Everything EOS EOS Node Operator Round Table, 29 Mar 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Node Operator Roundtable - March 29th, 2023

Description

This week, the EOS Node Operator Roundtable updates the community on Leap software development and has an in-depth discussion on 4.0.0-rc2 release notes feedback and solutions.

00:00:00 Start
00:00:03 Software development update
00:02:11 Leap 4.0.0-rc2 feedback
00:31:49 Leap 4.0.0-rc2 release notes feedback
00:45:01 Leap 5.0.0 features and options
01:00:26 End

The EOS Network Foundation coordinates financial and non-financial support to encourage the growth and development of the EOS Network. We’re harnessing the power of decentralization to chart a coordinated future for the EOS Network as a force for positive global change.

// EOS Network Foundation //
https://www.eosnetwork.com
https://twitter.com/EosNFoundation
https://www.linkedin.com/company/eos-network-foundation
https://t.me/EOSNetworkFoundation
https://discord.gg/eos-network
info@eosnetwork.com
#EOS #Web3

A

Welcome everybody: it's March 29th back for another note, operated Roundtable back after a little week, Break um since then, since the last call uh EOS or beep 4.0 rc1 is now out. So we'll uh get a little update on that. How about we start there? Actually, uh let's pass it over to Brian or Steven want to take this.

B

Yeah I can take this so um as I've been communicated through a few channels. uh Rc2 is out. uh We had an RC one that ended up being internally uh released only just because we've found an issue with it.

B

um We do anticipate having an rc3, I, would guess, uh probably within the next week to week and a half just depending upon uh some discovery that we're doing uh we're feeling a couple of issues.

B

um I think some stuff we made discuss today, but one thing that we know that we're tackling is some of the issues related to golf blogging.

B

We know we're at least going to support uh ipv4 and then we're going to do a little bit of toy box Discovery around IPv6 and then there's a little bit of feature, consolidation stuff that we're doing with a couple other contributors, but the other uh announcement that we have is just for next week, we'd like to do a little bit more of a demo format, uh just bringing some of the team members in to talk about some of their features, and you know just answer any questions.

B

People may have uh it's around that time that we anticipate that rc3 going out so I think it'd just be a good time to just kind of get on the same page about any additional stuff. People are looking for and uh move forward from. There I also anticipate, uh probably in the next week or so they're being uh CDT release uh more to come on that uh Bucky is kind of out of the office for the most part this week, so we're mostly tagging up early next week, but I think just about all that feature.

B

Development is now complete.

B

um That's it for uh updates from from my end, though,.

A

All right, thank you, Steven and, if I understand Brian, you were, you were thinking of doing a bit of an open Forum on feedback from folks who've had some experience so far. Is that right.

C

Yeah, so excuse me, uh for the last week, I kind of took my eye off of the um some of the antelope telegram channels and and whatnot so um I know that uh Mr Matthew Darwin is, is um you know actively working with the Ford Auto release candidate two um that he's been actively uh submitting some issues that he's found.

C

um I guess first I just wanted to take a pulse check, yeah first yeah. Thank you Matthew um yeah, but you know I wanted to take a pulse check and see you know if anybody else yet has um has started playing with it at the stage.

C

I know that sometimes Mafi plays the role of uh Canary for us, so.

D

I haven't had a chance to play with it over on our side, we'll, hopefully work on getting a build kind of maybe roll out to it. You win been able to preoccupied myself.

C

Okay, all right, so it.

E

Sounds like we're just uh I installed the binary and just kind of like read the release, notes and everything. Yeah actually I just had a small question about that, specifically about uh lox log stride.

E

So when you split that block log and it's no longer one big file and then it kind of like forms, this standard block log so to speak, can this be on different disks? Like, for example, can you specify multiple directories for that box, separate files? So, for example, the old blocks you can keep on say a traditional disk? That's not SAS speed, because you know you you'll get much less requests for that versus. Like uh you know, the latest uh block log.

C

Yeah, so um there is a differentiation between like basically how much block log do you want to keep, um uh but Kevin do you want to do you want to address the specifics of how you can sort it split.

F

Sure yeah so there's uh the options I'm just looking right now, so you can specify the stride and then the uh how many the max retain files that you want, but then their ability to specify a retain directory, which is location of of the directory of the of the stripes that you're maintaining and then the archive which are the ones that are just you know out there um that are not um accessible, so yeah.

F

Yes, you, you can specify a blockster which is the active uh one if you will or the the most recent one is more proper way of saying that and then a a lociation for the retaining directory, which can be separate, which could be on a floor disc for sure.

E

Got it so that, for example, if you have what doesn't Max retain lock automatically um move, you know like anything outside it like Auto, deletes it or moves it to the archive and then no longer it's no longer accessible. I meant more like. If you have like you want all the blocks accessible, but you want to know blocks behind a certain number uh on, like a slower disk and just keep the latest or, like you know, the one with the most up-to-date blocks on you know your fast nbme, for example,.

