VMware Velero Community Meetings, 28 Jul 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Velero Community Meeting/Open Discussion - July 28, 2020

Description

July 28, 2020

Status Updates
carlisia
Community support this week
PTO next week
Bringing the Helm repo up to date with v1.4 + more Velero Helm TLC
Continue work on the ServerStatusRequest migration to kubebuilder
ashish
Restore hooks:
Adding constants PR#2750
Updating RestoreSpec PR#2761
Backing up pod volumes using restic by default:
PR 2611
Excluding volumes mounting secrets and config maps
Opening Issue 2770. Support for label selector to choose pods to which the defaultVolumesToRestic should be applied. :thumbsdown: / :thumbsup: in the issue
PR reviews
nrb
Working on some CI improvements around assignment/reviewers, but ran into issues with how GitHub Actions expose secrets
PR reviews
Intending to get DeleteItemAction design doc updated today or tomorrow
Was on community support last week

Discussion Topics
Welcome JenTing Hsiao (https://github.com/jenting) as an official maintainer of the Velero Helm chart!
Let us know what parts of the Velero API you use - https://github.com/vmware-tanzu/velero/issues/2654

Contributor Shoutouts
@areed: Add hooks fields to restore context
@thejasbabu: Add Start and Completion Timestamp to Restore Status
Helm Chart
@jenting: fix: helm upgrade did not work correctly on BSL and VSL configuration
@jenting: helm 2,3 CI test + Upgrade CRDs to v1.4
@MrFreezeex: Fix Helm 2 CRD generation
@cesarokuti: add schedule labels
@tomahawk28: Helm 2,3 supports CRD at the same time
@ckotzbauer: add affinity option to velero deployment

A

Hello, everyone and uh welcome to another episode of the valero community meeting, slash open discussion today is july, 28, 2020 and uh per usual we're going to start with status updates and then dive into the discussion topics. Let me share my screen, so you all can see the meeting notes and let's kick it off all right. First, off on status updates, we have.

B

Calisia guess my first status update is that I have a new computer and, oh my god, the difference.

B

I was severely handicapped man.

B

So I'm on community support this week. I changed with ashish, because I uh planning on taking pto next week and I spent some time giving the pilero helm repo some tlc.

B

It is up to date now with version 1.4 of alero, because um it wasn't and um there's also news related to that we should be more, have more bandwidth to be on top of that, repo with you know, addressing requests or anything that comes up and besides that so support and continue finishing up the service status, request migration to queue builder.

B

That's it oh yeah! That's it.

A

Awesome, do you have anything that you would like some help with this week before you head off on pto.

B

Not particularly now.

A

B

A

uh Any questions comments for kalisia.

A

All right, uh ashish.

B

Oh, I think I should shut the skip.

A

C

Yeah yeah, it looks like he's on debt duty.

A

C

A

uh Do you want to cover any of the items there nolan, or do you want to jump over to yours.

C

um Just real quickly what he's got written here, I know he's been doing some stuff for getting the restore hooks work. Going um he's made some tweaks to the backing up pod volumes using rustic by default. um Some folks uh raised uh some issues with this.

C

In the slack channel. They were testing it and found um some some uh secrets and config maps were being included, and so they requested uh those get excluded. um This was the original pr, but then uh ashish made a follow-up this last week and uh excluded secrets and config maps, um and then uh I think yeah, I don't remember the number.

C

I think it's back on the update.

C

Maybe he didn't link it yeah, but yeah people people have been testing maine, which thank you very much. um Also he opened uh 2770, which was a discussion about uh how do we include some some of the pods uh pod volumes for rustic, um so there was discussion on slack and uh we would definitely like input from folks on this.

C

um It pertains largely to um figuring out if we want to add a new method for um selecting um volumes for inclusion in rustic, or if um this this was asking for a way to support um inclusion using our our standard tags or our I'm sorry, our standard and our standard inclusion logic that we already have so um yeah. If this, if this using label selection on against pods to get everything attached to it, um using the default default, rustic flag would be useful. uh Please vote on that.

D

um Hey question so uh at one point there was a move to move restick to be like a peer of the volume snapshotters, so it it's not like special anymore. So would we want to have the same mechanism for selecting rustic then.

C

um At that point, I think the selection would change. um I think steve had tried that and it was kind of like a a research spike. um I would have to double check with steve to see where that's at, uh but that's that's not merged. That's not like actual.

C

That that was investigative- and I don't know the current status of that, um but if we took rustic and made it a volume snapshot, I think that would make a lot of this stuff less special, um so yeah it would um it would. I think it would be a bigger design change than what this issue is proposing.

D

But if we're going in that direction anyway,.

C

Right, I don't know that we're going in that direction yet, um but it's worth thinking about now that undoing this wouldn't be a big deal right.

C

This particular one. uh I don't think so. This one yeah this one is, I think, basically, the the genesis was a user was asking uh hey. Can I use? um Can I use a pods labels to include um the volumes that are attached, which I think default volumes to rustic is going to get most of what they wanted anyway, which will come in 1.5, um which basically says if you have that vault.

