*** rloo has quit IRC | 00:05 | |
*** yolanda has quit IRC | 00:08 | |
*** gyee has quit IRC | 00:19 | |
*** harlowja has quit IRC | 00:23 | |
*** moshele has quit IRC | 00:25 | |
*** livelace has quit IRC | 00:35 | |
*** livelace has joined #openstack-ironic | 00:35 | |
*** phuongnh has joined #openstack-ironic | 01:19 | |
*** dmellado has quit IRC | 01:22 | |
*** rcernin_ has joined #openstack-ironic | 01:23 | |
*** rcernin has quit IRC | 01:25 | |
openstackgerrit | Tony Breeds proposed openstack/ironic-python-agent master: Add a UUID to the extra-hardware data on ppc64le https://review.openstack.org/594591 | 01:32 |
---|---|---|
openstackgerrit | Kaifeng Wang proposed openstack/ironic-inspector master: Wrap rpc server into oslo.service https://review.openstack.org/584758 | 01:41 |
*** rcernin has joined #openstack-ironic | 02:00 | |
*** rcernin_ has quit IRC | 02:03 | |
openstackgerrit | Kaifeng Wang proposed openstack/ironic-inspector master: Wrap Flask into oslo.service https://review.openstack.org/561823 | 02:23 |
*** zhangfei has joined #openstack-ironic | 02:28 | |
*** dmellado has joined #openstack-ironic | 02:38 | |
*** rpioso is now known as rpioso|afk | 02:43 | |
*** mmethot has quit IRC | 02:46 | |
*** edleafe has quit IRC | 02:46 | |
*** edleafe has joined #openstack-ironic | 02:47 | |
*** mmethot has joined #openstack-ironic | 02:49 | |
*** Bhujay has joined #openstack-ironic | 04:16 | |
*** galaxyblr has joined #openstack-ironic | 04:16 | |
*** galaxyblr has quit IRC | 04:38 | |
*** galaxyblr has joined #openstack-ironic | 04:39 | |
*** trungnv has quit IRC | 04:45 | |
*** moshele has joined #openstack-ironic | 05:00 | |
*** moshele has quit IRC | 05:02 | |
*** sdake has quit IRC | 05:13 | |
*** sdake has joined #openstack-ironic | 05:13 | |
*** sdake has quit IRC | 05:50 | |
*** sdake has joined #openstack-ironic | 05:51 | |
*** hjensas has quit IRC | 05:52 | |
openstackgerrit | Kaifeng Wang proposed openstack/ironic-inspector master: Wrap Flask into oslo.service https://review.openstack.org/561823 | 05:56 |
*** sdake has quit IRC | 06:02 | |
*** sdake has joined #openstack-ironic | 06:04 | |
*** moshele has joined #openstack-ironic | 06:06 | |
*** adrianc has joined #openstack-ironic | 06:26 | |
*** gkadam has joined #openstack-ironic | 06:31 | |
*** pcaruana has joined #openstack-ironic | 06:41 | |
openstackgerrit | Merged openstack/ironic stable/pike: Fix iDRAC hardware type does not work with UEFI https://review.openstack.org/588942 | 06:43 |
*** tssurya has joined #openstack-ironic | 06:47 | |
*** hjensas has joined #openstack-ironic | 06:56 | |
*** amoralej|off is now known as amoralej | 07:01 | |
*** zhangfei has quit IRC | 07:03 | |
*** dmellado has quit IRC | 07:04 | |
*** dmellado has joined #openstack-ironic | 07:06 | |
*** rcernin has quit IRC | 07:07 | |
openstackgerrit | Kaifeng Wang proposed openstack/ironic-inspector master: Wrap Flask into oslo.service https://review.openstack.org/561823 | 07:07 |
*** dmellado has quit IRC | 07:12 | |
*** phuongnh has quit IRC | 07:17 | |
*** dmellado has joined #openstack-ironic | 07:19 | |
*** dtantsur|afk is now known as dtantsur | 07:22 | |
dtantsur | morning ironic | 07:23 |
etingof | dtantsur, \o | 07:23 |
openstackgerrit | lei zhang proposed openstack/bifrost master: Migrate the link of bug report button to storyboard https://review.openstack.org/594876 | 07:39 |
*** jtomasek has joined #openstack-ironic | 07:42 | |
openstackgerrit | Ilya Etingof proposed openstack/sushy master: Add a virtual media resource https://review.openstack.org/570810 | 07:49 |
*** adrianc has quit IRC | 07:50 | |
*** jaganathan has quit IRC | 07:52 | |
openstackgerrit | Merged openstack/ironic-inspector-specs master: import zuul job settings from project-config https://review.openstack.org/592389 | 07:56 |
*** e0ne has joined #openstack-ironic | 08:01 | |
*** mgoddard has joined #openstack-ironic | 08:04 | |
*** w-miller has joined #openstack-ironic | 08:11 | |
*** amoralej is now known as amoralej|brb | 08:16 | |
*** pcaruana has quit IRC | 08:28 | |
*** pcaruana has joined #openstack-ironic | 08:30 | |
*** dirk has joined #openstack-ironic | 08:45 | |
*** amoralej|brb is now known as amoralej | 08:47 | |
*** sambetts_ is now known as sambetts | 09:00 | |
*** moshele has quit IRC | 09:05 | |
*** moshele has joined #openstack-ironic | 09:05 | |
*** amoralej is now known as amoralej|brb | 09:24 | |
*** moshele has quit IRC | 09:34 | |
*** dougsz_ has joined #openstack-ironic | 09:44 | |
*** moshele has joined #openstack-ironic | 09:45 | |
*** amoralej|brb is now known as amoralej | 09:56 | |
*** yolanda has joined #openstack-ironic | 09:56 | |
openstackgerrit | cheng li proposed openstack/ironic-lib master: Check GPT table with sgdisk insread of partprobe https://review.openstack.org/594922 | 10:11 |