F

uh Yes, so so the way that works is you have a specification for your blocks directory, which is going to be the the very most recent ones, and then you have a stride. So let's say: let's try to ten thousand. Well then uh the most recent ten thousand is going to be in the blocks directory and then after that, once ordered, then that are going to be. uh Are they going to be moved over into the uh which is can be somewhere different, they're still accessible um uh at that location?.

E

So I guess, as long as we don't specify a Max retained block, then we never have to deal with. You know the archive, which is not really required unless you want to keep them on off storage that if you want to sync on other nodes, for example, you may bring them back later, but for normal use case like uh we'll, be able to use a blocks directory and the blocks reset directly to separate uh or expand the blocks log. Basically, yes, exactly thank you.

C

Any other questions based on the oh I see something hang on in the in the chat here. um He plans to improve support for M1 Max um I. Don't really have an answer for that right now, thanks for asking, if.

D

Dune is always the answer. We talked about Mac support, yeah.

B

Yeah, so so on that uh there are efforts going forward to pick in better support, Max uh right now, I think we're also working to schedule a meeting next Thursday to discuss uh the experience with Mac users I'm happy to add anyone to that call if they would like.

F

Now I had that you know we are trying as best we can. uh Although we're not running active tests on Macs to not break that experience, we have at least one developer that that works on that platform. So we're trying to trying to you know keep that going as best we can until we figure out exactly what we're going to do. There.

D

C

All right, any other questions based on uh sort of first to look at the release, notes or anything.

C

E

Right um this is uh maybe a small question: I, don't think of them called wrote up in the past, but uh it the whole scheduling or snapshots and everything you do with snapshot. Gonna be always set in the producer Plugin or are there still plans to kind of separate that from the producer plugin itself.

F

It's definitely in the producer plug-in, um certainly I I think there's I've heard at least desire to move that somewhere else, so that you don't have to enable the producer API for things like pause and unpause um and other items there is that what you're asking for is the fact that it could be enabled separately than the the producer API plugin.

E

Exactly because generally you're not producing on a machine you're going to create snapshots from so just didn't, make like long-term sense. I guess.

C

Yeah yeah, so we've heard that feedback um both specifically about snapshots and, more generally, about um certain plugins being kind of overloaded in terms of when, when you would want to enable them or not right um but based on the the functionality, that's included in them, and producers is the big one.

C

um Excuse me, but uh so that's also you know it would be a perfect segue if I was going there right now, but not uh to to the sort of node types discussion, because that is part of the purpose of of why I you know we're coming through. That is um to potentially inform some.

C

um You know splitting of of notios, either literally or or by you know, options. You know that sort of thing so um yeah, but thank you. You know by the way, um if you ever have feedback, and you think the feedback has been shared before it's still always welcome to share it repeatedly.

C

um In fact, sometimes that's uh what helps us uh notice that something is indeed a high priority.

C

All right any other questions based on uh release, notes.

E

Another question just about the the peering with BPS is that I'm, assuming people are not going to give their necessarily their public IP or, like uh you know, like a URL, that's directly accessible to their VP and I'm. Assuming this is meant to be used where you have like a public peer that is connected within your land or within you like, uh like in close proximity to your BP, or is this actually expecting people to put their direct IP addresses or direct access to their BP nodes?.

C

Right so I think that that is left to the um to the block producer to just they're block producers Community to sort of decide how best to do that.

C

um But there's certainly no reason why it couldn't work with sort of a you know. A dedicated peer for you know, block producer um uh relaying if you will um uh yeah, but the the intention. Of course, just going back to the intention of the features to allow block producers to.

A

Sort of I guess a better hand off I, guess yeah.

C

I have better handoff to avoid um situations where, where they, you know, you get a late start on your first block uh because of latency issues and.

G

C

Sort of thing um by minimizing that latency, so that by ensuring that um you know you're allowing connection from um and and connecting to you sort of your your proximal.

E

And does it just allow connection like it just adds it to the curious so to speak and tries to connect to it or there's some sort of like prioritization, for that, like specific PR, to try to send a block to them. First, for example, so.

C

It's actually a separate um uh I, guess sort of it doesn't count against your like uh your max peers and it um and it's a sort of a separate config right. So it's you know you list so for BP nodes that you that you know you are aware of their sort of connection information and they're. Obviously their account uh information. You've got that in your I forget what the config is called right now, but you've got that in the config. For this feature um and.

D

um How do you configure that? Where does it get the next VP in the schedules Pier that should be? This is BP peer. Where is that advertised.

C

Aaron, oh you're, saying like where: how would you get the connection information? Is that what you're.

D

Saying, like I I know mine and my config, but I don't know Matthews when he's next to me, I, don't know um yeah, you know. Whoever else is this next in line? How is it broadcasted? Well.

C

There isn't there isn't really a discoverability feature with this at this time, um so that would be something that you would you'd either have to do some sleuthing or some sort of out-of-band. Okay,.

