janders | good morning Ironic o/ | 00:04 |
---|---|---|
opendevreview | Steve Baker proposed openstack/sushy-tools master: novadriver, don't use instance.flavor.id https://review.opendev.org/c/openstack/sushy-tools/+/840768 | 05:22 |
opendevreview | Steve Baker proposed openstack/sushy-tools master: Fix per-request memoization https://review.opendev.org/c/openstack/sushy-tools/+/840780 | 05:22 |
opendevreview | Steve Baker proposed openstack/sushy-tools master: Document SUSHY_EMULATOR_ALLOWED_INSTANCES https://review.opendev.org/c/openstack/sushy-tools/+/840804 | 05:22 |
opendevreview | Steve Baker proposed openstack/sushy-tools master: Use libvirt automatic firmware for UEFI boot mode https://review.opendev.org/c/openstack/sushy-tools/+/841653 | 05:22 |
opendevreview | Steve Baker proposed openstack/sushy-tools master: Add SecureBoot support to the emulator, all drivers https://review.opendev.org/c/openstack/sushy-tools/+/842266 | 05:22 |
arne_wiebalck | Good morning Ironic! | 06:15 |
opendevreview | Bernd Mueller proposed openstack/bifrost master: add sha256sum files for ipa images where to search for the offline git variables https://review.opendev.org/c/openstack/bifrost/+/842092 | 06:55 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost master: Remove support for all distributions with Python 3.6 https://review.opendev.org/c/openstack/bifrost/+/842272 | 07:36 |
opendevreview | Bernd Mueller proposed openstack/ironic-inspector master: Add DnsmasqFilter error - my fsm encountered an exception https://review.opendev.org/c/openstack/ironic-inspector/+/842275 | 07:39 |
dtantsur | the time has come folks: https://review.opendev.org/c/openstack/bifrost/+/842272 | 07:39 |
opendevreview | Dmitry Tantsur proposed openstack/ironic master: devstack: use CentOS 9 for DIB IPA builds https://review.opendev.org/c/openstack/ironic/+/842277 | 07:46 |
opendevreview | Dmitry Tantsur proposed openstack/ironic-python-agent-builder master: Remove support for openSUSE since Python 3.6 is no longer supported https://review.opendev.org/c/openstack/ironic-python-agent-builder/+/842278 | 07:49 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost master: Remove support for all distributions with Python 3.6 https://review.opendev.org/c/openstack/bifrost/+/842272 | 07:51 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost master: [WIP] Use libvirt-python from the OS https://review.opendev.org/c/openstack/bifrost/+/842282 | 08:05 |
janders | good morning arne_wiebalck dtantsur and Ironic o/ | 08:11 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost master: Use libvirt-python from the distribution https://review.opendev.org/c/openstack/bifrost/+/842282 | 08:34 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost master: Use libvirt-python from the distribution https://review.opendev.org/c/openstack/bifrost/+/842282 | 08:46 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost stable/yoga: Use libvirt-python from the distribution https://review.opendev.org/c/openstack/bifrost/+/842302 | 09:48 |
dtantsur | please check https://review.opendev.org/c/openstack/bifrost/+/842272, needed to fix bifrost master | 09:53 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost stable/yoga: [WIP] Use libvirt-python from the distribution https://review.opendev.org/c/openstack/bifrost/+/842302 | 10:02 |
waleedm | Hi guys, I'm working on a solution to update a fw and set some config using hradaware_manager in cleaning stage from a config json file, What is the best way to pass some parameter to IPA image, e.g. the url or the locateion of the json file ? | 10:31 |
dtantsur | waleedm: for a manual clean step you can actually pass parameters directly | 10:40 |
waleedm | @dtantsur, what if I want to add an autoamated clean step, where is the right place to start from ? | 10:47 |
dtantsur | hmm. ideally we should have something like driver_info, but for ports.. | 10:55 |
dtantsur | janders, arne_wiebalck, do you have opinions on ^^? | 10:55 |
* dtantsur will be in a training for the rest of the day | 10:55 | |
