Elsa Workflows Community, 23 Aug 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Elsa Workflows Community Meeting 21 ( 2022-08-23)

Description

Meeting 21
Topics
- Retention module optimizations
- Elsa 3 Versioning

Demos
- Auth0 Elsa 2
- Background jobs in Elsa 3
How to kill jobs?

A

Hi Jay welcome back from your holiday and your run glad to see you back so does any anyone. If you have any any topics or demos, you would like to demonstrate I.

B

Could demonstrate this version history thing, but uh uh it's it's even hurting, while I'm talking right now so I can do that. Maybe next week.

A

Oh yeah, of course, yeah. So gurkan has a bit of a throat ache sore throat. So it's hard to speak, yeah, yeah, okay, okay, no.

C

Demo for me, but maybe some discussion about the retention module, I post, an issue on this and I have just some questions about how we we manage and how we call the the database with the Entity framework, because I see we use a lot of uh the specification pattern that allows the developer to to write a good code. But when we, for example, when we want to delete some some elements, there is a different call to select data to get the ID.

C

And then we pass into some method to get again some some information and at the end we just need to delete some workflow instance based on only on the ID. So we don't need to make many call against the database and get a lot of data.

C

And what I see is a minimum of CPU and I've got many timeouts I can understand that I don't have to use this kind of database with small DQ in production, but when I have only 10 or 20 workflow I think I, don't I, don't have to get some exception exception or timeout execution with this SQL I think maybe we can add some element to to query to to make best better query against the database. Okay, that sounds good I, don't know if you have some times to uh to.

A

Look on the issue sure, let's go over this topic in a second. Let's see, I do have a demo about background jobs in LS3 and a r0 demonstration in LSAT 2.. It's something I worked on today as a proof concept for a client, so I figured it's it's interesting to demonstrate, and it this is about protecting the Elsa API, using whatever identity server you you want in this example will be out of zero. But then the question is: how can now the dashboard, which is a spa?

A

A single page application make calls to the API, because the API is now protected, so the client or the dashboard needs to send better tokens. So I would like to show one one approach to that. So uh that's that's a demo to look at but yeah. Let's start with the retention module sounds like it. There's some room for improvement there do you have you mentioned an issue that you opened yeah.

C

A

32 53 all right retention, module job could cost timeout when deleting large data or when using a small DB yeah patination. So the idea here is that it takes one page at a time for a database could have many many workflow instances, so loading them all in memory at once would be insane. It would kill your application. Probably so. This is vaccination using specification here doing the filter pipeline, adding the selected instant studies, ah so it returns Wi-Fi instances and you're saying. Could we maybe improve this by not returning the whole workflow instance?

A

Just just entities or the primary key values.

C

Yes, because the data column is huge and could be huge, yeah, absolutely and and other thing is the delete, many by IDs I. Think in this method. I don't post this uh this method, but we get all the dependencies, so the bookmark Etc to be sure to that. We we delete everything in this. We do a DP context, that's set and I, don't know when we do this. If Entity framework do a get to the database. Sorry.

A

I didn't get I didn't get that if we look at.

C

And we just open the solution and okay. So, if I take a look at just open the solution. Okay, if you take a look at the Entity framework, workflow instance store. There is a delete menu by IDs async and in this one we have a I do work callback with a when, where we we did the following: DB context.set for workflow execution, log records, bookmark and work for instance, and when we did this line, I don't know if Entity framework realized a secret role to.

A

The database, so this will invoke delete many by these I think, and this supposedly should delete all the workflow execution log records for this set of workflow instance IDs, all the bookmarks and all the workloads, and are you saying this is not happening? This.

C

Is happening but in terms of performance issue, when I was a when I get my time out. I get my time out sometime in this method and I was wondering if the dbcontacts that set that asquarable dot, where, if this command realize internally by Entity framework, a call to the database to get the workflow executioner record where offline instance ID is equal to and then it loads the element in the local context, and then he is able to delete using the ID or he did the The Entity.

C

And if it does like this, there is also a perf issue, opposite load or the entity. If I want to delete 100 work, for instance, I could download first many many Android or for flow executional records right yeah.

A

Etc, etc. Yeah, even one workflow instance, could have thousands if not more workflow execution log records, and this would try and select them all, although the bats delayed with workarounds. This is, it should happen server side, but even server side. It could be a problem so, and here it's even worse here- it's gonna project them or get them from the server into local memory and then delete them one at a time just for postgres, MySQL and Oracle, but even if it happens, server side with with my with sqlite, maybe could be an issue.