openstackgerrit | cheng li proposed openstack/ironic-lib master: Check GPT table with sgdisk insread of partprobe https://review.openstack.org/594922 | 10:26 |
*** jesusaur has quit IRC | 10:38 | |
*** adrianc has joined #openstack-ironic | 10:39 | |
*** jesusaur has joined #openstack-ironic | 10:41 | |
yolanda | hi... i have more problems testing standalone in ironic. Now i'm able to boot the ipa, and it seems to be stuck on deploying state | 10:50 |
yolanda | i can login into the ipa | 10:50 |
yolanda | and i see it just ends on finding the node, it finds the root node at /dev/sda | 10:50 |
yolanda | but after that does nothing, just hearbeats | 10:50 |
yolanda | what could be the problem there? | 10:51 |
sambetts | yolanda: is the node in maintenance? | 10:51 |
yolanda | sambetts, no, it's stuck in deploying state | 10:51 |
yolanda | in conductor logs, the last i can see is Node 37d158bc-821d-409d-88c0-d0bc9c24f4ea moved to provision state "deploying" from state "wait call-back"; target provision state is "active" | 10:51 |
sambetts | hmmm, only time I've seen IPA just sit and heartbeat without doing anything is when the node has got maintenance=true, it might be doing something in the background if its in deploying state, if your using iscsi deploy then once IPA sets up the iscsi endpoint then all the work should be being done by the conductor, IPA is hands off at that point | 10:53 |
yolanda | i cannot see any logs in conductor, just moving to wait call-back , then to deploying | 10:54 |
yolanda | and it remains locked until i restart the conductor service | 10:55 |
yolanda | i even cannot see any logs in the conductor, that says it tries to fetch the image, anything, although i enabled debugging | 10:56 |
sambetts | hmm how are you viewing the conductor logs? there should be logs touching on all of that | 10:57 |
yolanda | it's a container, with docker logs | 10:57 |
yolanda | and on the ipa, i enabled ssh access, so i see it directly | 10:58 |
sambetts | yolanda: kolla container? | 10:58 |
yolanda | sambetts, i am trying helm.. but yep, i think it uses kolla image | 10:58 |
yolanda | pike version, as queens was failing to install | 10:59 |
*** slagle has quit IRC | 11:01 | |
sambetts | yolanda: something I found when I was working to get ironic running in kolla-k8s was that the docker logs didn't contain much information, kolla put all the logs proper logs into a /var/logs/kolla/ironic directory, so you had to view them by doing docker exec with a cat /var/log/ironic/ironic-conductor.log or something like that | 11:01 |
sambetts | not sure if its the same with the helm version | 11:01 |
yolanda | let me check | 11:02 |
yolanda | i need to get more clues | 11:02 |
yolanda | i needed to do something like that on the http container | 11:02 |
yolanda | /var/log/ironic is empty, but i could configure it | 11:03 |
sambetts | any other directorys? kolla put everything in /var/log/kolla | 11:03 |
yolanda | i don't have a kolla directory there | 11:03 |
yolanda | just "io.kubernetes.container.logpath": "/var/log/pods/6b0fd3d1-a60e-11e8-8d7a-ecf4bbc0b534/ironic-conductor/0.log", | 11:04 |
yolanda | and that's quite empty | 11:04 |
sambetts | hmmm yeah might be best to configure ironic.conf to log to a specific location you can look at if it isn't already | 11:06 |
sambetts | something else I found when doing ironic things was that getting iscsi to work inside a container was a bit tricky, and I ended up having to run an priviledged iscsid/tgtd container alongside ironic-conductor to make sure the host kernel was setup right to make it work | 11:08 |
sambetts | but this was a little while ago | 11:08 |
yolanda | i could try with deploy direct as well | 11:08 |
yolanda | maybe gives less headache | 11:08 |
yolanda | i was thinking in something like that: permissions, communication | 11:08 |
sambetts | direct deploy should work with standalone you just need to have a http server to serve the images from | 11:09 |
yolanda | i even could reuse the ironic http one, right? | 11:09 |
sambetts | yolanda: should be able to, thats how non-containerised bifrost does it, it uses the same http server as for ipxe | 11:10 |
