** yangyapeng has quit IRC | 23:57 | |
openstackgerrit | Anup Navare proposed openstack/ironic-python-agent master: Build tinyipa with python3 https://review.openstack.org/495385 | 23:58 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!