VMware Velero Community Meetings, 4 Feb 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Velero Community Meeting/Open Discussion - Feb 4, 2020

Description

Feb 4, 2020
Agenda

Status Updates
----------------
Steve
prepping for v1.3.0-beta.1 release - will be pushed out later today
need to see if we should release new versions of the cloud provider plugins
working on some refactoring in the restore code related to restoring CRDs

Ashish
Coming back from community support
Checking in on https://github.com/vmware-tanzu/velero/issues/2069
Have some troubleshooting steps

Carlisia
On community support
CLI install redesign - https://github.com/vmware-tanzu/velero/pull/2202#issuecomment-577229282

Nolan on PTO

Discussion Topics
------------------
[carlisia] Drop CRDs manifest but install CRDs by velero install? https://github.com/vmware-tanzu/helm-charts/issues/44
[jonas] Would it make sense to create an ADOPTERS.md file, where people and organizations can PR themselves in? Similar to Linkerd and Harbor?
[carlisia] How to best deprecate CLI cmds, add/remove the word new?: https://github.com/vmware-tanzu/velero/pull/2202#issuecomment-577229282
[carlisia]
velero [resource] [operation] [flags]
OR
velero [operation] [resource] [flags]
https://github.com/vmware-tanzu/velero/pull/2202#issuecomment-581961860
[dymurray] Unfortunately need to miss the community meeting today but hoping to have a design doc PR up by EOD tomorrow for https://github.com/vmware-tanzu/velero/issues/1027

Contributor Shoutouts
----------------------
@jenting - prune unknown fields on Velero CRDs (https://github.com/vmware-tanzu/velero/pull/2187)
@cpanato - allow velero restore create --from-schedule to work for schedules in other clusters (https://github.com/vmware-tanzu/velero/pull/2218)
@cpanato - use GA topology labels on PVs by default (https://github.com/vmware-tanzu/velero/pull/2219)
@koehn - fix panic in restic status updating for empty output (https://github.com/vmware-tanzu/velero/pull/2231)
@shaneutt - add arm/arm64 container builds (https://github.com/vmware-tanzu/velero/pull/2227)

A

Hi everyone and welcome to this week's episode of the valar community meeting, slash open discussion today is February 4th, 2020 and per usual. We got a bunch of status updates and then discussion topics, so, let's dive into it. I'll share my screen, so you all can see the the meeting notes and if you want to follow along at home or in the chat rather.

A

All right, first up, we have Steven.

B

Yeah so so yesterday, spended part of the day getting ready to release the 1.3 beta. So for those who.

A

B

Back I got a spam call for those who weren't here last week, so we have a number of bug fixes particularly related to CR d backups and restores, but also a bunch of other things that have been merged into master since 1.2, and we decided that wow that, rather than waiting until the end of March or shipping, a 12.1 release with some of those bug fixes. We just push out a 1.3 with, with all of the fixes and small features that we've gotten in since 1.2 you so anyway.

B

The basically I've been getting the changelog and the docks and other things ready for the beta. So it's just about ready and I'm planning to push that release out later today. After this call, so that'll give everyone something to test for a while and will mediate any issues there and then get the one. The official 1.3 out within I'd say the next week or two.

B

One thing that we need to figure out is: if we want to release new versions. The cloud provider plugins I took a look at the chat. The commit history for these, and really the only substantive change to any of them was that we change. We made that fix to the the group that the container is run under I. Don't think we necessarily need to push out new releases just for that change, because they've all been working. Fine, as is so I, think it's probably fine to just wait on those until we have more changes.

B

So that's that's the beta and then I was also I spent. The afternoon yesterday, looking at doing some refactoring in the restore code, specifically around some of the fixes we put in for CRD restores so Nolan and I had talked over the last couple of weeks about a few different approaches to to the implementation of the code that we put in there and so I'm. Just looking at doing some, some cleanup there that I think will make the code a little bit simpler and an easier to maintain.