D

I mean I, understand I.

C

D

Automatic but manually how? How is it defined.

G

In the new peer-to-peer config line, there's a it starts with BP Name colon IP address colon Port, colon BLK or transactionality. You specify the the name of the account of the BP.

C

Right so I could I could imagine. For example, uh you know somebody having a um you know a a version of that file. That's um you know uh that they share with other BPS right.

D

That one I didn't see that in the the notes there that say, I mean I could see. Auto BPP or I just didn't realize that there was also a new mapping option. Yeah explicitly was BP name to whatever the other BP told.

C

You yeah, and so that config not only defines the connection information but sort of implicitly also gives them precedence. um It's a set. It's not it's a separate sort of prioritization if you will, but it basically says since I've included them here, I'm going to like allow connection automatically, assuming that the sort of sex schedule, proximity logic, is met right, um so, in other words they're uh they're within uh what is it two they're within two? uh You know block producers from me in terms of the schedule. The current schedule.

E

The event is BPP or flag in in the net API. That one is just derived from your config.ini. What you specify uh yourself on your machine and not doing any like checking other people's config or that's Central P2P, just based on your own config. So an ideal case say you have the top 30 BPS we write an entry is bppier for every one of the 30 BPS and then, ideally, it's some type of like accessible P2P node. That's within you know, proximity very close proximity to their actual producer.

E

Actually, that's how I could see used widely and be able to share those lists without being a security concern.

F

Yeah and that that's a good point- and maybe we should address that yeah we we do not in any way broadcast that information or that configuration information. So you know it if it contains.

F

um You know some sort of connection information outside of what you normally communicated when you connect to a to a node, obviously, uh but the actual configuration options or that can think those configurations are not are not broadcast or shared, and so yeah. It strictly meant to replace the manual process of where you change, like okay, I know that this BP is his IP address, I'm going to now add it to my connection so that uh you know I have better peering.

F

So it's strictly for for that that configuration option and it it's it interacts with, with the current lift in that. If you have a connection in your normal P2P peer list, um it will maintain that connection, even at the uh attempt to maintain the connection as the schedule changes. So if you have one in your P2P peer lift it, it won't like disconnect from that, because it's no longer uh in your it can you know directly uh in proximity to human schedule.

F

It's only the ones that aren't in your normal PDP configure list that that's in the BP configured list that it will automatically connect to or disconnect from as the schedule.

D

Changed and that's inside the standard, peer-to-peer plug-in b2p is there any level of enforcement if that information is found by a non-bp or it's still going to accept like a normal peer-to-peer connection, I mean it's kind of a back door around the peer-to-peer limitations, but well.

C

They'd have to be defined in that in that new config, so that the current rules still apply on their own sort of separately, um for them to have for them to be able to use this as a back door, so to speak, you would have had to configure them in the BP list um in your config. So sort of um you know, implicitly by putting some of these connection settings in there. You trust them enough to let them connect to you in the event that they are.

C

uh You know close to you in the schedule right, so it basically it uh implicitly, both you and the the connector I guess have to be. uh You know primary BPS or you know, uh in order, in order for this configuration to be meaningful, so.

E

I have the thing: that's concern is more if you've shared that, like you know, once you enable that, will you oh I see you know that anybody having that list can just deboss your like do whatever they want, if you're not putting another P2P in the middle and that P2P in the middle could also have its own limitations in which only accepts connections from the other 30 BPS defined in that list, and that way you can make that setup, secure and.

D

I now have an exposed, hopefully hidden just because nobody knows else knows about it, but twofold: that's actually the mesh that we try to use to make sure the BPS are staying. You know so I mean that's a main line into the the back mesh that the BPS are trying to keep. So that's why I mean yeah, so this I don't know. If they've been one simple thing to do that, then a mechanic that's already there is that he'd be hearing the key uh King limitations.

D

Maybe it would only accepts a BP peer that has the BP key, because you add that I mean then you would know it and be able to associate and maybe use that as anyway. That was the question about I'm going to expose it and then I'm going to have to publish it to my 30 friends out there. But if one of those friends is no longer my friend, it's more accidentally asked his peer list in public.

D

Once it's out there it's out there, so there's no enforcement to or just somebody sniffs it I mean they can just try it now, whereas you can just you know, stride stride, your IP.

C

E

It's something talk to anybody. I doubt anybody's got to put their BP nodes like directly publicly, but instead, as I said, they use like another P2P node uh that could have its own. You know like uh whether you see PRT or using just firewall, to allow only the users or the BPS to connect to that P2P node and that would connect to your BP node, but you would publish your P2P node as your you know: config shareable config to the other BP correct.

D

So I've already kind of got this other layer too, that you gotta lay on top of it's a further limited, and, secondly, that's the more private peer-to-peer mess you're trying to protect. You know, anyway, that I. That was my question absolutely.