C

If you have that option on any volume, any pod volume mounted on the pod will be backed up by rustic, with some caveats like inline volumes won't be won't, be um included um and free-floating pvs that aren't mounted to anything won't be included, but you won't have to do the annotation anymore.

D

Right that actually sounds good. That principle of least surprise.

E

And dave we are always taking a look at feature: enhancement, requests and new net new feature requests, um and so this is one that we have in the backlog. It's currently labeled as unprioritized, and just because we do have a few items prioritized for the next releases, but provided that there's bandwidth, um you know, and the amount of requests that we're getting in for a similar type of issue from the community.

E

We're always taking a look uh to make sure that we're triaging the priorities appropriately but sounds like you know the um current rustic functionality that's going to go out with 1 5 the improvement there should address a large amount of some of the use cases folks are looking for. So um we'll continue to make improvements on restic. You can see on our roadmap that we've actually got uh resting improvements to be defined that will be included in 1-6 also, so this may make it.

D

Sure sounds good.

B

Cool, so one refresher about that that issue that we we have open that we steve started looking at, we started talking about for making rustic work just like a another plug-in the trade. There is a major trade-off there, which is on one hand, but in a positive way. Things will be uniform. On the other hand, then we will have. We will lose some features that people use, so that really needs to be looked at.

B

I really like the idea of having things uniform, because we just keep these two parallel things and it's a lot of work and in rustic is always behind, but we have to see if it's worth losing the the features that we'll be losing, which I don't remember off the top of my head.

C

Yeah, I, the these things, are incremental building on top of what we have, whereas the the move to putting rustic in a plug-in volume, snapdragon plug-in proper, is a bigger lift and one that I think would take longer. So I think it's something when you we need to investigate and like properly define the trade-offs.

C

Excuse me and communicate folks uh hey this. Is you know this is going to affect data so like I, I don't want to make that decision now or say that that's what we're doing, I'm just going to put it out there that it it's an option for later, but, like I'm, not gonna, I don't wanna, say one way or another whether it would be like one, six, one, seven, one: eight, um it's uh yeah!

C

It's it's something we're thinking about, but I'm not sure exactly what all would change, because it's been a while, since I've looked at it.

D

Sure, fair enough, no, I just remember that we would talk. You know I'd heard about that that going on so I just wanted to see if that was still the thrust.

C

Yeah, I think that kind of that that kind of lost some steam, but if it's, if it's interesting to folks, we can, we can see if that slots in as because, as stephanie mentioned, like the road map item, is kind of very broad. It's improvements to rustic, uh so that may be an improvement uh we would have to we'd have to see.

F

um No, then this is a vishnu, I, I think probably one of the previous meetups I was saying but uh missed out putting an issue yeah well I'll, put it there. Basically, if you have a mixed set of volumes which have their own vulnerable gains.

F

Applying rustic uh or not, I mean it's like I think, right now it says it applies to everything every volume so.

F

I'll put out issue, I mean um I remember now I forgot to put it I'll, do it.

C

Okay, yeah yeah um was the concern that you don't want it to apply to every volume.

F

Yeah yeah, if um let's say I have a few local volumes and then few ab ebs volumes, so I mean it's, it may better to offload which you have a plan for it, rather than uh applying for every volume monitor to the pod.

C

In in your deployment cases, have you seen um users with multiple storage drivers like that.

F

um Mainly in the cloud case where we have uh aws and abs disk will be there and few, the local host path will be used for logs to store that kind of things which are kind of okay or debugging kind of things.

C

Okay, that makes sense that makes sense. Then.

B

C

Yeah, if you could file an issue for that, then we can. We can follow up there, um yeah yeah, I think, most times, um folks aren't really mixing and matching uh drivers, but in the in that case, where you have like a local um like a local volume that is like scratch or logs, and then you have the cloud for your actual actual data, um you could lean on the plug-in for that so yeah. If, if we get an issue for that for enhancing this feature, that would be awesome.

C

F

C

If we don't have anything else, I think that's it, for ashish's updates sounds good.

A

uh You want to continue with yours.

C

Yeah um so for mine, I was doing um a lot of internal stuff for some products uh that are utilizing valero on the open source front. I was looking at some ci improvements, hoping to get some workflow stuff automated.

C

Unfortunately, some of that doesn't work with github actions due to the way it manages secrets.

C

I was really hoping we could get some github actions that, like assigned reviewers and assigned to signees and all that, uh but it doesn't look like it's gonna work. The way I was hoping so I'll have to put that to the side for now um been doing pr, reviews and issue triage. um I haven't gotten your issue, yet I'm sorry, I will I'm going to start working from the oldest um as I can.

C

um I did not. We we had discussed last week about the delete action, slash, delete, item action, design, doc. I had not gotten a chance to get back to that and revise it.

C

um I'm intending to get back to that and either update it today or tomorrow, more likely tomorrow, if I'm being realistic, but my intention is to take that update it so that it will be backwards compatible with how valero works today, which is to say, if you don't, have any deletion item or delete item action, plugins installed, it'll just work the way it does today and then, if you do.