opendevreview | Radosław Piliszek proposed openstack/bifrost master: [CI] Test Kolla on Ubuntu https://review.opendev.org/c/openstack/bifrost/+/842332 | 10:58 |
iurygregory | good morning Ironic | 11:02 |
janders | hey iurygregory | 11:12 |
janders | waleedm (CC dtantsur) I don't think I have an opinion but I recall similar functionality in code: https://opendev.org/openstack/ironic/src/branch/master/ironic/drivers/modules/ilo/management.py#L582 - while I *think* this may be for manual clean step it may be a good reference - maybe? | 11:17 |
iurygregory | janders, o/ | 11:26 |
iurygregory | ok, funny we need to rebased https://review.opendev.org/c/openstack/ironic-python-agent/+/837039 to have it working... | 11:40 |
opendevreview | Jacob Anders proposed openstack/ironic-python-agent master: Multipath Hardware path handling https://review.opendev.org/c/openstack/ironic-python-agent/+/837039 | 11:58 |
iurygregory | ty janders \o/ | 11:58 |
jssfr | hi there! | 12:57 |
jssfr | where can I read about integration of neutron + ironic with SDN-capable switch hardware to facilitate configuration of layer 2 networks for metal as a service in openstack? | 12:57 |
*** mat_fechner is now known as matfechner | 13:18 | |
TheJulia | rpioso: o/ | 13:19 |
TheJulia | rpioso: Are there any support version matrixes of known tested/supported versions of idrac <-> ironic ? | 13:19 |
TheJulia | o/ jssfr | 13:19 |
TheJulia | jssfr: are you seeking steps? engineering mechanics? overview? etc. I guess trying to figure out where would be the best to send you | 13:20 |
jssfr | TheJulia, ah. some overview would be nice, steps might be good to understand how things play together, too. | 13:53 |
jssfr | context is that I just saw a presentation of a partner company where they want to provide HW-a-a-S, but with completely homebrewn automation and I'd like to understand if Ironic could be better used there | 13:53 |
TheJulia | Ahh, hmm | 14:03 |
jssfr | so far we only used ironic as a "dumb but not really" way to provision machines in something somewhat like an undercloud | 14:04 |
* TheJulia finds ironic docs on oracle.com | 14:07 | |
* TheJulia blinks | 14:07 | |
TheJulia | jssfr: https://ironicbaremetal.org/blog/bare-metal-neutron-ml2-video/ <-- overview https://docs.openstack.org/ironic/latest/admin/multitenancy.html <-- technical details of configuring. In that it lists Neutron ML2 drivers we're aware of.. but we likely need to revise. hjensas is working on netconf support, and there was a networking-ansible tool to basically facilitate the switch level configuration. | 14:10 |
jssfr | TheJulia, that document looks like what I was looking for as a starter, thanks a lot | 14:11 |
TheJulia | it is quite technical though, there are a few other things out there which might be good or useful. Not every ML2 driver for SDN integration has support for the VNIC_BAREMETAL type which is used in neutron | 14:13 |
jssfr | well, the folks involved are thankfully all very technical :) | 14:13 |
TheJulia | \o/ | 14:13 |
TheJulia | One final note, some SDN drivers are in weird states where the vendors are seeking to turn them into profit centers instead of just update/fix them | 14:14 |
jssfr | I had heard about that somewhere | 14:14 |
TheJulia | so.... Community solutions may be better but YMMV | 14:14 |
jssfr | non-networking question: If they just want to boot a live OS with ironic, they could do that via the rescue mechanism, right? | 14:16 |
jssfr | that would fire up iPXE which would then chainload that live OS? | 14:16 |
jssfr | (they had concerns about image sizes with plain PXE, but that should be no issue with iPXE inbetween) | 14:16 |
TheJulia | jssfr: for a already deployed node, that would be a little difficult in that your attempting to rescue something booted from disk. The deployment interface you want is the ramdisk interface | 14:16 |
jssfr | ah, ramdisk, thanks | 14:17 |
TheJulia | Just about everyone has concerns with image sizes and using raw PXe | 14:17 |
TheJulia | PXE | 14:17 |
jssfr | no wait, the ramdisk method just refers to using the IPA to deploy something? or does that allow to boot into a (e.g. glance-provided) image without touching the disks? | 14:18 |