yolanda | yep, httpboot | 11:10 |
yolanda | i was thinking on that | 11:10 |
yolanda | may be easier | 11:10 |
yolanda | i don't have any special requirement for iscsi | 11:11 |
sambetts | yeah for standalone I think it'll be simpler | 11:12 |
yolanda | this http works for sure as it's used to load the boot | 11:12 |
sambetts | yeah, and it takes the work load of the conductor too | 11:12 |
yolanda | trying | 11:14 |
yolanda | yay! | 11:18 |
yolanda | thx sambetts | 11:20 |
moshele | sambetts: hi | 11:21 |
sambetts | no problem :) it would be interesting to get some logs from the iscsi deploy to see what was failing and if there is something we need to fix there | 11:21 |
sambetts | moshele: Hi | 11:21 |
moshele | sambetts: can we talk about the smart-nic spec https://review.openstack.org/#/c/582767/? | 11:21 |
patchbot | patch 582767 - ironic-specs - Add Support for Smart NIC - 0h 42m 27s spent in CI | 11:21 |
sambetts | moshele: sure | 11:22 |
*** S4ren has quit IRC | 11:23 | |
moshele | sambetts: first of all we want that the plug of the port will be on the ovs neutron side see https://review.openstack.org/#/c/586252/. this is because neutron have all the bridge information | 11:23 |
patchbot | patch 586252 - neutron - Add SmartNIC representor port to integeration bridge - 8h 43m 16s spent in CI | 11:23 |
moshele | sambetts: second I am not sure how your proposl solve the deploy issue as we need to change the deployment flow. power on before plugging the network | 11:25 |
moshele | sambetts: also creating another network_interface seems like lost of code duplication ... | 11:25 |
sambetts | moshele: I personally don't think the existing OVS neutron mechanism driver should include that logic, I think that if we want to use the OVS mech driver we should behave like nova and plug and configure the downstream end of the connection to the baremetal, that way the OVS agent in neutron can operate exactly as it would for a VM and Ironic remains in control of how and when the baremetal | 11:32 |
sambetts | ports are connected to OVS, and we might be able to use https://github.com/openstack/os-vif to help us there too | 11:32 |
*** rh-jelabarre has joined #openstack-ironic | 11:33 | |
sambetts | moshele: my proposal for the deployment flow, adds a call to the network interface before the node is powered off after deployment, in addition to the one that is already called after power off doesn't that solve your issue? | 11:33 |
sambetts | moshele: I think the binding data for a OVS port is significantly different from the binding data for a baremetal vnic type port so I think there would be little duplication between the network interfaces right? | 11:35 |
moshele | sambetts: this an example of the change we need https://review.openstack.org/#/c/583573/5/ironic/drivers/modules/agent.py | 11:36 |
patchbot | patch 583573 - ironic - Add support for smartNICs - 9h 28m 24s spent in CI | 11:36 |
moshele | sambetts: for example in deploy the current code poweroff the machine/ but we need to reboot it to make sure we can connect to the ovs before we plug the ports | 11:37 |
moshele | sambetts: if we put a side who will be responsible to plug the port ironic or neutron. It will be almost the same just update the port host_id with the location the neutron agent | 11:39 |
moshele | sambetts: also we not just talk about ovs/ the can be any neutron l2 agent that can support it | 11:40 |
sambetts | moshele: I understand that, but sticking "if smartnics" everywhere isn't reusable in anyway and very specific, my proposal would add pre-reboot (power on) tasks to the offical API for ironic network interfaces so that other network interfaces can also add tasks at the point in the process if required | 11:40 |
sambetts | moshele: having other Ml2 l2 agents supprt it is exactly another reason why Ironic should use os-vif and not put the logic in neutron | 11:41 |
jroll | morning | 11:44 |
sambetts | o/ jroll | 11:44 |
jroll | don't have full context, but +1 for pre/post power off things in the network driver, changing the deployment flow to support smartnic seems hairy | 11:47 |
moshele | sambetts: not sure i understand. the deploy driver calls the network interface telling it when to switch to provision network. this today is done pre-reboot. for smart nic the power must be set to configure the network. having the deploy call the network driver before power does not help. are you suggesting the network driver control/affect the pow | 11:51 |
