We're looking for automated way to horizontally, vertically scale the pull of self hosted integration runtime virtual machines used in ADF.
Reading Microsoft docs does not provide answer.
Well, I don't have the experience, so I can only give you a theoretical answer, but maybe it's helpfull for you.
AFAIK, neither way is configurable out-of-the-box. For scale-out you'll have to deploy an additional IR machine yourself. So probably you'll want to create an image that you can provision from docker or kubernetes and has the IR and pre-requirements installed. The IR installation provides an PowerShell script that can be used to create an automated connection.
For scale-up/down, you'll have to run some script that scales your vm. In an IaaS solution (f.e.) Azure VM, that should be doable with an API call to change your VM.
For both cases you'll have to have some kind of montitor in place that monitors the IR loads and makes changess as needed. I think the measures provided in the Data Factory should do. Maybe you can use Log Analyics to monitor the loads.
I'm curious about your use case for this.
My solution is just for scaling out/in since the VM must be restarted if you are scaling up/down, which causes downtime and job failures etc.
At a high level this solution requires just 3 simple things:
Azure Metric Alert that fires when Scale-Out should occur (VM Start)
Azure Metric Alert that fires when Scale-In should occur (VM Deallocation)
Logic App that is triggered by Azure Alert and actually executes the Start/Stop of the VM, along with any other automation associated with this (eg posting to a Teams channel when Scale in/out occurs)
Here are more of the details surrounding how we setup the conditions for the alerts, but the main thing to keep in mind is (IR CPU %, IR queue length, Number of Nodes, and possibly IR Memory)
Scale-Out
Scale-In
Actions for Alerts
As you can see below we have the alert triggering 1 Logic App, using the payload that is passed to the Logic App, you can determine if the Logic App should be starting the VM, or stopping the VM. (As well as any other additional actions)
Logic App
There is a small chance that due to timing (and depending on how many ADF's the IR is shared to), that pipeline activities could be sent to Node 2 at the same time a deallocation command is sent to the VM for Node 2. I have not seen this as of yet, but adjusting the alert conditions based on your need could help avoid this. Feel free to play around with the conditions of the alerts, granularity, thresholds, etc. This is not a one size fits all solution.
Related
We have a service running as an Azure function (Event and Service bus triggers) that we feel would be better served by a different model because it takes a few minutes to run and loads a lot of objects in memory and it feels like it loads it every time it gets called instead of keeping in memory and thus performing better.
What is the best Azure service to move to with the following goals in mind.
Easy to move and doesn't need too many code changes.
We have long term goals of being able to run this on-prem (kubernetes might help us here)
Appreciate your help.
To achieve first goal:
Move your Azure function code inside a continuous running Webjob. It has no max execution time and it can run continuously caching objects in its context.
To achieve second goal (On-premise):
You need to explain this better, but a webjob can be run as a console program on-premise, also you can wrap it into a docker container to move it from on-premise to any cloud but if you need to consume messages from an Azure Service Bus you will need an On-Premise-Azure approach connecting your local server to the cloud with a VPN or expressroute.
Regards.
There are a couple of ways to solve the said issue, each with slightly higher amount of change from where you are.
If you are just trying to separate out the heavy initial load, then you can do it once in a Redis Cache instance and then reference it from there.
If you are concerned about how long your worker can run, then Webjobs (as explained above) can work, however, that is something I'd suggest avoiding since its not where Microsoft is putting its resources. Rather look at durable functions. Here an orchestrator function can drive a worker function. (Even here be careful, that since durable functions retain history after running for very very very long times, the history tables might get too large - so probably program in something like, restart the orchestrator after say 50,000 runs (obviously the number will vary based on your case)). Also see this.