B

So I'll probably continue working on that this week and I hope to get a PR up, but that's not something that I'm planning to try and get into one up three and that's about all for me all.

A

Right I mean question is on the one two three beta release here: the refactoring.

A

Question for you Steve D, so this is mostly a bug fix release. Do you think we should do a blog post about this one or mainly point people to the release Docs for.

B

I think we'd be fine either way I mean we might be able to no one wrote up a doc at the end of last week about the the handful of CRD issues that we were dealing with so could make sense to just take some of that and sort of put it into a blog post, just to help kind of explain what was going on, because there were like five or six different interrelated issues that were all affect things here: ID backups and restores.

B

But beyond that yeah I mean it's. You know it's a bunch of smaller bug, fixes and small features, so nothing major worth blogging about. But well you know we'll see if Nolan's up for putting together something next week when he's back yeah.

A

That makes sense, I think just notifying people. Although the major changes, the maker, fix's, okay, cool awesome, let's see, is Ashish on the call, no he's not so skip him for now so Kelis yeah.

C

So, as a end of day yesterday, I was caught up with peer reviews continued on top for the rest of the week on community support, I spent some time mentally blocked with that see a live design and but I am finally pushing the going to push the that they are.

C

The main updates will be a a set of details that, in the meeting we had to talk about this example of the kapalama files as soon consumed by customize it, which will be, which will represent the plugins for the deployments, so each layer will a separate day for each of the plugins that can be used. So have a couple examples of those and lastly, I still have to reveal the suggested see allies. I Steve listed on that PR just to see.

C

If what there is in the idea we can incorporate here, but I think this is shaping up well the way it is, though, so Apple questions on the discussions. They are minor, but also significant. So when we get to that section can go over there and that's it.

A

All right, awesome and Nolan Ison PTO this week.

D

Everyone on the lashes, I I was on the call, but I didn't know how to unmute myself on the phone. So now I figured the night.

A

D

C

D

Support from last week, I fired a few follow-ups, I have on the memory consumption issue that I think Anthony opened up. I, don't know, if he's what we call. So a quick update on that I have some more coming shooting steps on the issue.

D

So if if this is seen again, if folks can follow the troubleshooting steps and get back to us with more information, that would be super useful and there will be only update. I had yeah.

E

Then this is Anthony, so right now, I'm actually trying to reproduce following the the shtetl to continue. You have on the tickets, so once I'm able once I finish following the steps, I'll update the issue.

D

The experiment that I ran kind of simulate the work that you would experience.

E

Yeah so for us.

D

E

Saw that you know the details, there were no restores involved. So for us we were doing backups as well as restores and I just on Friday I was able to say that it seems like the memory jumps up when.

F

E

So I'll give it a try and try to break it down between I suspect, it's mostly on the wrist. Oh, but let me let me do some more digging around today and then I'll update the issue.

E

A

Awesome Thank You, ashy, Asian, Thank, You, Anthony.

A

All right, I think we're ready to dive into the discussion topics. Kaliesha. You want to start with yours, yeah.

C

I wanted to bring this to everybody's attention. Myselfin. No one have made comments on that issue, so the question is: should we instead of keeping the helm char, the volatile help chart in sync with the who is just sorry?

C

Let me catch up here so that way or another, do it the velocities in in in the way that they, if they are using at home, they can just for the CR deep, for they will just run and get that installed on the cluster instead of having that come from from the chart which will require continuous sinking of.

C

Anybody has a new. The idea, of course, is is tempter is appealing, because that synchronization requires there is nothing to trigger it. So it requires us to be attentive to.

B

Yeah I would I would say you know, I would I would want to look around and, as a few people said, that some other projects take this approach, so I'm not personally familiar with them, but I definitely like to see you know what those other projects aren't exactly how they approach it. I, just you know with all this stuff, I'd prefer to stay.