C

Exactly this isn't meant to like it doesn't replace the existing functionality. It's just intended to facilitate, like Auto configuration, uh something that we haven't tested that you know a enterprising person might want to play around with is is what happens if you set your max peers to zero, but have this configured, um because um you know in theory that would allow you to only allow connections from proximal uh BPS, so um something to experiment with.

D

Is that only allowing from BPS it's only allowing from a person who knows the information right, there's, no verification that the peer-to-peer that shoving transactions in is anybody other than somebody who has the right address correct.

E

I think anybody with an address from the sounds of it.

D

E

Now because you don't ask for the information of the config through the P2P right, so they don't know that you set this BP, it's a BPP or not. That's not transferred over P2P, so I think Collective saying is as soon as that information is available. The PHP endpoint anybody can connect.

D

It's just like another private public node from the exposure side of it. Now it's smarter that it is intelligent to try to figure out I need to talk to this person next in line, but the problem is in order for that to work. The door has to be left open on the other side or we're right back to where we started. It's just a private peer, that's shared between the pts and hopefully nobody steals the beans. Yeah.

F

Yeah in that, in that way, it is the same. It's it's an automated process for what you manually do today and, with the same caveat of your manual process that you have now there is uh just felt, was clear, I'm, not sure. If, if I heard a misunderstanding, there there's no requirement that this actually be on your VP peer.

F

So like you could have a separate P2P node that that's your relay note for the connection, your Canary nodes, let's say probably, where you'd have this configuration not on your BP node, but actually on a canary, though, where you really relay note whatever you what, however, you describe those and it's simply, it really is just simply nothing more than hey when I see this producer in the schedule connect to this pier right now. Presumably that is the peer that is associated with that producer.

F

It doesn't have to be right, I mean it could just be like well, when I see this producer connect to this random thing, you know it. That's all it is it's just when I see this pretty measure in the schedule.

D

It helps you, as an operator, try to more effectively manage your connection outbound, but, like I say in order for it to work. What now is going to have to happen is the same thing right now. I have to expose a separate producer, peer-to-peer match, be it the producer. Note itself were its own, dedicated node and try to secure it and try to make sure nobody else, because it's now a complete backdoor in the back system. Now it doesn't have any max limits, it doesn't have anything you figure it out.

D

You can completely smash it, which is the blood producer mesh. Does that make sense I mean in order to work you'd have to open it up, makes clients to.

E

D

Anybody it would not accept a normal connection via pure public whatever, but if somebody connects to it as a it does no enforcement of whoever's connecting to it, maybe it stops at 30.. Does that make sense I mean it completely bypasses the back door or it it lets along the back door. So.

C

I think to Kevin's Clinic, learn it. The important thing here is that it's it's actually the other way around. So it's I I think the way that I introduced this was sort of flawed. uh It's the other way around. It's it's! You are proactively going to connect.

D

um Black I get it it's the that's. What it's for and I like that I mean it, helps you as the operator, but my question is in order for it to work.

D

Every BP now has to publicly expose a or you know, optional, this new peer-to-peer that we offer. If anybody finds it out, they can just smash it.

E

But I think you can just use it like, for example, you know you can put say you know: C Niche, you put Nations it's as simple as you can put missions, even like public P2P. That is going to be much closer in theory to their block producing node they're. Not so that way.

E

Even if you put that as BP you can, you can guarantee a connection to that and that's more likely than not part of their internal land, and you do increase like your chance of propagating the block faster to that uh that kind of and then so to speak. And then you have the same benefits without any security concern.

D

Right, I I love the idea of it going out, but I, maybe I'm just misunderstanding: how it's not exposing.

D

If anybody else has that information, there's no Max peer limit, it's also the mesh of BPS that you're trying not to have people smash into and is accepting of transactions, and things like.

E

That you're, assuming it's a mess of VPS but like Kevin, said it's just that's how it's naming, but it's and in reality maybe it should be named like or whatever the name is. The idea is that when you have uh like the BP peer in reality, it doesn't mean it's actually the BP node, but a node within you know close proximity to that I.

D

Get the technical distinction that it doesn't have to be the block producer, but in maybe in referring to the terms of the type of tracking, there's public trash and then there's the block producers trying to talk and only maybe even blocks only relay types of things. I get the distinction. It doesn't have to be the block producer, but it's still the more private hidden network that I'm relying on to be not full of trash and smash to make sure I can get up, get it over to Matthew.

D

But if Matthew's peer-to-peer is getting smashed to death, because somebody found out that information number one I can't connect to it. Number two she's now relaying that right back into me at an almost public peer-to-peer level, because there's no validation that.

C

What yeah I would think of this two things, one one I think uh this probably deserves a more in-depth um session to sort of almost brainstorming between node operators of like how how this feature should be used. You know like um so maybe we put that in a parking lot, um but two I would just higher level I would think of this as a part of a solution right, Like, A, A, A Primitive.