TheJulia | no | 14:18 |
jssfr | (they're doing kernel development, they generally don't want to install stuff) | 14:18 |
TheJulia | ramdisk is "I want to book this kernel and ramdisk" after which we'll consider the machine active | 14:18 |
jssfr | ah, hm | 14:18 |
TheJulia | that kernel/ramdisk can be a live image | 14:18 |
jssfr | hmmmmmmmm | 14:18 |
jssfr | right | 14:18 |
jssfr | that should be similar to standard PXE, except that it doesn't use TFTP | 14:19 |
TheJulia | yes | 14:19 |
TheJulia | most machines will chain load through PXE just the iPXE binary, and from there it is all HTTP | 14:19 |
jssfr | perfect | 14:20 |
TheJulia | And yes, you *can* setup https, but you need to compile your own ipxe binaries | 14:20 |
TheJulia | I've only known a handfull of operators who have actually done that | 14:20 |
jssfr | yeah, we've got that on our roadmap. | 14:21 |
jssfr | but that's for another day :) | 14:21 |
TheJulia | the thing with ramdisk, is you'll want to ensure you know how your going to login or have credentials embedded | 14:21 |
TheJulia | or say an ssh public key... | 14:21 |
TheJulia | otherwise you'll need to use Metadata services | 14:22 |
TheJulia | it *is* possible to setup, fwiw | 14:22 |
opendevreview | Radosław Piliszek proposed openstack/bifrost master: [CI] Test Kolla on Ubuntu https://review.opendev.org/c/openstack/bifrost/+/842332 | 14:22 |
TheJulia | With ramdisks, we don't have configuration drives. | 14:22 |
jssfr | they have Pis associated with the nodes which serve as KVM and tunnel the serial console through and stuff | 14:22 |
TheJulia | Just something to keep in mind | 14:22 |
jssfr | yup, understood | 14:22 |
jssfr | though kernel parameters could be passeds | 14:23 |
TheJulia | cool cool | 14:23 |
jssfr | *passed? | 14:23 |
jssfr | (an ED25519 pubkey may be just small enough...) | 14:23 |
TheJulia | yes, I believe the option is instance_info/kernel_append_params | 14:23 |
jssfr | <3 | 14:23 |
TheJulia | enjoy! | 14:23 |
jssfr | let's see if I can prevent them from succumbing to NIH | 14:23 |
TheJulia | good luck! | 14:24 |
jssfr | thanks! :) | 14:24 |
opendevreview | Verification of a change to openstack/networking-generic-switch stable/yoga failed: CI: use pre-existing ssh key on multinode jobs https://review.opendev.org/c/openstack/networking-generic-switch/+/841472 | 15:21 |
arne_wiebalck | dtantsur: sorry, was afk | 15:31 |
arne_wiebalck | dtantsur: the question is what we use to pass params to the IPA? | 15:32 |
arne_wiebalck | dtantsur: waleedm: in our deployment we use "driver_info" for 'official' parameters which can be set (i.e. upstream code, like burn-in), but also the "extra" field on the node to pass some info to downstream-only clean steps ... both work for us and I am not super strict on where things go either :) | 15:35 |
* arne_wiebalck is not sure if this was actually the question ... | 15:35 | |
iurygregory | https://review.opendev.org/c/openstack/ironic-python-agent/+/842093 if possible can someone +W so we can fix our CI in bugfix 8.4 ? =) Ty | 15:43 |
iurygregory | dtantsur, are you against moving metalsmith job to non-voting in bugfix8.1 https://review.opendev.org/c/openstack/ironic-python-agent/+/841293 ? | 15:48 |
JayF | iurygregory: looking | 15:51 |
JayF | landed | 15:52 |
iurygregory | JayF, tyvm! | 15:53 |
JayF | np | 15:55 |
opendevreview | Jeff Young proposed openstack/sushy-tools master: aarch64/arm64 libvirt VMs should default to SCSI controller https://review.opendev.org/c/openstack/sushy-tools/+/842385 | 16:05 |
gmarcy_ | if I run ironic all the time on one of my homelab subnets to keep track of my metal machine inventory would I have the ability to augment the dnsmasq config to add non-metal machines on that subnet? | 16:09 |
dtantsur | iurygregory: not too much, but why? | 16:10 |
jssfr | oh no, they might prefer their 2.5kLOC of home-brewn "production ready" haskell-code | 16:12 |
jssfr | *sigh* | 16:12 |
dtantsur | iurygregory: I think I can try pinning keystone as well, hold on | 16:12 |
dtantsur | I really wonder why it affects only metalsmith, but not other devstack jobs.. | 16:13 |