A

Maybe what the solution should be is to go back to the so here's the cleaning up hosted service. It's doing it here so.

C

I I know we can check if there is a SQL query on a client side with when we debug uh we have some I have some question about. Can we uh do do? We have to continue to use specification pattern in this kind of method, or do we want to uh to use maybe some a specific, a SQL command.

A

Well, the um the reason for the specification pattern is to abstract away the underlying data storage, so we support mongodb Entity framework as an abstraction on relational databases. We support the yes SQL provider, so we have different providers and the specification pattern allows us to abstract away the concrete implementation so going dropping directly into SQL here. Does it make sense in the current architecture?

A

But what would make sense is to have a a new specification, so we Implement a custom specification called, let's say, delete workflow instances and then the Implement implementations of these of this particular specification can do the optimized querying. So if you implement the specification Handler for EF core that could have optimized SQL, the same goes for mongodb the mongodb Handler. For this particular specification.

C

A

Can we can certainly do.

C

That that's what my this is my first question and the the second question is: what about dependencies directly in in the database, using, for example, cascading delete? What about that? The cascading delete.

A

Yeah you want to use it, you mean set up cascading delete on the database layer. We.

C

Just have to delete it as a workflow instance, and all the dependencies are deleted by the.

A

Database itself, directly, yeah yeah that's an option. Of course you would have to be able to configure this dependency on the EF core level and also when using mongodb do we have a similar mechanism there? What about? Yes, SQL Etc, so you? So we need to think about these things as well, but uh if okay, if if there's a solution to that for sure just as long as it doesn't require users to specifically execute some SQL script, to make this happen for release databases and another script for mongodb.

A

Ideally, everything is taken care of from code. Okay,.

C

I have done some Evolution on the the first part in in.

B

C

The job to be sure to not load everything in memory, so I will submit a PR on this to to discuss and if you could just show me an example about the Handler to optimize. Firstly on SQL and then we can optimize on the other provider.

A

Right yeah, so let's take a look at EF core first I.

C

Don't know if you want to to show something on your brain.

A

I wasn't sharing.

C

A

C

A

Right so the place to look for in the example of the EF core providers, every provider implements one of these interfaces, so here this one implements I workflow definition store. We would be interested in the workflow instance, or this one and and this store will implement the following methods of the I store of some entity in that save is saying at the same Etc and they all receive, or some of them received. They receive a specification so in this case delete many.

A

So one way to go about this here would be to update the delete, I think or the delete menu or both where you did it by many many ID remember so this one gets a nice specification exactly so here it uses five many specifications, but what you could do is add a bit of code where you say: if specification is my retention specification, for example. Let's, let's imagine that's a new specification, then you do whatever you need to, and the retention specification can carry any data.

A

You need to perform the right query like the workflow instance, for example, or the set of workflow instances. You want to delete and then here you can do anything you like you, can inject all of the the other stores you can use the context directly. So the DB context you you can get this one directly and have access to all of the DB sets and, of course, execute SQL directly. So you can do all of the optimizations here.

A

If it's not and my retention specification then just execute the default logic now I have to admit this architecture is a little bit rigid. It's not very flexible. I wish I didn't use the specification pattern in this way. I just had the abstract word, for instance, or that wasn't this generic instead, if entity specific store would just have specific methods, then every implementation could have perfectly optimized queries to execute one housekeeping. Note guys if you're, not speaking, please mute your mics just to prevent background noise from getting into the call so yeah.

A

So you get the point right all right, so does that make sense Jay or was there anything else about this.

C

Yes, it's it's. uh We can discuss her over the time and then work on the on the under yeah sounds great I.

B

Have one question about the inversion history? In V3, you mentioned the execution workflows, execution log records I, just remembered that, like when I was working on this I noticed that we are not storing the execution log records based on the versions of the workflows, but only the workflow definition ID, so is that by Design, because when we delete a version, I cannot delete the log records because there is no versioning on that model.

A

So workflow execution log record should be linked to workflow instances but you're saying that's not the case or or you're saying it's. It's linked to workflow definition, but not to work for definition, version yeah,.

