Kubernetes Storage SIG, 9 Mar 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes SIG Storage Meeting 2023-03-09

Description

Kubernetes Storage Special-Interest-Group (SIG) Meeting - 09 March 2023

Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.ao8nq01jg6b7

Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage

Moderator: Saad Ali (Google)

A

All right today is March 9 2023. This is the meeting of the kubernetes storage special interest group. As a reminder, this meeting is public recorded and posted on YouTube uh today on the agenda, we're going to go over the 127 planning and uh if you have anything you'd like to discuss whether it's PRS that you want to get attention for designs that need review or anything else, please feel free to add it to the agenda um in terms of 127 planning.

A

The next upcoming Milestone is on March 14th, coming up fairly quickly, which is code freeze, and so we're going to go through the current items that we're tracking for the 127 release and get status updates on those.

A

So I'm going to switch over to our planning spreadsheet, add a column here for the current date.

A

And we can get started um first item we have is from haymont I, don't know if he's on the line.

B

A

Hello, uh so first item is recovering from resize failures: open apis to unblock quas.

B

Yes, I have have peers open for uh that updates. The API to newer versions like removing the resize status enum string and creates a map, so I would welcome any reviews, it's being tracked accurately, but I'm still waiting on reviews.

A

Awesome. Thank you. Next is issues related to assuming volumes are Mount points. uh Anyone have an update on this I'm, not sure if Jing is on the line.

B

Okay, I think the last thing that is left on this one is the Reconstruction assumes things are one point and there's a peer out to remove that code once that is done, I think we can close that once that Pierre merges then I think we can close this as done.

A

A

Awesome cool, thank you. Himan next is volume group API.

C

Right so we're trying to cut a CSS back release. Do you think we are in agreement here? Can we.

A

C

A

Think I have no concerns from my end, I was waiting for you and uh James to see if you all are okay moving forward.

C

Yeah I think he has uh they've given plus one this morning. I see.

A

Here, oh perfect, okay,.

C

Yeah so I think we should be fine. Yeah sounds.

A

Good okay, then I will go ahead and plan to cut today. Okay,.

C

Great yeah, so we do have some um follow-up actions just to take a look of the script. Change I've moved to GitHub actions, I, don't know if that has affected this. uh You know the compare comparison of the generated output, I, don't know what happened there so.

D

C

Take a look at that yeah um and also in addition to the CSS back release. uh uh We also have the API um definitions, PR merged um and the work is in progress to write the controller and a sample. Csi driver implementation.

A

Awesome, this is I, think the flagship feature for this release. So thank you for pushing a shank thanks. um All right next up is provision volumes from Cross namespace uh snapshot. uh Pvc continue, Alpha work, um any updates on this one.

D

uh No I haven't seen anything here: I'll go ahead and ping the um feature owner to don't check if they're gonna have time to work on the remaining items sounds good.

A

Thank you Michelle. uh Next up we have CSI volume, Health, additional metrics and or events uh anything new on this one.

C

uh No, unfortunately, not I have not seen the E3 test being updated. It's still working progress got.

A

E

B

A

uh Next, we have change block tracking. uh This was designed for this cycle. Anything new here.

C

uh Yeah, we still are having design discussions on this. uh There is a new approach that we are discussing, so uh we're going to do some pocs and continue with that.

A

Got it cool sounds like good progress. Thank you. uh Next, we have new rwo access mode being moved to Beta um Michelle. Any updates on this one.

D

um I think the feature is pretty much ready to go all the PRS are in, and the dock, the dock and the um blog PR's our placeholders are opened.

A

Awesome sounds like this is pretty much uh pretty much done. Yeah.

D

I, don't know if we have a status almost done.

A

I think we can say make this one green and leave this one yellow sure uh all right cool. Thank you. Michelle next is runtime assisted mounting.

A

uh I think this is the same uh same as last time. I haven't seen any updates on the CSI spec side for this one so copy that over uh next we have CSI proxy for Windows transition to privileged containers out of tree. uh Any update on this.

A

Okay, I'm gonna mark this one as no update moving on. We have node expansion secret, um so I think last status here was end-to-end test and review anything new here.

C

The test is submerged.

B

C

um But I think we need more. This is just one test got it yeah.

A

Cool. Thank you. Shane next is uh SC Linux relabeling using Mount options. Csi driver API change.

A

uh The last status was we'll need to move feature gate to Beta. Anyone know the status of this xion.

F

Yeah so today, I submitted actually the pr to flip it to Beta the features and I submitted also a few end-to-end tests and like that I promised in the.

A

Some thank you. John.

A

uh Next, we have CSI migration related things, uh so remove entry GCE. This is a placeholder for 128, so we'll skip that uh we have remove entry, AWS EBS, anything new on this one yeah.

E

You can mark this one as complete um the pr for it was merged. Nice.

B

F

Provider is fine, removing clouds with the feature gate if their feature gate being Alpha.

F

F

Provider, they have their own feature gate to remove the cloud providers and it's still Alpha, and it seems it's okay- to remove Cloud providers without graduating that feature to GA. It works for me, I'm. Just a bit surprised.

A

Got it does anybody have any more information on that.

E

Yeah so um I attended the Sig Club provider meeting and um they were okay with removing it as part of that PR. um So yeah.

F

E

Also found it a bit surprising, but we decided to move forward with that.

F

Okay thanks a lot.

A

Cool, thank you both. uh So next one is just for tracking we'll skip over that to remove entry, Azure disk. Any updates on that one.