TheJulia | jssfr: *sigh* | 16:14 |
jssfr | but they'd really like to use neutron for the networking which I don't see how that'd work | 16:14 |
jssfr | ah well, meeting next week, let's see what I can do. | 16:15 |
dtantsur | iurygregory: mm, very interesting. devstack jobs use keystone stable/xena even though we didn't explicitly specify it. | 16:15 |
JayF | jssfr: that's like saying "I don't need a hilt, I'll just grab the sword by it's blade" | 16:15 |
iurygregory | dtantsur, ack, wow devstack *magic*... | 16:15 |
jssfr | JayF, well, given it's haskell and not nodejs or so, I'd say it's more like grabbing it at the crossguard. ;-) | 16:16 |
TheJulia | JayF: I believe the Forged in Fire judges would not approve. | 16:16 |
JayF | jssfr: I was referring to neutron without ironic ;) | 16:16 |
jssfr | oh, yeah | 16:16 |
jssfr | that's definitely grabbing it at the blade. | 16:16 |
*** gmarcy_ is now known as gmarcy | 16:16 | |
TheJulia | ++ | 16:17 |
TheJulia | err | 16:17 |
TheJulia | wrong window | 16:17 |
TheJulia | well, that was *wrong computer* | 16:17 |
opendevreview | Dmitry Tantsur proposed openstack/ironic-python-agent bugfix/8.1: Use Xena constraints for bugfix/8.1 https://review.opendev.org/c/openstack/ironic-python-agent/+/841293 | 16:21 |
dtantsur | iurygregory: let us try this approach ^^ | 16:21 |
dtantsur | now, where was I wrt fixing the bifrost CI? | 16:21 |
iurygregory | dtantsur, tks I will keep an eye on it | 16:21 |
gmarcy | dtantsur I got rid of the trailing underscore, does that make me an IRC power-user? for my homelab question would bifrost be better than just running the BMO tools/run_local_ironic.sh ? | 16:23 |
opendevreview | Verification of a change to openstack/ironic-python-agent master failed: Multipath Hardware path handling https://review.opendev.org/c/openstack/ironic-python-agent/+/837039 | 16:24 |
iurygregory | NO CI! | 16:24 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost master: Use libvirt-python from the distribution on Red Hat systems https://review.opendev.org/c/openstack/bifrost/+/842282 | 16:25 |
TheJulia | and sadness spread across the land | 16:26 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost stable/yoga: Use libvirt-python from the distribution on Red Hat systems https://review.opendev.org/c/openstack/bifrost/+/842302 | 16:26 |
iurygregory | I'm getting a deja vu with the failure.... | 16:26 |
* dtantsur hates libvirt-python | 16:26 | |
dtantsur | also folks, looking for some reviews to fix the bifrost CI: https://review.opendev.org/c/openstack/bifrost/+/842272 | 16:29 |
* TheJulia glares at google's presentation tools | 16:32 | |
gmarcy | dtantsur there are copy/paste errors in your CI fix | 16:32 |
dtantsur | gmarcy: mm, I think I know what you're talking about | 16:33 |
gmarcy | dtantsur kernel where you want initramfs | 16:33 |
dtantsur | yep. fixing. good catch! | 16:33 |
gmarcy | trying to setup login now | 16:33 |
gmarcy | didn't have ubuntu-one acct | 16:34 |
iurygregory | I love the fact that the unit tests failed in 3.8 during gate but it was green in check... | 16:35 |
iurygregory | maybe we have some race condition? .-. | 16:35 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost master: Remove support for all distributions with Python 3.6 https://review.opendev.org/c/openstack/bifrost/+/842272 | 16:40 |
dtantsur | gmarcy ^^ | 16:40 |
dtantsur | I wonder if anyone will be sad if we remove the suse support completely | 16:40 |
gmarcy | nice factoring out the distro - reads cleaner | 16:41 |
dtantsur | yeah, should have done that from the start | 16:42 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost master: Use libvirt-python from the distribution on Red Hat systems https://review.opendev.org/c/openstack/bifrost/+/842282 | 16:46 |
opendevreview | Dmitry Tantsur proposed openstack/bifrost stable/yoga: Use libvirt-python from the distribution on Red Hat systems https://review.opendev.org/c/openstack/bifrost/+/842302 | 16:46 |
gmarcy | just fyi, I've been using centos-8-stream and installing python39 explicitly. still finding some gaps in support for 9-stream where it has not reached "stable" status | 16:53 |