B

That's uh yeah, that's what I'm saying like uh there is a on the workflow execution log record and there is workflow definition ID, but not the version. So I was wondering if that by Design or just a mistake.

A

Let's see so here we have to work for definition ID in society. So maybe you can explain to me why we need a version ID for the workflow execution log records, because baby.

B

A

Mistake but I can't I, don't see why it would be necessary, but then again I'm not implementing the versioning, of course delete.

B

The version of the workflow I told like we should also delete the low records regarding that version.

A

I see so yeah that makes sense, but of course you will also want to then delete the workflow instances. So if you were really to workflow definition, all of its instances need to be deleted and this entity.

B

A

Dependent on its workflow instance, but I can imagine if you delete a specific word for definition version. Maybe the query could be optimized by being able to delete all workflow execution log records based on the workflow definition ID and its version, so that you don't have to first go over each work.

B

A

The instant study before you can delete this one, so yeah probably makes sense for optimized deletion.

B

Okay, yeah, okay,.

A

Yeah good question: thanks all right guys anything else, all right and I will continue with demos, I'll start with off 0 for LSAT 2.. Let me undo these changes first. So let me give you some context. As you host Elsa apis online, you probably want to protect the API endpoints. Otherwise anyone could invoke those apis and create or delete workflow definitions or access cute workflows, for example, and that's not good for obvious reasons. So, just to give you an example here are some of those endpoints.

A

This one is allows you to save a workflow definition or publish or unpublished, is so ported as well. These API endpoints they exist for client applications like the Elsa dashboard. But of course, as I mentioned, you don't want to expose this to the Internet, so you should protect it somehow and there's various ways that you can do the most typical way to do it with aspin at core applications with your Elsa workflow server is typically is to use asp.net core authentication middleware.

A

But then, if you protect your API endpoints, you need a way for your clients, in this case the Elsa dashboard to authenticate itself or at least provide whatever credentials are necessary to make the API calls to the back end. So what I did here is I set up proof of concept for a client and I would like to demo what that looks like for so as an example that you could use as well I'm using off zero, but it would work for any oauth application, so would also work with Azure ID or any identity provider.

A

So, first, let's take a look at the back end or actually, let's take a look at of zero, the the setup there. Actually, that's, not a good idea. Let's start with the back end, that might be easier to explain so so I set up a sample here. So this is an asp net, core application. It's ASP using.net6 framework, I'm, just referencing this packets, the jwtwt bearer and the also server API project. It's a very simple application. All it does basically is host Elsa.

A

So when this application runs, we have a workflow server running, but it has a couple of interesting things. First and foremost, I'm adding the application. Services, of course, like the health checks, it's optional, but it's a good idea to have cross-riching resource sharing, I think is what the course stands for.

A

This is important if your workflow server is hosted on a different origin compared to your dashboard, so adding this by default during development, I'm hosting on different origins in any case, so it makes sense, but you need to be aware of the security Ram notifications if you allow any headers and any origin, so you may not want to do this in production in production. You may want to explicitly specify what Origins are supported here: I'm, adding else API, endpoints and internally.

A

It does a couple of things like adding controllers, which is from asp.net configuring, Newton soft configuring, some routing. So this is very opinionated and it's used as what else I requires in the current version. In the new version it will also be opinionated because the the client will expect Json in a certain format, for example, or.

C

A

We'll make it configurable where you need to make sure that both the server and the client are using the same Json serialization settings and it's adding API version in.

B

A

But just to give you an idea what else API endpoints is doing. So that's that this is adding asp.net core authentication services, and here we're configuring it using JWT Bearer defaults as the authentication scheme and then.

B

A

We are configuring, JWT, Bearer and by reading some configuration from of zero, but this is not specific to r0 It's. Just in this example, it's specific because I'm using this term of zero. But if we look at the up settings, that's just the section name. You can name it anything you like, and these are all standard fields that that you can configure on the options object.

A

You get here as a as an argument, the Jade, with the bearer options so basically on the server side, there's nothing oauth specific because of or sorry off zero because of zero implements the oauth standard and open ID connect standards. This is very nice here, I'm, adding some more services. Currently this example doesn't use author authorization, but you you could, if you wanted to and finally adding the necessary Elsa services. So this is all depends injection setup.

