| JayF | https://maixhub.com/discussion/100733 :( tl;dr whatever I read that said they had redfish/ipmi appeared to be wrong or future looking | 03:26 |
|---|---|---|
| JayF | still cool but not useful for ironic purposes seemingly | 03:26 |
| JayF | it's basically a RISC-V linux server though, so it's possible if not implemented | 03:27 |
| opendevreview | Verification of a change to openstack/ironic master failed: dedup reboot request in redfish bios path https://review.opendev.org/c/openstack/ironic/+/933020 | 05:05 |
| rpittau | good morning ironic! o/ | 07:44 |
| rpittau | JayF: nice toy you got there :D | 08:26 |
| iurygregory | good morning ironic | 11:40 |
| dtantsur | TheJulia: I could find https://helm.sh/blog/storing-charts-in-oci/ which refers to https://specs.opencontainers.org/distribution-spec/?v=v1.0.0 | 12:57 |
| dtantsur | which seems totally different from what we discuss, at least at the first glance... | 12:57 |
| dtantsur | hmm, at the second glance, it may be the same thing but via the ReST API | 12:58 |
| dtantsur | interestingly, helm uses the oci:// scheme, not docker:// | 13:00 |
| dtantsur | I wonder if it's a different name or a completely different approach | 13:00 |
| dtantsur | TheJulia: the claim to use https://oras.land/ (which is something Red Hat participates in, apparently) | 13:01 |
| dtantsur | This ORAS thing claims to support basically all popular registries, which sounds promising | 13:02 |
| dtantsur | https://oras.land/docs/how_to_guides/pushing_and_pulling looks easy to use | 13:04 |
| dtantsur | TheJulia: can we just say we use ^^^ and stop caring about the exact implementation? | 13:04 |
| rpittau | that looks interesting | 13:07 |
| TheJulia | good morning | 14:22 |
| TheJulia | I'll have to look once I've caffinated | 14:22 |
| dtantsur | morning TheJulia | 14:35 |
| JayF | rpittau: yeah, it's neat, but I'm sad it doesn't redfish outta the box | 15:06 |
| opendevreview | Julia Kreger proposed openstack/ironic master: docs: note ipv6 is a good idea with neutron interface https://review.opendev.org/c/openstack/ironic/+/936951 | 15:49 |
| opendevreview | Julia Kreger proposed openstack/networking-baremetal master: DNM: Test CI https://review.opendev.org/c/openstack/networking-baremetal/+/936955 | 16:46 |
| rpittau | good night! o/ | 17:22 |
| TheJulia | dtantsur: so, that might work, and is actually very similar. It hides the top level details and uses the docker python api client to do all interactions though. The model of use is a little different, so I guess the question is what are we going to prefer. I'm really digging the idea of an oci url with a sha hash witch points to the manifest for a payload for highly specific items. | 18:11 |
| TheJulia | oras seems to largely be modeled on a specific version tag to be the driving delineator. Where as with similar prior art they kind of have a model for the hidden selection details. Oras seems like it is kind of all front-loaded on the tag but podman's got all the hidden details like architecture selection hidden under the hood of the container. | 18:21 |
| TheJulia | The other big aspect difference is podman is applying zstd to like... everything | 18:43 |
| TheJulia | so the type is that, instead of the native type | 18:43 |
| cardoe | TheJulia: when I get back. Just randomly popping in while getting warm at a cafe in Prague. | 21:50 |
| TheJulia | Looking at oras-py library, definitely not compression aware compared to what podman does, so if we want to align with other practices, the question is much more along the lines of "how do we want to store the blob in the store", and and the details is much more an implementation detail since it is just walking the second->third layer data. The only thing which we might want todo which is different is if we determine we | 21:50 |
| TheJulia | don't want to try and walk into the third layer to grok it. I think at the end of the day, Ironic shouldn't prefer or require specific tags, maybe assume "latest" if you lack one. Challenge with oras is the file name is *separate* from the container store, so you need to know your hunting for my_file_in_the_store.txt upfront... really | 21:50 |
| TheJulia | cardoe: enjoy! | 21:50 |
| TheJulia | The tag is just the starting point though | 21:51 |
| TheJulia | I'm pretty sure, no matter what we do, we're likely best off writing our own client code. The print and the "removal of our own client" stuffs from the oras-py in the commit log doesn't make me feel great. | 21:59 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!