C

If you will that you can use it's it's more about sort of, um what's the word uh being a good citizen, if you will to help other BPS receive blocks on time and um yeah, so we'll we'll go through that um in more detail on another in another um session. I think you know pairing this with. You know other um uh other approaches. You know like networking security stuff.

C

um You know you have a complete um uh solution for sort of minimizing uh latency. uh You know safely so, but this was just to go back to like the the reasoning. This was in response to um you know, sort of learning that block producers were often doing this manually anyway, like frequently reconfiguring to make sure that um other block producers were in their.

C

um You know peer configuration uh for for certain nodes, and so this is an intention to to automate that process to some extent um so that you know it's sort of built into the software, but yeah uh I'm writing that down as a future topic, um because I think we could go into a bit more detail of like exactly how it works, uh preferably with a diagram and- um and you know, then discuss uh some of the uh real world practicalities of what what other things you would want to have in place to make full use of a feature.

C

C

Okay, um so speaking of release, notes, um sorry I wanna make sure I wrote that note. My my mouth got ahead of my brain.

C

Okay, um So speaking of uh release, notes um I did want to sort of seek feedback on on the release, notes themselves. So I know one thing that came up um you know: Matthew gave some feedback on was around um some of the more sort of technical documentation, things that were not included in the the release, knits themselves and so I kind of wanted to have a a brief discussion on um almost sort of information architecture and like to what extent release Nets should be.

C

You know, verbose and complete in and of themselves versus to what extent they should link out to useful information right and that's more technical documentation, oriented and whether that technical documentation.

C

um You know what the nature of that technical documentation should be. It's a sort of best support, uh node operators, you know hitting the ground running when new versions are released, um so I that's kind of a vague question, um but like it, you know for those who have really read the release notes, um you know what what do you need to have included there and what do you need linked to that? Would that would actually? um You know allow you to successfully uh upgrade to the to a new version. Does.

E

That make sense, I think the new one uh like how it's set up now makes a lot of sense. For me. Obviously anything breaking changes needs to be like put bare any new API endpoints, any new features, kind of like how you're doing and- uh and then you know, for the subject really need a lot of more details or examples. We just linked it like you have with a printing snapshot scheduling. We have a link and somewhere else. You have a link when there is more examples or information about it.

E

I think this is quite well. Ideally, you also have examples somewhere where where's appropriate to have examples, but that will also be in links and not in the release. Notes. In my opinion,.

G

It would be great if it linked to the actual documentation portal, but the documentation portal is not up to date. Doesn't talk about 4.0 anywhere yep.

C

G

A this is a like: the documentation should come out at the same time as the release. Right, ideally,.

C

Right um yeah: this is something that we, you know from uh release process perspective haven't, haven't nailed yet is how to how to synchronize those things so um yeah I agree.

E

I think it's important to come out at the same time if it's a stable release, I think it's 30 RC not to. Instead, you have it out because you might have to repeat all the work or things that change.

G

Yeah but the the documentation should be already ready like if the documentation should be developed in parallel with the new features right, ideally as a Sprint, a done done feature means it's documented, so uh presumably yeah the release, notes might not be or the documentation might not be final, but if the the new features is describing. If, if the release has a new feature, then really you want to point the documentation and not write it all in the release, notes and then redo it all by adding it to the documentation.

G

So it would be less work to logically and lesser. Maybe not you know practically depending on how the team works, but.

C

Yeah I, don't know if you had documentation is like uh effectively a totally separate product right um now. You know uh from an external perspective, uh I I'm, in complete agreement with you that that makes sense. It is part of the product right.

C

um This is part of why I'm asking this question is to to figure that out. So one piece I heard is like at least when we go to stable release. uh We should link, but then you know we do have folks like uh Matthew, who frequently play the role of canary um and so I. Think, there's a um you know, there's a there's, also a consideration there for like what does a um Intrepid Explorer uh need to. You know to be well respective as well right if.

G

You want people to actively test your release candidates, then you need to make it easy for them to you know and not you know, Michael was confused how to configure the BP configuration peer-to-peer because it wasn't like I figured it out by doing nodioa, nodia print default config and then it's like. Oh that's how you can figure it right, but it wasn't obvious from yeah like.

D

The in the release, now the the scene is expected. Oh.

C

Yeah, it does seem. Oh sorry, real, quick I just wanted to make this point like you know, linking to documentation is one thing um and I did say that, like you know, documentation is effectively a separate product internally like process wise right. um Man pages are an interesting sort of caveat to that. Right, like you know, more um uh more built into the tools stuff, so I wonder if they're do.

G

People still use Mac ages well,.

C

I, don't I don't mean Man pages, literally so much as like the help that is, output by the tool itself. um I wonder if there's you know a preference there. You know from you know: web documentation versus sort of CLI documentation. You know.

E

I, thank you generally need both, but personally I I much prefer to rely on the local documentation like if you ever use the excuse or something it doesn't matter. What you read online you're, going to get need to get familiar with all the different options that are only available in the you know, dash dash help uh in the CLI.