A

Here we are going to build the application and then configuring the request pipeline, so we're using the course middleware routing, authentication, middleware and authorization middleware, which again in this example is, is optional and then here in the use endpoints, we are mapping the health checks at the root root. So if we start the application, it will perform a health check and return the response or the result of the health check. And then here we are mapping controllers, and this is the key to protecting all controllers in your application, including the ones exposed by Elsa.

A

So it's uh it will protect everything which may not be ideal in all scenarios, although maybe you can, let's say your application- has custom controllers that are supposed to be accessible by Anonymous requests. I think allow Anonymous will work, which is an attribute. You can apply to your controller but I'm, not sure, but you can you can try it, but in any case this will protect the Elsa API endpoints.

A

So now the server is protected, it's hosting on 11 000, and if we start it, we see the health check response, which is an endpoint that we did not protect, which is good because it may be useful for health check services that need access, Anonymous access to this endpoint, but the controllers they are protected.

A

So, let's see, if that's true, so let's say we want to list all of the workflows and let's not provide write any tokens, so no authentication and it's sent so here we get a response that the request is unauthorized, which makes sense because we protected the endpoints so to make it work. We need to provide an access token. So let's take a quick look. What that looks like so it's protected using JWT, Barrett token authentication mechanism, so we need to provide a better token, which means we need to provide an access token.

A

So here, I set up a simple example using the wrong client ID. Let me fix that. So this takes us to the half zero dashboard. So in the dashboard I set up a thing called an application which is basically configuration of a client. That's allowed to make calls to the of zero identity server and it typically comes with a client, ID and a client Secret in this case to to be able to request an access type. We need to provide the client ID and a secret and the audience.

A

So we are going to use this client ID this one here and we'll take the client secret. The audience is correct and the grant type is client credential. So so, if we submit this request, we get an access token, which I'm using using this test script, which is very convenient of Postman. It's a way to execute a bit of JavaScript in response to the request. So we get this as a response body. We are parsing this into a Json object and then getting the access token field, which we receive.

B

A

And storing it in an environment variable called access token, which we can then automatically use in other requests such as here yeah I'm gonna pass in that fresh access token. So now, when I hit send we get a successful response. Of course there there are no workflows in the system, but the response is valid. So now the question is: what do we do with the dashboards? So, let's switch to the dashboards Let's Pretend This isn't here for a second and just start it as you would normally do all right. So we see a blank screen.

A

If we look at the console, we see something similar. We see a 401 response when requesting some endpoint, so this makes sense. Our backend is protected. So now, let's see how we can allow the user to authenticate themselves so that they can actually use the portal or the dashboard, and for that I wrote. A small plugin called off zero plugin, which works like this. So here we are looking at a simple HTML file that hosts the designer. So that's this style you may be familiar with.

A

It's just uses the L Studio root, custom element and someone else's Studio dashboard components pointing to the back ends. We are running so that's this file just for context. What this allows you to do is it allows you to install custom plugins and that's what we're doing here at the bottom.

A

So here we are listening for the initializing event of the LC Studio root element, this one and then here there's a new function called register plug-in Factory, which lets you pass in a Lambda or function that returns a plug-in and using this approach you can return an instance of a plugin, that's initialized with application specific settings, so here I'm gonna pass in a client ID, but this is not the same client that he has used for the server, so every application needs its own application representation in F0 with its own client ID in its own settings and its own authentication flow.

A

Basically, here we were looking at the Elsa API application and now we're going to look at the the dashboard application. That's this one and, as you can see, it has its own client ID we're going to copy this. One actually, this is the correct one. Then the domain is also already correct in the audience as well domain, client, 30 and, of course, these settings are important for your applications like what are the allowed callback URLs.

A

So since the dashboard is being hosted on localhost Port, this one, it needs to be the same same for this all right and, as you can see, I just changed the settings here in this HTML file and because of how to reloading the page got refreshed, and now we do see the dashboard, which also means. If we look at the the network tab, we see that it invoked a token endpoint. That's an endpoint hosted on my tenant on out of zero, so it's requested a token and a response included. Okay, this is the payload.

A

The response was the access token here and after that, in the initialization we can now make authenticated. Api calls like getting features or if we go to workflow definitions, as you can see, it's now able to make successful calls and looking at the request, headers. It makes sense because it's, including the bearer token that we got from of zero and a better token, looks something like this. So it includes an issuer subject audiences and some other claims- and this is extensible of zero, very extensible.