gmarcy | thinking it may be time to make the switch to 9-stream and just fix what I have problems with | 16:54 |
opendevreview | Merged openstack/ironic-python-agent bugfix/8.3: SoftwareRAID: Use efibootmgr (and drop grub2-install) https://review.opendev.org/c/openstack/ironic-python-agent/+/826914 | 17:25 |
opendevreview | Merged openstack/ironic-python-agent bugfix/8.4: Run Jobs on bugfix 8.4 https://review.opendev.org/c/openstack/ironic-python-agent/+/842093 | 17:25 |
TheJulia | gmarcy: I was thinking similar recently... just $time | 17:27 |
iurygregory | https://zuul.opendev.org/t/openstack/build/88f0090869314cfaad5bc82b966307d6/logs wtf zuul "This build does not provide any logs" O.o | 17:29 |
gmarcy | thinking more on how to use ironic best in my homelab. if I have a machine to dedicate to that role (not a big machine and something without a BMC so out of the game already) would it be best to run ironic in containers, a VM or a SNO cluster? | 17:31 |
gmarcy | trying to think about what gives me the most flexibility to add on more services | 17:32 |
iurygregory | by more services you mean openstack services? | 17:38 |
gmarcy | or metal3 | 17:41 |
gmarcy | something along the lines of airship/starlingx to take the available h/w and use it to provision k8s or openstack deployments | 17:43 |
gmarcy | not sure if my homelab is big enough to get too fancy | 17:44 |
* TheJulia raises an eyebrow that the foundation is tweeting a translated version of an article I wrote | 17:44 | |
gmarcy | TheJulia - that seems... uncool | 17:45 |
TheJulia | I mean, we'll all be there in a few weeks so... I should be working on my short german responses | 17:45 |
TheJulia | but yeah | 17:45 |
TheJulia | I think homelab wise, there are just a ton of variables. | 17:47 |
TheJulia | Personally, mine has been off for ages | 17:47 |
TheJulia | I think what killed it was the ethernet interface on my NUC dying | 17:48 |
TheJulia | like... no longer shows up in lspci/lshw | 17:48 |
TheJulia | NOR the bios | 17:48 |
gmarcy | interesting thing about my homelab h/w is that all the BMC machines have wi-fi power monitoring switches | 17:52 |
TheJulia | nice | 17:53 |
TheJulia | I... am seriously considering building a huge battery bank at home, and thinking how an I turn on some micro-inverters to suppliment my power utilization from the battery bank | 17:54 |
TheJulia | and it basically will all come down to needing contractors I can easily control | 17:54 |
TheJulia | But at that point, I might as well just wire it up old school | 17:54 |
gmarcy | so ideally I would power all the machines on, register them with ironic, then power them off until needed. but that would mean I would (1) need ironic to remember I still exist and (2) when I need to access the BMC that it would run a script to make power available to it | 17:54 |
TheJulia | there *has* been some chatter about making a command or script interface for things like power | 17:55 |
TheJulia | since NobodyCam had this... different case | 17:55 |
TheJulia | When he explained it, it kind of made sense, but it was still like "wow, did not expect that!" | 17:55 |
jssfr | (that would be really handy… we have a bunch of legacy hardware which is power-controlled via a inhouse API which controls some arduino switching relais) | 17:55 |
TheJulia | NobodyCam: you should like... join this discussion | 17:56 |
NobodyCam | yawn... huh...???... | 17:56 |
TheJulia | NobodyCam: command interface for power | 17:56 |
jssfr | (we're gradually decommissioning it, buuut there's still some deployments where we can't easily, and making those usable with ironic would be very interesting for ephemeral test setups) | 17:56 |
TheJulia | since you have the "need to turn off one port turn on another and flip/flop the state to undo" | 17:57 |
NobodyCam | oh for controlling evil devices | 17:57 |
NobodyCam | yes! | 17:57 |
* TheJulia is amused they are labeled "evil devices" | 17:57 | |
jssfr | hey! ;-) | 17:57 |
TheJulia | That wouldn't be like my flipper zero sort of evil device right? | 17:57 |