B

You know more or less in line with how the rest of the community is doing it, recognizing that there's not just one way, but you know if this is approach and approach, other projects are taking and it seems somewhat mainstream, then I'm, not I, don't necessarily have any opposition to it. I just wouldn't want to turn us into a special snowflake.

C

Points or that well then, if that's, if that's not the case, I don't see any dump downside to doing this either. So if anybody on the call later knows of anything, you can jump on example, there'll be helpful. I also take upon myself to to find some examples and see you get an idea. If this is well known and well used and then I guess, we have to come up with a new command, because that valera installed is going away command specially it's just for that.

C

A

Hey, do you want to go through your yeah, your other discussion topics here and then I can talk about mine, okay,.

C

Ashish me I mean to the CLI installed redesign fear and that one a gracious plan so for my grading to the new set of commands, so the the Larry Starr is going to be worried and we're going to have the little in Natesville their config yeah. That's it. The substitutable are installed. So the question here is: do we name the new commands?

C

So we proceed the names of the new commands with the word new and then remove that once the deprecated and that's going to happen with the next major first, just 2.0 or not so is the not just a guess for personal preference. Add the word new, because then we have to remove it and just still read the commands, no, like normally, of course, the deprecation message will be in in the commence to be in this document, which will be deprecated in any flag. That will be deprecated, and my reasoning here is so we don't.

C

There is no churn, there is no add remove later and also the commands are very suiting, there's nothing that will that right now in the list of commands that looks like the lower it needs and the little big. So they depends for confusion there. That was I mean those commands I knew in they don't. If there was a possibility of confusion with existing commands, then that could be appealing to me but I. Don't it's not how but I'm open to hearing if people think otherwise.

C

D

Yeah I made the I'm in the comment. I was thinking about how like saying g-cloud puts out it commands. There are some hidden commands and declawed beta blah command. So I was like trying to see if we can. That seemed like a good way to put kind of introduce new commands. So that's that was the thinking behind it. Did you think those call for making that comment.

B

Yeah, that was that was exactly what came to mind for me.

B

Was the G cloud beta sub command, so you know I think if we're gonna push the new commands out and call them beta, then I certainly think like nesting them under a beta command could make sense and could signal that hey these things aren't GA yet, but if, if you know they're, if they're GA, if they're ready to go and we push them out, I tend to agree with Carly Co that adding you know, nesting them under the new command just means more change down the road which is more work for both us and users, so I had given, given the arguments that Carly's he laid out.

B

I'd, probably you know, prefer not to mess them. In that case, yeah.

D

That makes sense to me as well I. Think, though, if you are going to have another install, come on or a lot but somewhat come as I said, I agree, yeah.

C

I know because we have to make sure that they are GA when we've released them so we'll take it well. Brother I rather focus on that making. Surely enough have to use any paid or new keyword.

C

The next one is also suggested by Ashish that we use first of all, my initial for anyone who did read the display are I in size, command and I use flags, but they should have been I use flags. The should have been commands, so my bad there in might have caused confusion, and so there's one one of these formats is UPS.

C

The way to go in my the people brought this up because it was then that I realized I was using flags and anyway do we format the commands in a way that it's very resource, a parade of resource operation and then the flag or.

C

Operation resource and then the flag now to jump ahead, the little Valero does have a presence, so happy swept ping the operation and the resource.

C

So now is going to continue to be true for this, so we don't have to discuss Tekken's we're going to do that now, the implicit that when we do Valero help what is appear on the terminal is going to be whatever cup is the first item so that item be resource or should the item be operation, Terra, Barry and I I'm, making a case here that when we do a little help, what we see there are world at the world's operation, suburb operator, action and the same thing with Cube control and in another case that I think the list of verbs will always be shorter than the list of resources, so that help listing is going to be shorter.

C

If we just go with the verbs first but I say.

C

Input is always welcome.

B

Khaleesi you're cutting out a little bit there. So I may have missed a couple of parts, but are you suggesting not supporting both forms going forward.