moshele | er state? | 11:51 |
*** mgoddard has quit IRC | 11:52 | |
jroll | moshele: what sam is saying is that attach_tenant_networks() or whatever would still be before power on, and some new post_power_on_work() method would be after power on, and the smartnic driver would do its work in the latter | 11:52 |
sambetts | I was thinking the other way around, a pre power off action (before we shut down IPA), and then the existing one, because we need to unplug the provisioning network before we shutdown IPA if the server needs to be on | 11:54 |
jroll | well, same idea | 11:55 |
jroll | :) | 11:55 |
sambetts | yup :) | 11:55 |
jroll | though moving the existing one feels wrong, if there's a driver that depends on that, it's suddenly broken :) | 11:55 |
sambetts | the existing one is already after we shutdown IPA, but before we power it back on to the guest OS | 11:56 |
sambetts | my proposal adds an action before we shutdown IPA the existing one doesn't move | 11:56 |
jroll | oh, I need more coffee | 11:57 |
jroll | I get what you're saying now | 11:58 |
jroll | but it sounds like moshele needs the existing one to be after power on, for the same reason, the NICs need power | 11:58 |
sambetts | my suggestion is that the smart nics would have a new network_interface (as its quite differnt to the existing implementation) and they would only implement the pre-power off function not the existing one | 11:59 |
jroll | so tenant networks would be plugged while IPA is up? | 12:01 |
jroll | sounds sketchy :) | 12:01 |
moshele | sambetts: let say the baremetal is poweroff state and I am doing dypoly does it means that my new network interface should now do power on from the network driver | 12:02 |
sambetts | no | 12:02 |
sambetts | I'm saying their will be new function in the netowrk interace that will be called once the power is in the right state | 12:02 |
jroll | deploy driver: new_pre_power_off_networking_foo(); power_off(); current_networking_foo(); power_on(); done() | 12:04 |
jroll | yes? | 12:04 |
sambetts | more or less yup, we might need more points in the process, pre-IPA-shutdown, post-power on into tenant image, e.g. unplug provisioning network needs to be done before shutting down IPA, but plugging tenant network might need to be done after powering up into the tenant image | 12:05 |
moshele | sambetts: so the deploy will tell the network driver that power is on and the network driver will imply that the meaning is that it should plug the network? the network driver should not need to guess what to do based on power state. the deploy sets the flow, and the network driver does what it is told to do when it is told to do it. | 12:06 |
*** amoralej is now known as amoralej|lunch | 12:06 | |
sambetts | pretty much, which is why I wonder if this starts heading in the deploy steps direction | 12:08 |
sambetts | e.g. a network interface could just change the priority of the plugs and unplugs, wdyt jroll ? | 12:08 |
jroll | ... that might make sense | 12:09 |
jroll | again, don't have full context | 12:09 |
jroll | the basic premise is that smartnics need power to connect or disconnect a network, right? | 12:09 |
sambetts | yup, but the current deploy interfaces always power off the server before calling the network interface and there no way to tamper with that unless you make your own deploy interface | 12:10 |
jroll | right | 12:11 |
* jroll will try to have a think about it | 12:12 | |
* jroll also thinks we have a lot of spaghetti he'd rather untangle before introducing new spaghetti | 12:12 | |
sambetts | yeah... I feel like deploy steps making the deploy process composable pretty much solves this issue without introducing a whole bunch of extra driver functions etc to ignore/implement depend on what you need, however I am out of touch with the progress on that work | 12:13 |
jroll | basically we've got it done, but there's only one deploy step right now | 12:14 |
jroll | so we need to map things out and break it down | 12:14 |
moshele | can you point me to the deploy steps spec/ implementation? | 12:15 |
jroll | http://specs.openstack.org/openstack/ironic-specs/specs/11.1/deployment-steps-framework.html | 12:15 |
jroll | via http://specs.openstack.org/openstack/ironic-specs/ ctrl-f "steps" :) | 12:15 |
*** trown|outtypewww is now known as trown | 12:21 | |
moshele | jroll: so as I understand all the in tree deploy interfaces are now just one step, and the plan is the break them in the future right? | 12:27 |