* TheJulia is sufficiently evil good to use "evil" | 17:57 | |
NobodyCam | kinda like that... not quite ;) | 17:58 |
* jssfr googled flipper zero | 17:58 | |
jssfr | what am I even looking at | 17:58 |
jssfr | is this like the usb killer but in cude | 17:58 |
jssfr | *cute | 17:58 |
* TheJulia giggles | 17:58 | |
NobodyCam | we have devices attached to pdu that could benefit from being able to be controlled outside of the nodes power process | 17:58 |
jssfr | also, does it run rockbox | 17:58 |
TheJulia | jssfr: custom firmware it looks like in C that is actually kind of easy to read | 17:59 |
TheJulia | and of course, open source | 17:59 |
jssfr | I was more wondering if there's a rockbox port for it ;) | 18:00 |
TheJulia | no idea | 18:00 |
jssfr | oh, an stm32 | 18:00 |
TheJulia | Now that I have sucessfully derailed the conversation! | 18:01 |
* TheJulia gives herself a gold star | 18:01 | |
jssfr | you did! | 18:01 |
NobodyCam | hehheehe | 18:01 |
jssfr | well done! | 18:01 |
TheJulia | so, command interface for power | 18:01 |
jssfr | but in all seriousness: a command interface for power would be pretty neat actually | 18:01 |
TheJulia | volunteers?! | 18:01 |
TheJulia | should be minimal code really | 18:01 |
TheJulia | like, super thin vaneer levels of code | 18:01 |
jssfr | we've had this on our rough todo list for a while now, but as it only affects legacy hardware ... it isn't really worth it for us to *develop* it | 18:01 |
TheJulia | this is the kind of thing I *might* do while I'm in the air or something, but... I'm spread so thin at the moment with the summit coming up I may just curl up in a ball in my uncomfortable economy seat and try not to drink the entire way | 18:02 |
NobodyCam | I can see it becoming a bit more then a vaneer | 18:03 |
NobodyCam | ie. say the node is ipmi / redfish for deploy and main power but secondary port is snmp? | 18:03 |
jssfr | if I was deeper in the ironic codebase, it could be something for the train ride, but I suppose this time it'll be too large a group for working anyway | 18:03 |
TheJulia | NobodyCam: I feel the need to suggest dinner tonight, so I can sit at a table and glare ;) | 18:04 |
NobodyCam | LoL | 18:04 |
TheJulia | fine, layered vaneers | 18:05 |
NobodyCam | ;) | 18:05 |
* TheJulia raises an eyebrow that it looks like her flight schedule has changed... again | 18:06 | |
NobodyCam | I keep wanting to leverage the chassis object, but that is never the answer | 18:06 |
* TheJulia begins to wonder if NobodyCam created the chassis object | 18:07 | |
* NobodyCam oh look a squirrel | 18:08 | |
TheJulia | I guess... ground squirrels count. | 18:10 |
NobodyCam | hehehehe | 18:10 |
TheJulia | Speaking of.... I've not seen any this year | 18:10 |
NobodyCam | we have bunnies | 18:10 |
TheJulia | But we've got a cat that is hanging out in our yard | 18:10 |
TheJulia | bunnies++ | 18:10 |
TheJulia | I just watched one hop by the window | 18:10 |
NobodyCam | ++ | 18:10 |
jssfr | :-O | 18:10 |
TheJulia | jssfr: if it wasn't obvious, NobodyCam and I live fairly close to each other | 18:11 |
jssfr | but bunnies | 18:11 |
TheJulia | okay, corporate travel and airline say the same thing. App on my phone doesn't... :\ | 18:13 |
NobodyCam | if I under stand the original issue there are nodes with bmc that are connected to a snmp like device that pulls all power from the node so the bmc is powered off and unable to respond | 18:14 |
jssfr | (the original issue *here* is that there's no BMC whatsover with those devices :)) | 18:15 |
NobodyCam | in order to provision the node we would need to engage the power for the node; then use ipmi/RF to control the node and deploy | 18:16 |
NobodyCam | oh so only pdu controlled | 18:16 |
TheJulia | NobodyCam: I guess that is also because there is bmc firmware updates, it might be absolutely totally necessary to toggle the device's complete power | 18:16 |
NobodyCam | ++ | 18:16 |
NobodyCam | The use case I have is a little different, where I would like to power off secondary devices durning deploy / teardown then turn when node is active | 18:18 |
TheJulia | NobodyCam: going to SCaLE ? | 18:18 |