C

No yeah, sorry about my internet I'm gonna move next month, so I stopped worrying about it because it's not gonna get best me it turned out here is not the best um I mean in my area. You know no, this weapon thing is going to, but when we type a little hell whatever we get there, the in the listing is going to be either the verb or the resource. So that's. Why that's why it makes a difference. How we this up. Does it make sense so.

B

Right now they both show up. So if you do Valero help, you see, you know backup, for example, as a sub command for the Valero back up get, but you also see get as a sub command. If you're gonna do I.

C

Stand corrected, I was actually wondering is I both and I thought. No okay, yeah.

B

I just ran it to check it so.

C

B

Yeah, I think, is, if you look at how we're like the way that we're supporting both is by wiring both up through through the Cobra and so because of that yeah.

C

B

Think yeah both forms show up and now, okay.

C

I'm still going to go with the verb first cuz, that's under somebody thing, so otherwise I wanna just for documentation purposes and when one takes precedence over now that I forget how the cold it's laid out. I think will make it compatible with the way cube. Control is laid out or organized. That makes no difference.

C

Right I don't see if they're friends.

A

I'm, just thinking here like for get, if you want to do something, would get you do like a get remote and then add so you have the resource and then the operation I.

A

Think when, when doing a get, help or Valero help, I think it in my mind, it makes more sense to have the resource first, because I want to do something with plugins or I, want to do something with conflict and won't do something with backup and I want to do. Bolero backup something instead of bolero get.

F

Back up and I agree with Jonas 100% when you think about operations, addition comes first and then it's what you want to do so I know I also.

C

Agree, I also agree, but we we are trying to follow, cube control organ the way they organize it and if you type just keep control help you're going to say the basic commands are excuse me, the basic commands create expose, explain, get added, delete, deploy roll out scale outer scale and in this bunch of resource links, and then the troubleshooting is this store logs but then is attached. As so guess. It's all over the place, advanced supply, Patriot Place, wait, comfort, customized, it's mostly verbs and.

A

Gets a little confusing in my mind, with the burbs.

C

No so so I personally am on board with you, but we do have this competing re, which is to be in line with what cube control does, but but I, don't think it makes any difference at all. Points in our case is because we do the swapping so.

D

Yeah, so when I, let me give you my thinking when I made that comment for for when I was looking at it. It felt like not all the resources expose all operations. There are some operations that are like very specific to say, plugins, so we say plugins configure, so that is. That is what makes me think that we can have resources as the first sub command, and then we can have like the operations that did that are specific to that resource as what follows. So that was my thinking when I made that comment.

F

And I've seen that also work with across other see allies like power shortages, immensely popular in IT and and automation, is that it makes a lot of sense to have the research first, because if you want to see what applies to resource both operations, you can do what verbs it makes sense to basically run that so the latter then plugins and then type hop and then see what can you do with plugins the white cube CTL is I, see there they needed to be raised, kaliesha to be consistent, but I think if CTL is not architected right, I think it's confusing, even though works and a lot of people are using it.

F

It's not. The easiest to use I, take some chemistry so having the model where the resource comes first, and then you decide what operation you can do. It's it's more more easy to consume easier to understand my opinion. We don't need to overthink this so can make a decision in one now.

C

I would I would suggest that would make a call at some points, but in this case it's a milk point because we do the swap it's they are going to be list for better for worse.

F

C

Was I was wrong and initially I thought only one of the other. Only the first one will be listed. I.

B

C

My goods are, we.

B

We can also customize the help output, so if we want to actually you know, group things like haven't have the first section that shows the resources and then, as an alternate section, that shows the verbs. We could do that right now, we're just using the default help output, which just dumps everything alphabetically, but we could do that just for historical context. We, you know we initially only supported velaro resource verb and we got a lot of requests from users to support the other way around as well.

B

So I think you know there are strong opinions on both sides. I'm, not sure that one is more right than the other I mean.