jroll | moshele: correct | 12:27 |
jroll | there's also talk about making deploy 'templates' that define a set of steps - so you might have a 'raid5' template which will set up a raid 5 before deploying and a 'standard' template that does not | 12:28 |
*** rnoriega_ has joined #openstack-ironic | 12:38 | |
*** mgoddard has joined #openstack-ironic | 12:38 | |
*** jcoufal has joined #openstack-ironic | 13:01 | |
TheJulia | Good morning everyone | 13:01 |
sambetts | o/ | 13:01 |
*** rloo has joined #openstack-ironic | 13:05 | |
*** amoralej|lunch is now known as amoralej | 13:05 | |
dtantsur | morning TheJulia, jroll | 13:06 |
TheJulia | I'm kind of concerned about changing the api interface signifigantly with more logic. Why not have simeple helper methods that look at interface capabilities and do the needful dance through based on the helper. That allows us to have one logic location each pertinent major chunk that different deploy interfaces could call | 13:06 |
jroll | morning TheJulia and dtantsur | 13:06 |
TheJulia | s/more logic/more logic and more methods/ | 13:07 |
sambetts | something like driver.network_interface.needs_power_on = true/false, for the interface capability, then the deploy logic can call things at the right point? I still think that deploy steps priorities might be better, because then we could just make the priority of the plug X network step, before or after the reboot to guest OS step :/ | 13:09 |
TheJulia | That is kind of the same problem we're trying to solve, but the step would have to be aware that task.driver.network_interface.capabilities has 'smartnic' or something burried in the list, should ideally sort that, because then if it is entirely left to the operator to figure that on a template they define, then we're in for a world of hurt if we don't have detection logic burried all over the place to throw errors | 13:12 |
TheJulia | that their template might not work as desired. | 13:12 |
TheJulia | then again needs_power_on = true/false is another possibility :) Same meaning in the end, just different names and slightly different implementation details | 13:12 |
*** rh-jelabarre has quit IRC | 13:13 | |
jroll | TheJulia: I believe sam is implying the driver author defines the priority, not the operator | 13:13 |
TheJulia | so then we come up with two different priority matrixies? | 13:14 |
sambetts | I had imangined these would be steps that are defined by the interfaces with prioritys like the clean steps are (again I'm super out of touch) so in that case the smartnics network interface would define the same steps as the neutron network interface just with different priorities so depending on which interface is loaded on the node things occur at different points | 13:14 |
*** mjturek has joined #openstack-ironic | 13:14 | |
jroll | I don't think we've discussed operators defining deploy step priorities at all, only which steps happen or don't happen. but I could be wrong. | 13:14 |
TheJulia | I remember we even started discussing an api for it... | 13:15 |
TheJulia | but we also tabled all of that to get the actual steps work done | 13:15 |
TheJulia | so we can revisit and try and figure out what works, but tl;dr operators do want to be able to provide control of steps, invocation of optional steps, and the priority. The templates were to have inferring logic to take traits and turn them into actions as well | 13:16 |
TheJulia | those actions might be different steps being turned on/off too | 13:16 |
* TheJulia gets a headache | 13:16 | |
TheJulia | we should likely decompose this a little bit more to figure out what our real mvp is | 13:16 |
jroll | let's just throw everything out and only support the ansible driver | 13:17 |
TheJulia | lol | 13:17 |
jroll | then operators can do anything they want | 13:17 |
* jroll is sick of inifinite customization | 13:17 | |
TheJulia | but how do you support anything they want? :) | 13:18 |
* jroll steps away to play with dogs instead | 13:18 | |
TheJulia | I think the mvp was likely just traits turning on/off steps | 13:19 |
TheJulia | so it could be a "smartnic" trait about the node that we have | 13:19 |
mjturek | gm ironic! | 13:21 |
*** Bhujay has quit IRC | 13:21 | |
*** etingof has quit IRC | 13:21 | |
*** etingof has joined #openstack-ironic | 13:22 | |
*** galaxyblr has quit IRC | 13:28 | |
*** hjensas has quit IRC | 13:28 | |
rloo | good morning ironic'ers, mjturek, TheJulia, jroll, sambetts, dtantsur | 13:31 |