E

So um for me it has to be part of the actual, like binary, that, like you know the equivalent of man or that just help, but uh for for Mass adoption yeah a lot of people try to read uh like you know, on a web page and- and you know, graphics and all that fun stuff.

G

You're, like I'm doing training so just to Echo that I'm doing training for fire hose telling node operators how to run a fire hose ecosystem, and the thing I include in the training is buyer- follows dash dash help so I'm training them to look at the help that comes out of the binary right. It's required. This is the documentation.

G

C

Wonder if there's a strategy where we improve our CLI, you know sort of embedded, local documentation and uh and sort of that's the you know, that's what's included for like early release candidates and then at stable.

C

You know the um the documentation gets included in a you know. Prettier format with you know, whatever diagrams are helpful, Etc in in the actual documentation resources like online and whatnot um with something like that sense.

G

Yeah potentially yeah I.

D

Mean outside of different aspects.

G

I mean I've always grumbled about the documentation for nodio, as it doesn't tell you how to use an option right right. It just says this option does X well yeah I kind of guessed that from the name of the option but like why, in which conditions do I need to change this? It's never clear right. So the only way I know is ask Kevin.

D

Well, on the main night you can go in, you know what I used to remember, going and looking up some of the details on eos.io, where it was more broken out. Yes, some of them are yeah, but other ones, identity, correct, I, tend to be more of a manual online. Can you know be able to find it versus that? That's help everything I hate CLI, as opposed to getting my mouse, but um I mean I, got kind of a different aspect on that, so what I'm? What we're running into and and I'm curious?

D

If we can do anything about it, is like right now, we've still got people that are going from the 2.0 stack to the 3.0 stack, which are getting burned by a lot of these new little gotchas that all along the way got added in there or ultimately changed as yo resource monitoring. Now triggers you know, that's burning everybody and um thank you. Michael yeah I know right I.

D

That I will eternally my last parting gift to still breaking, um but what's- and this was the problem when we went from 1.8 ultimately to 2.0- and it's now- you've got 4.0 on the horizon- is a way to kind of effectively both from the node operator side, as well as from an API side I'm not as fluent with it but send transaction to or whatever these new little feature. Yeah enables and RPC.

E

D

Almost have a a version of op, Matrix or features or whatever even classified I mean yes, I understand. You can read this note in this note, and then this note, and then this note but they're cumulative and it's brutal. You know I mean especially now that you guys are working so much magic. Y'all are spewing out stuff, you know constantly it. I I've already been through the first wave of it where people didn't know.

D

The resource monitor I'll, use that as an example and now you're about to have another way where it is, you kind of had a way of saying, look, I'm going 2.0.9 2.0.6. Some of these chains are still running and they don't even add something. But later whenever and here's all of as a node operator you need to now know the resource monitor, is whatever Prometheus.

F

D

Man, that's great, that's more of a feature type of thing, sorry to minimize it Matthew, but some sort of a way of a critical path, type of thing and even for the API operators and I, don't know subjective billing. The way that's kind of you know but to say you're going from here to here and assume the latest and greatest type of thing.

D

But if you had the feature Matrix, could you then categorize them as this is where it came in and it's been flagged, as you know, even retroactively, hey it's a great idea, but ironically, it's breaking a lot of out of the blue it'd, be nice to remind them and even have like a quick type of whatever anyway I'll stop, but that that was something that popped into my mind, especially as you've got. You've got different classes of maturity right. They just wait, wait until the best settles and make it hop.

D

Now we got two hops and an.

E

Ideal and an Ideal World you'd have like a documentation site or a migration site that you were nearly do a drop down of on this version. I want to buy a rad version. It breaks down everything, that's depreciated in between what are all the new features and what old requests mean to change like you know, add use old, send RPC, kiosk or something to that effect.

C

Obviously, we're talking more specifics, but it seems like the broader theme here is just um that the upgrade process could still be improved quite a bit like all the way from knowledge about. uh What's going to change to the actual process of upgrading itself right.

D

Correct and realizing you know, we live so much Matthew bleeding edge. You know, I mean we. We live in the channels. Oh yeah I forgot about that. By the way, don't forget, PSA your shit's gonna shut down at 90 by default, because this new feature kicked in, but most of us have forgotten about it. It if there were a way to you know at least kind of knowingly say from this version to this version learned a couple different ways: either information or critical path.

D

You know type of thing anyway that, as as we're just you were discussing the documentation, yes I mean you should have these notes that are detailed, like matching City.

C

So I appreciate the uh the feedback there. I've got some some action items taken.

F

C

That I'm going to Ridley switch gears because there was some stuff that we wanted to ask so for some feedback on so um Uh Kevin. Do you want to go into the um you know, timeline compression uh for lack of a better term uh discussion right.

F

Okay, so we're transitioning we're going to talk about a 5.0 feature, um so September October.