F

I think you didn't see any PRS.

C

There was a PR submitted by editing.

F

C

It should be reviewed.

C

uh Let me find a tag you there.

A

Cool thank you for the update uh we'll skip over the next, because we uncommitted to them for this cycle. Next is control volume, conversion between source and Target, PVC.

A

um I think the remaining work here was docson blog. uh Any any changes here.

C

Yeah, that's still the same uh at the placeholder PR is submitted for the blog.

A

Got it cool? Thank you, Shane uh next is actually let me put this as almost done.

A

uh Okay, next is better default storage, class, GA and 128 tracking work. uh uh This was just for tracking, so, okay, let's skip that and robust volume manager, reconstruction.

F

So the feature has been flipped to Beta. Well, not free. It was introduced as beta and I have one PR pending sick Note review. I am adding some metrics to their code and I need their approval, which is hard to get actually the other, so yeah I'm bringing people we will see in the worst case. We do not have the metrics and we will have the feature gate as beta interesting.

A

Michelle, do we have someone from cignode who might be able to help here.

D

um This is for reviewing the Reconstruction change, correct.

F

No, no uh reviewing the metrics for or you know we have these log messages orphan Port found because I could not remove the directory because of blah blah blah. So I did metric for that. So we don't need to grab logs. We can look at the metric okay.

D

um I can find some node folks to take a look.

F

That would be nice. I will send it to the chat.

A

All right, thank you. Both next is PV. Last phase transition time um looks like cap was merged, but no updates. Since anything new here.

A

Okay, Mark that as no update.

A

All right, the next set are co-owned between Sig storage and other sigs. First, we have address issue PVC created by stateful set, will not be Auto removed.

A

um No update. Last time is Matt carry on the line or anybody yeah, hey Matt, yep,.

G

Hey I'm here sorry picture there, um so the um uh the um so I'm I'm trying to um summarize the the issue has been getting the end-to-end tests working, um it doesn't seem feasible to use kind So. The plan is to do a um manual run of the uh GCE and end tests with the um mutation detector enabled um and see. If we can then see if we can deploy the pdcsi driver with the cloud provider gcp end-to-end tests, um so that when this is beta, it will be uh tested.

G

um If, if we can get this test stuff sorted out, then we should be good to enable the feature gate, which is the last um PR that needs to be done so I'm trying to get this done now before the freeze next next week,.

A

Got it sounds like a complicated set of tests? uh Thank you for working on that Matt yeah. Next is volume expansion for stateful sets um and we haven't had an update here for a while a month. Do you have anything here? Maybe.

B

uh No nothing today I'll try to find some time to review it. Okay, no.

A

Worries. Thank you. uh Final uh item here is non-graceful node shutdown. Oh, this was for tracking, so we can skip over this. Okay. I think we are done with status updates. I'll go back to our agenda. Doc uh looks like we have one miscellaneous uh item: K does gcrio is deprecated move to registry.k.io uh there.

G

A

Significant image pulls for AWS uh driver no driver, registrar, liveness probe, um who posted this. This.

D

Is me sorry, I forgot to put my name no.

A

D

Yeah, so the kubernetes in for folks are, you know, trying to get people to switch from the deprecated kates.gcr.io to the registry.cutes.io and trying to get users to stop using the old one.

D

um They did a recent I guess analysis and they found that these three images are like top 10 images being pulled.

D

um I think they're, trying to figure out how to best deal with this I think the two approaches they're looking at is either just straight up deleting the images um or they're trying to look at implementing some sort of redirection, um but yeah I think either way both of them have the potential to break users.

D

um So yeah I think this is just a FYI that everyone should be moving off of these registries.

A

Yeah, thank you for the uh important note there Michelle um call out for everybody on the call. If you are aware of the usage of any of these, please please um try to take action on these as soon as possible.

D

Like try to update all the documentation everywhere already so I'm suspecting these are just like. You know: people running older versions of the tripod and haven't upgraded yet.

H

Wait can anyone hear me yeah? Oh okay, sorry it wasn't letting my name yeah for what it's worth over at the AWS and yes, CSI driver we're aware of this. um We have a warning at the top of our readme and in previous release notes.

H

You know telling people that it's possibly going to go away, um but obviously, if you're CS, if you develop a CSI driver here, you know you very likely, presumably use the node driver, registrar or live news probe, so you know worth making sure that you don't have any instructions floating around or manifests using the old images.

H

um But you know we've been warning people for as long as we have been aware of this issue, and um you know, unfortunately, some people's clusters might break eventually if it gets cut off. But you know, if that's what they have to do to save the money for the infrastructure. You know that's what they have to do.

H

um We've tried to warn people for as much as possible.

D

H

E

Good yeah I.

D

Think you know from most of the our sidecars and also the drivers, we've already updated the manifesting documentation. So it's really a matter of people, sort of upgrading and um getting off of the old usage.

C

I think there is a one thing we need to check is when we cut the release. We we have this container location, I, don't think we went back there and changed that some of those are still the old location so like that.

D

D

I see yeah, we might need to go back and update all our release. All release, notes, yeah.

C

I can't yeah I can take a look at that.

A

Cool. Thank you uh for the follow-up they're saying good call out any other uh questions. Comments. Concerns related to this.

A

Okay, if there's nothing else, we can end uh early today and give you all about 40 minutes back. Thank you. Everyone.

B

D

Thank you yourself. Thanks.