dtantsur | morning rloo | 13:31 |
rloo | TheJulia: doesn't this need a reno? (even the master patch) https://review.openstack.org/#/c/592247/ | 13:33 |
patchbot | patch 592247 - ironic (stable/rocky) - Fix not exist deploy image within irmc-virtual-med... - 5h 48m 56s spent in CI | 13:33 |
*** jcoufal has quit IRC | 13:33 | |
TheJulia | rloo: just yanked my +2, irmc results still not posted and yeah, since after 11.1 we do need a reno | 13:36 |
rloo | TheJulia: maybe add one in master and squash in the backport? whatever works, I am easy :) | 13:37 |
TheJulia | yeah, we can do that too | 13:38 |
TheJulia | we could also post the reno after the pach lands | 13:38 |
TheJulia | its not a big deal, the real issue is irmc not voting | 13:38 |
openstackgerrit | Merged openstack/ironic-specs master: import zuul job settings from project-config https://review.openstack.org/592398 | 13:45 |
*** jcoufal has joined #openstack-ironic | 13:47 | |
*** r-daneel has joined #openstack-ironic | 13:48 | |
*** dmellado has quit IRC | 14:01 | |
*** rh-jelabarre has joined #openstack-ironic | 14:06 | |
*** S4ren has joined #openstack-ironic | 14:07 | |
*** hjensas has joined #openstack-ironic | 14:09 | |
*** moshele has quit IRC | 14:10 | |
*** jiapei has joined #openstack-ironic | 14:16 | |
openstackgerrit | Merged openstack/ironic-python-agent stable/ocata: import zuul job settings from project-config https://review.openstack.org/592430 | 14:27 |
openstackgerrit | Merged openstack/ironic master: Minor fixes to contributor vision https://review.openstack.org/593736 | 14:27 |
*** cdearborn has joined #openstack-ironic | 14:38 | |
*** jrist has joined #openstack-ironic | 14:45 | |
*** mmedvede is now known as mmedvede_ | 14:55 | |
*** mmedvede_ is now known as mmedvede | 14:55 | |
*** tonyb has quit IRC | 15:04 | |
*** rnoriega_ is now known as rnoriega | 15:07 | |
*** pcaruana has quit IRC | 15:10 | |
*** rh-jelabarre has quit IRC | 15:11 | |
*** r-daneel has quit IRC | 15:23 | |
*** r-daneel has joined #openstack-ironic | 15:24 | |
*** scas has joined #openstack-ironic | 15:25 | |
*** openstackgerrit has quit IRC | 15:31 | |
*** cdearborn has quit IRC | 15:34 | |
*** dmellado has joined #openstack-ironic | 15:45 | |
*** openstackgerrit has joined #openstack-ironic | 15:46 | |
openstackgerrit | Merged openstack/ironic-inspector master: import zuul job settings from project-config https://review.openstack.org/592386 | 15:46 |
*** pcaruana has joined #openstack-ironic | 15:49 | |
*** gkadam is now known as gkadam-afk | 15:52 | |
*** tssurya has quit IRC | 15:53 | |
openstackgerrit | Merged openstack/ironic-inspector stable/rocky: import zuul job settings from project-config https://review.openstack.org/592460 | 15:56 |
openstackgerrit | Merged openstack/ironic-inspector stable/queens: import zuul job settings from project-config https://review.openstack.org/592448 | 15:56 |
*** rpioso|afk is now known as rpioso | 16:02 | |
rpioso | Good late morning, ironicers | 16:03 |
* rpioso was occupied downstream | 16:03 | |
openstackgerrit | Merged openstack/ironic-inspector stable/pike: import zuul job settings from project-config https://review.openstack.org/592437 | 16:05 |
openstackgerrit | Merged openstack/ironic-inspector stable/ocata: import zuul job settings from project-config https://review.openstack.org/592427 | 16:05 |
openstackgerrit | Merged openstack/ironic-inspector master: switch documentation job to new PTI https://review.openstack.org/592387 | 16:05 |
TheJulia | good afternoon rpioso | 16:08 |
rpioso | TheJulia: :) | 16:09 |
rpioso | And thanks, again, for your help getting the idrac UEFI bug fix merged everywhere. | 16:09 |
openstackgerrit | Merged openstack/ironic-inspector master: add python 3.6 unit test job https://review.openstack.org/592388 | 16:15 |
*** rh-jelabarre has joined #openstack-ironic | 16:28 | |
*** moshele has joined #openstack-ironic | 16:28 | |
TheJulia | rpioso: happy to help | 16:33 |
*** jiapei has quit IRC | 16:35 | |
rpioso | dtantsur: You, too :) | 16:45 |
*** mgoddard has quit IRC | 16:47 | |
dtantsur | you're welcome :) | 16:48 |
*** gkadam-afk is now known as gkadam | 16:49 | |
*** trown is now known as trown|lunch | 16:52 | |
*** gkadam has quit IRC | 16:54 | |
*** S4ren has quit IRC | 16:59 | |
*** dougsz_ has quit IRC | 17:00 | |
*** w-miller has quit IRC | 17:00 | |
*** gyee has joined #openstack-ironic | 17:03 | |
*** NobodyCam has joined #openstack-ironic | 17:06 | |
*** e0ne has quit IRC | 17:07 | |
*** rh-jelabarre has quit IRC | 17:15 | |
*** sambetts is now known as sambetts|afk | 17:23 | |
sambetts|afk | night all o/ | 17:23 |