A

You can configure many things, including Scopes, and that can this can include application, specific Scopes, which you can map to permissions that are required on your backend. So it's very, very flexible. So let's take a look at the plugin, the off zero plugin. So it comes with the Elsa core, but you could implement it yourself as well, using stencil or playing JavaScript. It's the same as any other designer plugin. When you use typescript you implement Elsa plugin, you implement a Constructor. In this case.

A

This Constructor accepts of zero client options with map to this object here and I just store it in a private field and then in this Constructor I, start initializing, some event: bus events that I'm interested in specifically the initialize event, which is fired by the root components.

B

A

This allows you to do things like, in this case creating an r0, API clients testing. If the user is already authenticated, because then we don't need to do anything, but if we are not authenticated, we need to get an access token and the way that works is by basically logging in to to alph zero. So we redirect to the off serial login page where the user can enter their credentials and once the user did, that officer will redirect back to this page using the origin, and that means we'll get back to this somebody's making.

A

Some some noise disconnect.

C

A

Yeah, so when we get redirected, we get through this code again, of course, because the page is reloaded and this time we will have have code, query string, parameter and if that's the case, we let the off zero client API, handle that that code, so it will interpret it, probably store some cookie or whatnot. It's all that's all handled by the officer, API client, which is very convenient, and here I'm just updating the URL, basically because otherwise you would see some ugly code equals and some value and something else.

A

So it's basically going to remove that from history, which is also better for security. I assume. So that's one part. So that's getting the access token or actually getting the user authenticated. And this will this handle redirect callback will store the access token somewhere and it will make it available to the application to Elsa dashboard through the get token silently it will internally cast the access token and presumably also refresh it I'm, not sure, on the implementation.

A

Details I have yet to look into that, but it will give you the access token and by the way this is configure of middleware, which is invoked from the HTTP client credit event. Bus, and this is invoked by Elsa as soon as it's about to instantiate an anxious client, which is like an HTTP client. This event gets fired, giving all plugins the opportunity to set up the HTTP client, for example, XCS supports middleware, which is very convenient for us, because that means we can attach custom headers to its outgoing requests.

A

So in this case we want to always attach the access token. So that's what we're doing here and that's basically it that's.

C

Great and in your demonstration, we we do not see the login page, because you are using the same user as the o0 yeah.

A

That's a good question, so actually that's because I was already signed into my tenant on r0, but let me show you by clearing the data close all windows and now that all my my caches emptied I, don't have any cookies anywhere anymore. We will now see the login page, so we'll go to localhost this and now we do see the login page because you see the domain here. So this is my tenant on of zero and I need to log in with a user, that's known to have zero.

A

So in the standard I created one demo user using my personal email address, so this one. So this is my email, my password and now I will sign in and I'm signing yeah all right. That's great!.

C

I think we couldn't then add more custom plugin on the dashboard or open ID in.

A

General, exactly a more generic component would makes a lot of sense, so we could have an open ad connect module where everything is configurable through the Constructor like here I did it's very specific to file zero, but you're right, we could add an open ID connect generic module, making it very flexible and accessible to many users, because this is a much requested feature.

C

For on my side, I already use a0, but I use this also an identity server as a knight entity broker between my solution and zero. So I need an open, ID, plugin, so I think I I will take a look at this implementation of all zero and derive an open, ID plugin. For this.

A

That makes sense, yeah and that would be great to have if you want to contribute another module called open ID. That's uh that's wonderful! Let me know if you need any help with that.

C

Yeah and I think after this is the next step, it could be to add some role and policy on the different API or to be able to make some policies, for example, administrator a viewer to see the workflow instances, for example, and be able to get the role in the token and only authorize some people to access to access. Some part of the dashboard.

A

Yeah, absolutely that would be terrific, would probably require some changes to the API. For example, if we look at the else server API project, where we have the endpoints and we look at let's say: workflow definitions delete so right now it's it's public, but we protected this one in program here using this. So this is protecting everything, but doesn't take into account policies or roles, and that kind of thing, although you can provide policy names, but this would be provided to all controllers.

A

But if you want more grinder earlier control over what certain users can do using their permissions, we need something like authorize and then here we should either provide a policy name and or a set of roles, but this would be very opinionated right. It would still be helpful. We can't just Define a set of policies and then it's up to the application implementer to configure the requirements for each policy here. So.