C

Give the download the the backup tarball and give each relevant item to the plugins that are interested, um but I will get that design, doc updated um and then I was on community support last week and I don't I don't think I have a whole lot else to bring up here.

A

Do you have anything that um you would like to reviewed or any blockers this week.

C

We've got the um restore hooks work going on, and uh I know uh andrew I don't know if he's on the call doesn't look like uh andrew was working on that, um so that that stuff should be that that's been moving moving along pretty well um ashish and andrew have been driving a lot of that. I did a little bit of it.

C

um Did a lot of the helm stuff? Thank you carlesia, so I don't think I have anything in particular at the moment. That's that's. A burning need to get reviewed. um The the delete item. Action thing is like uh the highest thing on my list and that's on me to get updated. So I don't believe so.

A

All right sounds good. Anyone have any questions, comments or no one.

A

All right, then, we're diving into the discussion topics.

A

Who wants to take on the first one.

B

A

B

So we noticed that I apologize if I'm mispronouncing, which I probably am genting, was uh doing a phenomenal job on the helm, repo. He was asking people to make changes that were needed. He was opening prs. He seemed very knowledgeable and very engaged, so it was very natural and easy for us to decide to invite him as a maintainer which he accepted, and now we have besides the valero.

B

Maintainers we have two external maintainers for our helm, ripple, genting and carlos. So between the three of us, we knocked down a bunch of uh prs and so we're. I think, we're in a good place now we're not at zero inbox zero. Yet so I think we're gonna get there pretty soon.

A

Fantastic news.

A

I think this one is your versus walter alesia.

B

Oh yes, so this is a recur I spaced out. So this is a recurring issue that we're going to have in the discussion topics.

B

Every meeting um we we weren't very intentional about making codes not exposed to public use, but now we're trying to change that and put our codes inside the internal folder for people who are familiar with go know that that's just internal doesn't get exposed, doesn't get, is not accessible to be used as libraries, so we do have some some code that are intentionally made public for people to reuse and it's listed on this uh issue.

B

If you are using codes, if you are using there as a library and importing code that is not in those packages, please let us know we are. We are just trying to avoid breaking code for anyone. Well, what dresses? Somehow?

B

First of all, we need to know.

A

Awesome, thank you any uh does anyone else have anything that they would like to bring up on today's call before we dive into the contributor shout outs.

B

I just wanna say one thing: I completely forgot to compile a list of people who made contributions on the home repo. I am just realizing. I should have done that. But after we finish, I will do that and we'll do that next week.

C

Nope, it's there. It's already there.

B

Oh he's already, there.

C

I updated my little shout out script.

B

Oh wow awesome.

A

All right, let's uh dive into them,.

C

um So the first first couple, um we've, like I said andrew reed, um has been working on the restore hook stuff, so um the the first round of of restore hook, work has been on um just adding fields to our our data structs. um So this is kind of laying groundwork. It's not the actual, restore hook um logic, but um thank you very much andrew for engaging there, and I know he and she should have been discussing stuff on other pr's.

C

The next one was from the josh.

C

This was addressing another open issue that someone had filed about um adding start and completion timestamps to restores um which we had on backups, but we didn't add to restores so um this pr was very much appreciated because this that allows people to look at how long a restore took so um very much appreciated. Thank you, um the helm chart, I don't know carly cf, you wanna you're a little bit closer to these. So I don't know if you want to go through these.

B

Yeah, I can't I can't go through that. um So the first one was uh it's just when you upgrade the release of a release of a helm chart, there were things that needed to to be included in our chart to make it work that were in there. So that's what this pr did. It's just some pre-hooks, some post hooks and some other changes so yeah.

B

That was very helpful, uh so that was by genting in another pr that he submitted and got merged was to actually do the upgrades to bring the uh the current helm chart to 1.4, because there were some crds that changed and another thing you say on this pr. No so so he also spruced up the test. The ci test and up.

B

Here I don't remember what that is.

C

Oh, this one was the the white space issue.

B

Oh yeah yeah. Yes, you know it's it's a small pr, but that makes all the difference.

B

So that was by mr friezack's physics, uh hard to tell what the name is arthur: okay, uh so cesar submitted spr to add schedule, tags schedule, labels.

B

And so this was neat pr, so before we had the crds for when we migrated to home, three, we needed to have a directory with the crds for home too, in another directory would serve this for him three, so this person found a way to just have one directory, which is perfect, so that cleaned up things a little bit so greatly appreciated tomahawk 28.

B

Okay, and that was also adding a label uh adding having the ability to add a configuration for dfinity, not a label, an affinity configuration for the deployments yeah, that was uh there was our great those were all great pr's and uh yeah. Like you see you as you see, there were a bunch of them and we have about five. Now a couple are duplicates we'll see. Maybe we'll get them knocked out this week.

A

Awesome, thank you. Does anyone else have anything to add before we close up the meeting.

A

All right well have a fantastic rest of the week. Everyone have a good one and see you next week. Bye, bye, bye,.