F

Today we have its Valero resource and they're bright. It's.

B

Both we support us. We.

F

Support Valero.

B

Get backups or Valero backup get either ones valid, yeah.

C

That's what I meant when I a this whole time, I've been saying with SWAPO both are supported and both are listed in the help. Right now we can change. That's the precedence. I mean.

F

If we suppose, then we cannot make a change right now, even though we're we actually see a lion causing new complex some things, I don't want us to break a lot of things or solstar have built automation and backup policies on terminal error, so I, no subfolder both my mistake, I didn't understand that earlier collision. So in that case, unfortunately, we need to continue that design. Is there a huge cost to us to support continued support in both junior Soleri architecture?.

F

C

The library we use makes it very simple, in my opinion. Well, this might have different meanings. We can rethink the way we displayed the help. That's in maybe we'll list resource first but Steve just said. We have a different section for an alternate way to do the same thing, but right now everything is out. The boss get this fight. If that makes sense, got.

F

It yeah, let's keep doing the same thing, makes a mess of a resource. If we started down that path, we should continue one day by the way when we actually shape the letter to the door. When we can actually make our course breaking API changes, then you can do it right. Jonas see, you smiling sounds great.

C

F

Need to standardize eventually having both is, we should keep one. It will reduce our test abilities. Our documentation will reduce customer confusion, but well I.

C

Think I think I, don't know, be pretty hear this, but there's also a strong case to be made to stay within the culture of your tech stack it. Everybody who uses cabernet to use cook control and that's how come come, show use for first for most commands yeah.

F

And that choice might be that we actually follow what the cube CTO does so that so that's that that's fine! We don't need to make that decision today. So let's keep both right now and keep going copy.

C

F

Need to run in one minute to go and compete.

A

All right yeah did you get the answer. You were looking for: Curly's yeah cool all right next up, I have something that I wanted to chat with you all about I'm gonna share my screen here, so you can all see you as well. The so we've had a github issue running for quite some time, asking velaro users how they use Valero.

A

We also did a survey last year and we're publishing a blog post about that very soon, and one thing to make this easier for us and for our community here on the call to kind of boast about how you are using Valero I was suggesting creating and adopters MD file, similar to linker, D and harbor. So it's a simple adopters MD file that just lives in the root of the the project folder, where people can PR themselves in with their name linked to their organization for harbor.

A

We went a little further, so for those who wanted we can they can also add in how they are using harbor and I. Think this would be great. We've seen a bunch of really cool use cases John. You talked about one. Last week we've had maya data we saw yesterday. Bonzai cloud had a really cool one, so I want to make sure that we capture all of all of those, and then we can use the data here and actually feed into a adopters page on the site too small.

A

But I don't want to ask our community to edit the site, but rather just PR themselves in here in a simple MD file. What do you think.

B

Sounds good to me.

C

B

C

To make to $1 is if nobody really looks like this, we we only have three users will be sort of sang. We know a lot of. There is a lot more than that. You know what I mean yes say. Yes,.

A

C

Not have it listing only three companies well.

A

C

A

Here is to to populate it right: I'm, not gonna, populate all of it, but rather reach out to our awesome community members here on the call and on github and on slack to kind of feed that and make sure that everyone gets their their logo in there. So they can get boast about their usage. Essentially.

A

All right, I'll get this set up as a PR leader today, and we can roll with that sounds good cool I got a bunch of minor changes that I'm gonna do to the site as well. So I'll put those into.

A

Right, that's it I also just wanted to shout out Dylan Murray. He couldn't make the the community meeting today, but he will have a design doc PR up by end of day tomorrow, for this specific issue, talking about supporting custom, CA bundles, that I know, we talked about on the calls here a few times so you're gonna have a design doc up for that which is gonna, be really awesome. So maybe we can talk about that in the community, meaning next week.