C

A

Still very flexible, yes,.

C

Okay to see if we can or if we have to create some generic role in a description to add some attribute in the controller and allow the example to map. If we decide, for example, to create some basic role and add authorized attributes, we have to let the developer and the implementer to map his own role on our generic.

A

Code yeah exactly something like that: yeah.

B

I think one of the options currently that we can do right now is that we can write just custom authorization Handler, like one big class, to handle all kinds of requests, and then you can just put the conditions. If this is the end point allow or deny so it could be helpful if we can just break it down, for example, to operations whether it's a create edit delete or update.

B

Then in that case we can just write our own custom Handler, without even forcing any policy in the specific controllers, just by defining some operations that we are trying to achieve, and then we can just write our own handles and Define. Whatever logic that you want.

A

Yeah exactly that's another approach. You could take yeah great okay, nice. Unless there's any additional comments or questions, then I will share their background jobs, yeah.

B

Yeah, maybe because I missed the initial I wanted to see the current credential manager, because I try to run the demo myself, but since I'm currently away I couldn't run it.

A

A recurrential manager, the new feature you mean that was recently.

B

A

uh Yeah yeah, so I didn't demo this one and I have yet to try it myself, I've seen some demos from the developer.

B

But I'm gonna ask.

A

Them to uh to see if they may be interested in in demonstrating it on.

B

A

That would be, that would be pretty cool, otherwise I will go through it. Myself and and I can do a demo as well instead, but we will see. Thank you all right. So the background jobs in elsa3 I mentioned this.

A

uh The last meeting, I think or the meeting before that, and it's basically a way to as a developer, to implement some work that you want to execute in the background, so not in response necessarily to an HTTP request or not based on a timer you, you may want to be able to schedule it manually or in response to a timer or in response or a request, but it should be non-blocking. That's basically the idea there are solutions out there like Hang Fire, like quartz.net, and that's that's great. Typically.

A

The way you would work with that is. You would have some code that then schedules a job and the job gets handled by The, Hang, Fire or bikewatch.net. What else are three? We have that as well, using Hang, Fire or using quartz.net, but with this feature that I'm about to show these jobs now become available as activities. So you can orchestrate what jobs you want to execute and then block the workflow until the job is finished, and then you can continue the workflow. So it's a nice way of an easy way of orchestrating work.

A

So yeah, that's what I want to show. So first, let's take a look at an example here. I have an example: job called index blockchain, and it's going to do something very simple: gonna write something to the console out, wait for five seconds and then write something else, just to show that this is a long running process in the background and how you can use it from a workflow.

C

A

So, let's see this one, so here we look at the designer. Here are the activities, and here we see this job that I created. So it's under the jobs category I will add. It here add to activities for demo purposes. This will say start this will index the blockchain, and this will tell us when it's done and the reason I'm showing the start and end activity is to demonstrate that this workflow will be suspended.

A

While this is executing and for that to to properly demo, that I should probably increase the time from Five Seconds to let's say 15 seconds, and then we will publish.

B

A

Copy the definition ID, so we can run it from the API all right so before I run. Let's take a look at the console here. Let's clean this up and send so here we see start and the first line of the job which is now running. ah I was a little bit too late. It finished already. Let's start again, of course we see the that it's finished and it ended, but I want to show that it was really suspended. I'll refresh this- and here we see this workflow- is it see the state?

A

Has the status of running at sub status is executing yeah, so this is, it is suspended. It's out of memory, it's a little bit hard to prove it without actually going through a debugger, but I'll leave that as an exercise to you but yeah. It's finished the work, and then the cool thing about this is that you don't have to create a custom activity.

A

You just Implement a job by implementing or inheriting from the job, Base Class and the way this works is that there's there's an activity, type provider that looks all job implementations and then basically wrap its job into a an activity that you that you can then choose from the from the Picker here and the first time I mentioned this I think. Was you Muhammad to ask the question? What's the difference between a job activity and a regular activity? Yes,.

B

A

And really, the only difference is that this job executes in the background, while an activity runs synchronously right. So when this executes workflow execution doesn't continue until this is done- and this happens synchronously in memory, the workflow remains active Etc, but it did make me think so more that why not take it a step further and allowing activities themselves to declare or to be configured to run asynchronously. So in parallel it would, for example, allow you to Branch out to multiple branches of execution.