*** etingof has quit IRC | 17:25 | |
*** rnoriega has quit IRC | 17:25 | |
*** sambetts|afk has quit IRC | 17:25 | |
*** mmedvede has quit IRC | 17:25 | |
*** jlvillal has quit IRC | 17:25 | |
*** dims has quit IRC | 17:25 | |
*** Guest92900 has joined #openstack-ironic | 17:34 | |
TheJulia | Goodnight! | 17:35 |
TheJulia | sadly it is not nap time... | 17:35 |
*** dims_ has joined #openstack-ironic | 17:36 | |
TheJulia | o/ dims_ | 17:36 |
*** rloo has quit IRC | 17:46 | |
*** mmedvede has joined #openstack-ironic | 17:58 | |
*** rloo has joined #openstack-ironic | 18:00 | |
*** etingof has joined #openstack-ironic | 18:01 | |
NobodyCam | hi | 18:12 |
NobodyCam | oh its working now | 18:12 |
NobodyCam | I was registered but still could not join :( | 18:12 |
TheJulia | o/ | 18:15 |
TheJulia | yeah, freenode did a whole sweeping restart yesterday it seems | 18:15 |
*** trown|lunch is now known as trown | 18:21 | |
*** dmellado has quit IRC | 18:22 | |
openstackgerrit | Julia Kreger proposed openstack/bifrost stable/pike: Only install libvirt-python and python-lxml via pip https://review.openstack.org/595280 | 18:27 |
openstackgerrit | Merged openstack/ironic-specs master: Remove the duplicated word https://review.openstack.org/592892 | 18:43 |
*** dtantsur is now known as dtantsur|afk | 18:43 | |
*** moshele has quit IRC | 18:46 | |
*** amoralej is now known as amoralej|off | 18:51 | |
*** mjturek has quit IRC | 19:11 | |
*** adrianc has quit IRC | 19:16 | |
*** cdearborn has joined #openstack-ironic | 19:32 | |
*** eandersson has joined #openstack-ironic | 19:42 | |
*** pcaruana has quit IRC | 19:47 | |
eandersson | What is the most place for nova-compute-ironic. Do most people just put it on the control plane together with other nova services? | 19:52 |
eandersson | I am guessing that compute_driver = libvirt.LibvirtDriver has no effect on services like nova-scheduler etc right? | 19:53 |
*** e0ne has joined #openstack-ironic | 19:57 | |
*** MattMan has quit IRC | 20:12 | |
*** tonyb has joined #openstack-ironic | 20:26 | |
*** e0ne has quit IRC | 20:27 | |
*** harlowja has joined #openstack-ironic | 20:42 | |
*** jcoufal has quit IRC | 20:49 | |
TheJulia | eandersson: people either run it on the control plane when they use ironic, or they run it on the nodes they are running ironic-conductor on | 20:53 |
eandersson | I see - thanks | 20:54 |
TheJulia | eandersson: nova-compute process feeds information into scheduling/placement | 20:54 |
*** trown is now known as trown|outtypewww | 20:54 | |
TheJulia | but in ironic's case, the virt driver distributes nodes using a hash ring | 20:54 |
*** jtomasek has quit IRC | 20:57 | |
anupn | Hi TheJulia: Any idea if we add proxies separately inside the glance image? I have added proxies including no_proxy to my env variables, still seeing ProxyError when IPA tries to download instance image registered with glance | 21:07 |
TheJulia | anupn: no idea, I think more information is required | 21:08 |
anupn | TheJulia: I am trying to launch an instance over BM and I see this error inside IPA logs http://paste.openstack.org/show/728634/ | 21:09 |
TheJulia | "Temporary failure in name resolution" | 21:10 |
TheJulia | no dns to reach a named proxy? | 21:10 |
TheJulia | or is the proxy an ip address? | 21:10 |
anupn | TheJulia: But i am able to curl www.google.com | 21:10 |
anupn | so assume nameserver should not be an issue? | 21:10 |
anupn | proxy is the URL | 21:11 |
anupn | XXX.intel.com | 21:11 |
TheJulia | what kind of ramdisk are you using? | 21:11 |
anupn | default that devstack use - agent tinycore | 21:12 |
anupn | TheJulia: And i see http_proxy is getting added while the request is made, but it is not adding 'no_proxy' | 21:13 |
TheJulia | I'm not sure anyone has tried the proxy code with tinycore before | 21:13 |
jroll | is the proxy hostname publically addressable? | 21:15 |
jroll | er, publically resolvable? | 21:15 |
anupn | jroll: Yes | 21:15 |
jroll | might be using a nameserver that doesn't know about the proxy's hostname | 21:15 |
jroll | hrm | 21:15 |
TheJulia | but tl;dr, from with-in the running IPA image, it needs to be able to perform a dns lookup to be able to connect to the proxy, it is basically saying it is unable to | 21:15 |
anupn | on running curl with that address, I see it getting connected | 21:15 |