G

Hi this is Ranger and I'm. The DMC engineer and work with Anthony and I just want to talk something about you know just manager about adopters. Actually, I really agree with that, because I just read about the harbor but officer, maybe because I come from China III saw that there are a lot of success stories of the company this their Chinese companies, I. Think yeah, I. Think if you ever call my neighbor from several markets for about the quality I think you Valero can I mean show something about these titles.

G

The tech stories I think this is a very quick thing over. They were I mean yeah, I, really I, don't wait, wait, wait! Okay, awesome, awesome.

A

Thank you yeah and just add add in the success stories that you have in there yeah definitely awesome. Thank you so much.

A

All right any other discussion, topics or comments. Anyone want to make.

A

Right then, we're diving into your computer, shoutouts Steve.

B

Yeah we've gotten a bunch of pr's in the last week or so merged from from community contributors, so I'll go through these, so first of all, Genting has made a few different PRS for us, but in this past week we got a PR merge.

B

Yeah support pruning unknown fields on our CR DS, so this is basically setting the preserve unknown fields to false, which has the effect of removing fields that are not part of the schema.

B

Part of the reason for this is that it makes upgrades to the v1 custom resource definitions, API group, smoother, because the true value is not actually supported in v1, but so this is a good change and keeps us up to date with what's going on in this er any world, so appreciate that change, Genting and then see Panaro another one of our Carlos another one of our ongoing contributors.

B

They've got a couple of PRS in this week, so the first one so on the Valero restore create command, which can also be Valero, create restore for everyone's knowledge.

B

If you use the the from schedule flag prior to this PR that the schedule that you specify there had to exist in API server- and this meant that if you were trying to do a restore from a schedule in another cluster or if you're, trying to do a restore from a schedule that no longer existed but that that still had backups in your cluster, the validation would basically prevent that from happening.

B

So Carlos changed this code to, instead of actually checking for the existence of the scheduler API object, it would use the label selector on backups to see if there are any backups with the schedule name label which essentially allowed us to support both of those use cases that I just mentioned, not working previously so another another nice little feature to have so appreciate the change Carlos and then the other one is from Carlos was it's kind of a tech dead item.

B

But there are some new topology labels that go on persistent volumes, basically for labeling the region and zone that the persistent volumes are at and the Valero code was previously using the beta versions of these. So Carlos updated to use the GA versions of the labels by default and only if they're not found to fall back to looking for the beta versions.

B

So this should keep us up to date with the evolution of these labels, but also keep those backwards compatible as well so another nice one to have thanks a lot Carlos, and then we had a bug fix from cone. I'm sorry I'm not sure that this contributors name off the top of my head, Brad cone, hopefully I'm pronouncing that right. Apologies, if not, who ran into a bug with rustic and I, can't remember if it was backups or restores it looks like backups and we had.

B

We basically had a bug that wasn't handling an empty output from the rustic binary correctly, so he logged the issue and then I was able to actually get a fix in and, as she shalt make sure that it was. It was good to go so he's been able to test this and it solved his his issues. So I really appreciate the bug, fixes it's great when you know, when folks can sort of help fix their own issues. That they're encountering so really appreciate it.

B

And the last one that just got merged yesterday, Shane utt added support for armed and armed 64 container builds so we've had the infrastructure. The plumbing in place for doing multi arch container builds for a little while now, and so we know that we've had folks who are interested in arm-in-arm 64 builds, so we finally got that support merged, and so, if you look at the master tag of the Valero container image, you'll see that it supports.

B

You know: MD amd64, arm arm 64 as well as power, Power, Architecture, so I think we're in a pretty good stay here and appreciate the contribution and I. Don't know that we'll make it out into another 1.3 release.

A

So that could be another big portion of the 1.3, the blog post time as well, yeah for sure awesome, cool loans, the last shoutout anyone else have anything to add to the call today.

A

Alright, then that's it for today. Thank you all for for joining in and yeah. If you have any questions, you know where to find us on slack and with that have a fantastic rest of the week. Everyone bye.