A

Have multiple activities execute without its Brands having to wait for the other one? To finish so that's something I'm research, saying as we speak, and this way we could have. Let's say you have a custom activity that makes an API call to some Azure function, but it's potentially very slow could take minutes.

A

Maybe you don't want the entire workflow to be in memory, while that is happening so those kind of activities you want to execute asynchronously in the background, like a job I'm, not sure it's going to be implemented like a job, but I am wondering what would be a good way to configure an activity to execute in the background, would it be a setting here, a settings that you can configure for even for the right line activity, for example, to execute in the background, or should it be kind of activity?

A

So if you look at a given activity like this, one I think it has a property called activity kind, but maybe it's not implemented in three I. Don't.

B

A

It here, but in lc2 there's this concept of activity kites or something like that here is an enumeration, so an activity could be an action could be a trigger like HTTP endpoint or maybe it could be a job. This is currently not used, it's not implemented. So that's a question if.

B

A

Any thoughts on that, please do. Let me.

B

Know so so, basically, and like an action activity, if it is running for a long time when I look at the workflow status, I will see that it is running. It will keep running until the activity finish. But what I understand in case of a job the workflow becomes suspended, and it will wait for the background to finish. Then it will resume execution of the auto.

A

That's right, although you make a good point there, because even though the workflow will be put out of memory and persisted the status in Elsa, 3 will remain running, but it will not be actively running. It's just a status that the workflow isn't finished. Yet even it may it may be waiting for user input, and while it's waiting for that, the workflow doesn't exist in memory, only ones that.

B

A

Provided will it be loaded from memory, but the logical status will be running and the subsidies will be executing? Let's see.

B

What? Because, in answer 2, if I see that workflow is running for a long time, I start to get worried. Yeah.

A

Yeah yeah, absolutely because in Elsa 2 it should be suspended. If, if you expect it.

B

A

For something it should not be actively running.

B

A

Three, it will be running and there's a bunch of sub statuses uh suspended, so it could be. The main status could be running, but then the sub status should be suspended if it's waiting for a job. To finish, for example, we can pass some data to the job, not right now, but that will be added as well. That's very important and not just pass in data. You should also be able to provide a result or some output can I terminate a job not currently, but that's a good question.

A

You should be able to terminate it somehow at the very least, it should be an API to do so and then, of course, maybe from the workflow. Maybe you want to split execution and then start a job on the one hand and have a Timeout on the other hand, and if that one fires, maybe you want to kill the job, for example, or maybe jobs should have a Max timeout or a timeout value that you can configure. But that's a that's a good question. Let me let me write it down so yeah.

A

That's demo about Job activities and, as I said, I'm investigating ways to implement activities that.

B

A

Run in the background, which would be uh would solve a lot of issues actually that I know of yeah.

B

But if some activities have that assumption that they keep they stay in the memory for the course of the execution, maybe it will cause issues I'm, not sure.

A

Yeah, if we, if we make this configurable, so maybe some activities will execute in the background and others won't and then we know up front which ones we could mark them with the.

B

A

Kinds maybe yeah yeah, I'm, leaning towards that as well.

B

A

And once we have that, then maybe there's no need for a job itself to be exposed as an activity. At the same time, it could be an interesting option because in some scenarios you just want to execute a job outside the context of a workflow, as is the case with you know when you implement Hang Fire jobs or quartz.net yeah.

B

Definitely, the concept of the government is very interesting yeah. If you can manage it from the workflow, it will be very helpful, so I can just run something and then wait for it to finish and come back to it without worrying that my workload is going out of control. That is really interesting.

A

Yeah, exactly all right, I actually need to go. If there's any questions, let me know yeah.

B

Okay and so 2.9.

A

Is else 2.9 there's still issues outstanding? Let's take quick look yeah. We still need to fix these issues. I haven't had time.

B

A

Into them, but yeah I I, don't think we should release without these ones done.

B

Yeah yeah, that's fine, so probably in September, then it was August is a bit difficult. Yeah.

A

I think so we are already 23rd yeah yeah. Let's.

B

A

Aim for September sure, and in fact let me update the Milestone I- don't want to put too much pressure, so I'm gonna put.

B

In the last day of September.

A

And, of course, if you, uh if you have any availability, please take a look if you can and we'll try to we'll try to help awesome thanks. So much guys thanks for attending today and we'll see you next week.