NobodyCam | I we thinking of how we could could `chain` power actions across more then one power port | 18:19 |
NobodyCam | I had not planed on it | 18:19 |
TheJulia | interesting | 18:19 |
TheJulia | power "steps" | 18:19 |
NobodyCam | +++ | 18:20 |
TheJulia | FWIW, we're going to go to SCaLE in July in order to plot appropriate world domination. | 18:21 |
NobodyCam | nice | 18:21 |
TheJulia | Although, we're not going to do anything like take the RV. | 18:21 |
NobodyCam | when Isi t? | 18:22 |
NobodyCam | is it even | 18:22 |
TheJulia | I feel like it is july 27th-31st | 18:22 |
TheJulia | or soemthing like that | 18:22 |
TheJulia | Hmm, there is an RV park nearby... | 18:22 |
TheJulia | LA county, and I get a 404 when loading it. Woot! | 18:23 |
gmarcy | here is a picture of what all my h/w is plugged into - https://imgur.com/aGxCebk | 18:24 |
TheJulia | gmarcy: nice | 18:24 |
gmarcy | includes servers and network switches | 18:24 |
NobodyCam | nice | 18:24 |
gmarcy | if I don't need a 10gbe network for the workload I'm running I can just turn it off | 18:25 |
NobodyCam | Alex ... deploy ... hummmm | 18:25 |
NobodyCam | Alexa | 18:25 |
gmarcy | so idea would be to get all of the h/w "under management" and not need to keep the power on. when there is a request that needs say "server2" it can run a "power-on server2" script | 18:26 |
gmarcy | BMC should be accessible very quickly after that and can confirm that the h/w hasn't been changed vs. what was last seen | 18:27 |
gmarcy | don't mind if it takes a few steps to get it into service, saved all that electricity when it wasn't in use | 18:28 |
TheJulia | I am not worried about the electrons at the moment.... | 18:28 |
gmarcy | interesting, last time I talked to them they seemed worried about you ;) | 18:29 |
TheJulia | well... I'm pushing a *ton* of power to the grid now | 18:29 |
gmarcy | see, that's what not running your homelab gets you ! | 18:30 |
TheJulia | Hmm, only pushing 7kw at the moment... I guess batteries are still charging | 18:30 |
TheJulia | generating 12.5kW | 18:31 |
jssfr | from? solar? | 18:35 |
TheJulia | yeah | 18:37 |
TheJulia | If they manage to push through the solar tax that is being talked about, we may go completely off-grid and add some wind at some point | 18:46 |
TheJulia | too many unknowns | 18:46 |
jssfr | ... solar tax | 18:46 |
jssfr | do I want to know? | 18:46 |
TheJulia | no | 18:47 |
TheJulia | Just the utility companies trying to remain super profitable | 18:47 |
TheJulia | and prioritize themselves over the future of the planet | 18:47 |
jssfr | ah then all is well /s | 18:48 |
opendevreview | Merged openstack/ironic-python-agent bugfix/8.3: Use a pre-defined partition UUID to detect configdrive on GPT https://review.opendev.org/c/openstack/ironic-python-agent/+/840470 | 19:38 |
opendevreview | Julia Kreger proposed openstack/ironic-tempest-plugin master: WIP: Secure RBAC Test Concept https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/842427 | 19:42 |
TheJulia | hjensas: so ^^^ is kind of what I'm thinking, the needful for like project reader as executed under tempest is that it can't create stuffs. | 19:43 |
TheJulia | hjensas: and that if it tries to read, then the list is empty | 19:43 |
TheJulia | hjensas: I *think* we're going to need to teach the baremetal client code in the plugin to "expect failure" | 19:43 |
hjensas | TheJulia: cool, I'll take a look after call with jparoly. | 19:44 |
TheJulia | hjensas: lmk if you have any questions or need additional insight, it didn't take me too long to hack that up, presently thinking of how to handle the "expect failure" | 19:45 |
iurygregory | TheJulia, I'm a bit puzzled with the unit tests failures in https://review.opendev.org/c/openstack/ironic-python-agent/+/837039/ https://zuul.opendev.org/t/openstack/status#837039 running locally I can't reproduce... | 19:53 |
TheJulia | uhh, I guess the global var use is trashing it | 20:00 |
TheJulia | hmmmm | 20:04 |
TheJulia | yeah, so that test does try to exercise the global, but it can *always* get reset before then | 20:06 |
TheJulia | or even mid-flight | 20:06 |