TheJulia | running curl where? | 21:15 |
jroll | yeah, definitely a dns problem | 21:15 |
anupn | TheJulia: curl from the devstack machine, from where I am running nova commands | 21:16 |
TheJulia | Sounds like a DNS problem where your provisioning network is unable to resolve DNS | 21:17 |
TheJulia | or the DNS name of the proxy specificially | 21:17 |
anupn | TheJulia: Umm, may be I will check from within the network namespace if is able to ping the proxy server | 21:18 |
TheJulia | anupn: but it also depends on what neutron is handing out to on the provisioning network for DNS resolution | 21:19 |
*** cdearborn has quit IRC | 21:20 | |
anupn | TheJulia: But doesn't same route is used when downloading deploy kernel and ramdisk for provisioning, because that time I didn't see such error | 21:21 |
TheJulia | devstack uses ip addresses | 21:21 |
TheJulia | and ironic likely uses ip addresses as a result of devstack | 21:22 |
anupn | TheJulia: So do you mean it is better to use IP address for proxy? | 21:23 |
TheJulia | you won't have dns resolution issues if you use that, at least in development. It is good to make sure dns works though | 21:24 |
anupn | TheJulia: Ok, thanks. Will check that, and replace URL with specific IP, and see if it makes a diff | 21:26 |
*** dansmith is now known as htimsnad | 21:34 | |
TheJulia | htimsnad: is it almost friday? :) | 21:38 |
*** vcgi has joined #openstack-ironic | 21:40 | |
vcgi | I wonder what will be the best setup option for openstack ironic for a new enviroment. All servers have dual nics. Option 1: one for the Bare Metal Provisioning Network, and one for external connectivity. Option 2: nic 1, nic 2 bonded with multiple vlan setups. | 21:43 |
openstackgerrit | Merged openstack/ironic-python-agent master: switch documentation job to new PTI https://review.openstack.org/592394 | 21:43 |
anupn | TheJulia: in my resolv.conf, I see nameserver as my hostname itself | 21:45 |
anupn | TheJulia: I feel no_proxy should get added when IPA tries to make a request | 21:47 |
TheJulia | anupn: but your resolv.conf on your workstation is not what is present to the deployment ramdisk while running | 21:48 |
TheJulia | vcgi: guess it depends on your hardware and security posture/requirements | 21:48 |
TheJulia | vcgi: does your switch support LACP fallback to a single link ? | 21:49 |
anupn | TheJulia: Yes that's why I feel if no_proxy has has my hostname IP address is added to the ramdisk, might solve the problem | 21:50 |
anupn | know? | 21:50 |
vcgi | @TheJulia yes they do support lacp fallback to a single link | 21:50 |
TheJulia | vcgi: so concievably, you can do bonded/portgroup links with vlans. I think most neutron ml2 drivers lack the portgroup concepts though, so it would likely need to be manual static switch side configuration | 21:52 |
TheJulia | but then you can do vlan as access ports on the bonds I guess | 21:52 |
TheJulia | from a security standpoint, that would be ideal. A static back-end provisioning network is not really a great idea since only one of those nodes needs to be compromised to begin to allow the rest to be targeted and pivoted through. | 21:53 |
TheJulia | anupn: It doesn't work that way | 21:53 |
TheJulia | well | 21:54 |
TheJulia | anupn: if there is direct connectivity, then you don't need a proxy, then yes, that shoudl work. It won't fix dns resolution issues though | 21:54 |
* TheJulia thought you meant for the dns issue | 21:55 | |
vcgi | #TheJulia what if I don't use lacp for bonding ? | 21:56 |
TheJulia | vcgi: pxe booting is likely to fail. You can just do vlan enabled ports and not have one of the two ports connected to anything | 21:57 |
anupn | TheJulia: direct connectivity with what? | 21:57 |
TheJulia | vcgi: specifically PXE chain loading through tftpboot (with UDP packets) won't survive the traffic distribution algorithms that switches try and transmit down the wire | 21:57 |
TheJulia | anupn: direct connectivity between you the node (in this case, I believe it is your VM) that your trying to provision | 21:58 |
anupn | TheJulia: yes | 21:59 |
anupn | TheJulia: There is a direct connectivity in that sense | 22:00 |
*** livelace has quit IRC | 22:24 | |
*** rcernin has joined #openstack-ironic | 22:34 | |
*** livelace has joined #openstack-ironic | 22:37 | |
*** slagle has joined #openstack-ironic | 23:17 | |
*** rpioso is now known as rpioso|afk | 23:18 | |
*** r-daneel has quit IRC | 23:58 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!