C

Thank you for that context, so we don't thoroughly confuse everyone right.

F

So this is, you know this is not uh don't be in 4.0, so, let's again 5.0, so um what we've done I I put a link there is that we have um in in a PR it's looking good. So it's about to be merged into what will become 5.0 is we've taken the uh the CPU effort and the last block offset and the block Offset you know, and the last CPU epic right.

F

There are four options: um the the CPU percentages, ones and then the offset ones which control the exact same thing, they're just different ways of specified. You can do it either as an absolute value or as a percentage, whichever one you prefer, and it always takes the conservative of those values.

F

What we've done in this PR that are linked here is we've basically done away with all those and left one option, which is a CDU effort percent Works um in this change, is you have your your 12 blocks of your round and uh what currently happens for those 12 blocks is according to those settings. Is that you always start your block production um uh 500 millisecond before the block is scheduled to be produced uh so that uh if, depending on your settings, you're going to release that block um a bit early?

F

So by default, it's 80, which is 100 milliseconds, so by default, you'll release each block 100 milliseconds early uh from, and that gives it enough time hopefully to make it to the next block producer.

F

And then you guys often take the last block of that round, and you make it much much tighter right where you make it in a lot of cases, it seems like almost zero size to make sure it has plenty of time to make, but we've effectively done away with all that we've compressed it down into one seti and we don't have any time between blocks.

F

So you'll start your first um block of your 12 block rounds, just like normal 500 milliseconds before it's scheduled production time, but then, according to CPU effort percent or if you've actually uh exhaust the consensus, Max CPU in a block, whichever happened first, um it will then relieve that block producer release. It find it all the things it does send it out.

F

On the network and immediately start the next block, which, um presumably, if the, if you have the default settings, um will be 100 milliseconds earlier than it would have started in the current code and then the next block is in 200 milliseconds earlier than in the current code in the next block, 300 milliseconds, and that adds up where your last block figure out the map. There is what one point two two seconds early um by default would be thin out.

F

Now, we've changed the default because that's a ton of time, uh so we've changed the default from 80 to 90 in the yard. Well, so that you'll need uh for the last block would actually be 600 milliseconds early by default. Now, of course, you can tweak that setting until we get back to 80, it would be 1.2 seconds to get to the to to the next block producer.

F

So we we've basically taken that timeline for the blocks and compressed it all into the early part of the window um uh to to produce these blocks out the door. You still get the same throughput.

F

um You should get the same type of latency uh when the uh when the blocks are are um are are getting uh filled up, um there's a little bit of difference in latency near the end of that that window, if you're, if you're, not producing full blocks because there's you know, there's a little bit uh larger Gap there than it would normally be, uh depending on your configuration settings, uh of course, so all that hopefully explains what we're moving to uh by default.

F

That would be what we're doing the question came up is: is there a a desire or a use case for maintaining the current um operation as an auction? um Knowing that that would you know, complicate uh a number of things, um I mean it happened to work both things uh at once, and you know all the testing to make sure it worked both ways. Everything worked both ways and all that kind of stuff.

F

uh So, but that's the question: is there any desire or any use case for the current uh mechanism, where you always sold the car mechanism again? Is you always start your your blocked production, a 4A given block at 500 milliseconds before it's set to release all it? That's always the case now, according to your clock, it always started 500 milliseconds before it's supposed to be released, every single block and and the new Behavior we're talking about here.

F

If that's not the case, you simply can press it all into the early part of the window and do we do it as fast as we can when we're overloaded.

C

Yeah, so a good explanation, let's, let's sort of like restate one more time with you know um more specifically what what feedback we're seeking um so just try to facilitate this. Is there like a 10 word question a high tooth: did you state that, as a haiku company, just.

F

So is there a desire to maintain the the current functionality of starting each block at an exact time, um as opposed to compressing blocks within the 12 block window?.

D

With the time stamps still be, every half a second I mean yes, it nothing other than the producer note itself would notice any difference. Old, Network, new version, validating blocks right I mean it's really just correct. A timing, type of thing, correct.

F

We already re uh produced and released Blocks Early so and that's why you see negative late right sometimes have people like what? How can you have a negative latency? Well, we we calculate that according to when the block should have been produced until we always produce Blocks Early. It's just with this change. The blocks there at the end of the window are really early if you or earlier than they are currently possible in in the current solid.

G

Yeah I think the only the only issue is the the monitoring systems will go like what, because it's really early, that's all otherwise, it's good, because now we have much more time to get that last block to the right place and the blocks can be Fuller. Yes, they can be bored and bad news are mystery. Solutions, because now we have Fuller blocks.

F

Again my VP used to it I mean you can regret correct the the thing down. If you like, uh slow down like the by setting the smaller steep.

G

Clean version I was kidding, yeah well, I mean I, mean it changed right. So a deal.

G

No I don't see I, don't see any for that.