TheJulia | since it is a global | 20:06 |
TheJulia | and that whole unfortunate boundrary between them | 20:07 |
TheJulia | iurygregory: is it just happening with py39 | 20:12 |
TheJulia | or.... others? | 20:12 |
TheJulia | iurygregory: my only real thought at the moment is to remove the one test | 20:16 |
TheJulia | if we're sure it works otherwise | 20:16 |
TheJulia | some things are just not "easy" to test | 20:16 |
opendevreview | Verification of a change to openstack/ironic-python-agent bugfix/8.1 failed: Use Xena constraints for bugfix/8.1 https://review.opendev.org/c/openstack/ironic-python-agent/+/841293 | 20:40 |
stevebaker[m] | good morning | 21:22 |
stevebaker[m] | TheJulia: hey I had an evil idea and I need to be talked out of it. Its not urgent if you don't have the spoons | 22:13 |
TheJulia | At the vet with a cat… maybe a little later | 22:17 |
stevebaker[m] | ok all good | 22:18 |
opendevreview | Merged openstack/ironic-python-agent stable/xena: Use a pre-defined partition UUID to detect configdrive on GPT https://review.opendev.org/c/openstack/ironic-python-agent/+/840348 | 22:18 |
hjensas | TheJulia: ETOTIRED, I will get sleep and look at the RBAC tempest stuff tomorrow. | 22:34 |
iurygregory | TheJulia, before it happened to py38... (sorry was out doing exercises) | 22:35 |
TheJulia | iurygregory: I say drop the test, minimal value and I’ll even +2+A | 22:42 |
TheJulia | Well, the offending test | 22:42 |
opendevreview | Steve Baker proposed openstack/sushy master: Update jobs names https://review.opendev.org/c/openstack/sushy/+/838662 | 22:46 |
opendevreview | Verification of a change to openstack/ironic-python-agent master failed: Multipath Hardware path handling https://review.opendev.org/c/openstack/ironic-python-agent/+/837039 | 22:50 |
iurygregory | ok, let me do that | 23:09 |
opendevreview | Iury Gregory Melo Ferreira proposed openstack/ironic-python-agent master: Multipath Hardware path handling https://review.opendev.org/c/openstack/ironic-python-agent/+/837039 | 23:26 |
iurygregory | now if something fails I will have to flip my table | 23:27 |
iurygregory | (╯°□°)╯︵ ┻━┻ | 23:27 |
TheJulia | heh | 23:33 |
TheJulia | stevebaker[m]: so...w hat do I need to convince you not to do... or what should I put my evil hat on and encourage? | 23:33 |
stevebaker[m] | TheJulia: ok, so. Neutron based DHCP allows the initial binary to be entirely determined by the node's drivers, but dnsmasq based dhcp like with bifrost is basically static. You have to choose at install time whether to serve all nodes iPXE, grub or whatever. | 23:39 |
stevebaker[m] | TheJulia: How about we implement a dnsmasq based BaseDHCP? There could be a dnsmasq.d enabled with dhcp-optsdir and per-node files written which match on the MAC, tag with the node UUID, then do a dhcp-boot specific to the boot interface. It doesn't even require a SIGHUP after writing the file | 23:39 |
TheJulia | intruiging | 23:40 |
TheJulia | somewhere between lawful evil and chaotic good | 23:40 |
stevebaker[m] | I haven't yet looked into how new dnsmasq needs to be to support dhcp-optsdir. My fedora man page documents it, random internet man page doesn't | 23:40 |
stevebaker[m] | this is prompted by a dtantsur comment in this review https://review.opendev.org/c/openstack/bifrost/+/807220/32/playbooks/roles/bifrost-ironic-install/templates/dnsmasq.conf.j2#100 | 23:41 |
TheJulia | stevebaker[m]: what is documented on thekellys ? | 23:42 |
stevebaker[m] | what is a thekellys? | 23:42 |
stevebaker[m] | ah | 23:43 |
stevebaker[m] | its there https://paste.openstack.org/show/b6u1pFv2XsOSIJI7qLkC/ | 23:44 |
stevebaker[m] | https://thekelleys.org.uk/dnsmasq/docs/dnsmasq-man.html | 23:44 |
TheJulia | cool | 23:44 |
stevebaker[m] | It looks like it'll be in centos8 onwards | 23:45 |
stevebaker[m] | TheJulia: alright, I'll put it on the backlog. The code might be simple compared to getting everything to integrate with it | 23:49 |
stevebaker[m] | this would help metal3 also I think | 23:52 |
stevebaker[m] | TheJulia: ok, back to eating lunch, thanks | 23:53 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!