C

Yeah, it sounds like the um no dissent that here that, like we, don't really need to maintain the current functionality we couldn't. We can move to the the purpose.

D

Does it affect the time offset on the front side? I think that's how we wish the the front side of things around. Does that there's a setting that allows you to I think I'll set the pool no.

G

That got rid of the first time, didn't.

D

G

Got rid of that already.

F

um Yeah left not I mean there's the the.

F

uh What is it the block offset, there's a block offset with offsets each individual uh Block in a round um and so yeah that that setting would go would go away and it would all just control by deceit, effort percent and I. Guess that's the other part of that that that we didn't bring up Ryan was you know he there's two ways that we can do this right. We can.

F

We can mark the options that are basically ignored after the pr goes in as deprecated um and but lead them there and just like when you do help that do deprecated and if you do the like. These settings are awful uh configurable via HP endpoint and and leave them there as well, or we can simply remove them if we remove them the the only um kind of the drawback of just removing well, a good thing is removing we're done with it. We can move on.

F

We don't have to worry about those anymore, but if we do remove it, that means when we do, when you guys do the bi.l update you'll, have to uh change those settings as part of that upgrade process. Ethical review.

D

Just what happens, people restart their node and then go oh I got this network version match, that's no longer a valid parameter, or now it's the B1 history I mean it it. It does happen, but that part of where that previous statement about hey those are the types of things that, if you just simply knew we can go from this version to this version.

D

You need to know that when you restart it or you should prepare, this is no longer you need to have in like in my guide, hey guys, but there's a way to do this.

E

But that's too easy to be handed with the CLI, instead of just saying that gustako EOS you're upgraded to fight, and you start now with EOS, you have the old settings, it tells you, these settings have been depreciated or you know incorrect, Plug-In or or unknown plugin, but also it detects that the old plugin anything to do with that uh you know offset. It has to also write. You should be using the new. You know CPU effort, and maybe you know like a small example of what the default is uh to use.

E

At least people can get back up and running quickly and then explore it further. If you want to customize it.

F

Yeah, so I think we actually did better and released out from this regard. um Let me drop the ball on some other revarts, but in this regard actually in the 4.0 release, node uh rc2 release notes. There is a list under upgrading from prior release and there is a list of uh options that um have been. uh You know that that been removed that that you have to remove from your config configuration now that's a manual process and not a like. You know, upgrade script that you run or anything like that.

F

But uh at least it is document.

D

Oh well, yeah buried way down there in a text. Wall.

C

Well, the whole thing: that's actually why I was asking the question about like how verbose things should be versus linking out right, like you know, I think I think there's something to be said on both sides of that right. The concise I could I could even even imagine, going the opposite way in the release, notes being practically just a bullet list, but linking to things you know.

C

um So you know I I think there's like a valid approach in both directions, um but yeah it is. It is a textual anyway,.

C

Yeah we're at the time before we go I just wanted to check. Is there a clear sort of Direction forward on that uh that conversation, Kevin um I, didn't I, didn't really grasp what it was other than it seemed like there wasn't any concern with removing the um uh removing the current functionality, um but I wanted I could follow up with you. Offline I just want to make sure that you, you feel like we got what we needed out of that Commerce.

F

I think so I think I. We we heard that. Obviously, if anyone um you know Waits up middle night, saying, oh, my goodness, no, we need to maintain the previous Behavior because of X. Then you know, certainly let us know.

E

Right I think it's always easier to remove things, especially in order like accrue a lot of this Legacy stuff, especially for newcomers like when, when, like you know, I was trying to help someone set up the fuse just doing the true start that that's helped. They were just like overwhelmed.

E

They wanted like a long time just to read through these, and even though, in that case they weren't Legacy ones, they had all actively used, but you don't want to make the large harder for new people uh by just adding the old stuff just to you know, because there's really no no valid reason to keep it. In my opinion,.

F

Yeah and that's I think that's exactly what Brian's trying to do with these whole like um documentation of various modes right I mean we can simplify this whole thing and just say: here's what you need to do for an API note: here's what you need for a BP node as opposed to here's, a million options, good.

C

Luck yeah either. That's either like clear documentation for each mode or even at some point, separate, separate uh applications or or options for nodeeos or whatever, but anyway, I do we should. We should stop here, we're four minutes over I'm handing it back over to Daniel. Thank.

A

You thank you, excuse me and quick call to action for, for those of you that are block producers on the GEOS Network. There's an msig up right now uh to create the account to enable the launch of eocvm uh I, don't know, is Danny still on the line and roofies anything to add it, but that's pretty much it check it out if you're BP uh msig is out. Definitely if you're in the top, 21 or or you're gonna stand by on EOS would be, would appreciate the review and a signing of that.

A

So we can get evm launched in time for what is it? April 14th to Target I believe something like that. Maybe even earlier, but anyways check it out um and thank you. Everybody we'll see you back again next week and uh we'll be sharing the video recording and and uh summary in the chat later today have a good day guys.

C

A

E