If you want to add to this, the constrain of portability then you can run this function in a docker image that can be run in an AKS cluster in Azure. This might not work well for durable functions (try it out, who knows :) ), but will surely work for the worker functions (which would cost you the most compute anyways)
If you want to bring the workloads completely on-prem then Azure functions might not be a good choice. You can create an HTTP server using the platform of your choice (Node, Python, C#...) and have that invoke the worker routine. Then you can run this whole setup inside an image on an AKS cluster on prem and to the user it looks just like a load balanced web-server :) - You can decide if you want to keep the data on Azure or bring it down on prem as well, but beware of egress costs if you decide to move it out once you've moved it up.
It appears that the functions are affected by cold starts:
Serverless cold starts within Azure
Upgrading to the Premium plan would move your functions to pre-warmed instances, which should counter the problem you are experiencing:
Pre-warmed instances for Azure Functions
However, if you potentially want to deploy your function/triggers to on-prem, you should spin them out as microservices and deploy them with containers.
Currently, the fastest way would probably be to deploy the containerized triggers via Azure Container Instances if you don't already have a Kubernetes Cluster running. With some tweaking, you can deploy them on-prem later on.
There are few options:
Move your function app on to premium. But it will not help u a lot at the time of heavy load and scale out.
Issue: In that case u will start facing cold startup issues and problem will be persist in heavy load.
Redis Cache, it will resolve your most of the issues as the main concern is heavy loading.
Issue: If your system is multitenant system then your Cache become heavy during the time.
Create small micro durable functions. It will be not the answer of your Q as u don't want lots of changes but it will resolve your most of the issues.
I have an azure cloud service which scales instances out and in. This works fine using some app insights metrics to manage the auto-scaling rules.
The issue comes in when the scales in and azure eliminates hosts; is there a way for it to only scale in an instance once that instance is done processing its task?
There is no way to do this automatically. Azure will always scale in the highest number instance.
The ideal solution is to make the work idempotent and chunked so that if an instance that was doing some set of work is interrupted (scaling in, VM reboot, power loss, etc), then another instance can pick up the work where it left off. This lets you recover from a lot of possible scenarios such as power loss, instead of just trying to design something specific for scale in.
Having said that, you can manually create a scaling solution that only removes instances that are not doing work, but doing so will require a fair bit of code on your part. Essentially you will use a signaling mechanism running in each instance that will let some external service (a Logic app or WebJob or something like that) know when an instance is free or busy, and that external service can delete the free instances using the Delete Role Instances API (https://learn.microsoft.com/en-us/rest/api/compute/cloudservices/rest-delete-role-instances).
For more discussion on this topic see:
How to Stop single Instance/VM of WebRole/WorkerRole
Azure autoscale scale in kills in use instances
Another solution but this one breaks an assumption that we are using Azure cloud service; if you use app services instead of the cloud service you will be able to setup auto scaling on the app service plan effectively taking care of the instance drop you are experiencing.
This is an infrastructure change so it's not a two click thing but I believe app services are better suited in many situations including this one.
You can look at some pros and cons but if your product is traffic managed this switch will not be painful.
Kwill, thanks for the links/information, the top item in the second link was the best compromise.
The process work length was usually under 5 minutes and the service already had re-handling of failed processes, so after some research it was decided to track state of when the service was processing a queue item and use a while loop in the RoleEnvironment.Stopping event to delay restart and scale-in events until the process had a chance to finish.
App Insights was used to track custom events during the on stopping event to track how often it completes vs restarts during the delay cycles.
At the moment we are running our application on an AWS Beanstalk but are trying to determine the suitablilty of Azure.
Our biggest issue is the amount of wasted CPU time we are paying for but not using. We are running on t2.small instances as these have the min amount of RAM we need but we never use even the base amount of CPU time allotted. (20% for a t2.small ) We need lots of CPU power during short bursts of the day and bringing more instances on line in advance of this is the only way we can handle it.
AWS Lambda looks a good solution for us but we have dependencies on Windows components like SAPI so we have to run inside of Windows VMs.
Looking at Azure cloud services we thought using a Web role would be best fit for our app but it seems a Web role is nothing more than a Win 2012 VM with IIS enabled. So as the app scales it just brings on more of these VMs which is exactly what we have at the moment. Does Azure have a service similar to Lambda where you just pay for the CPU processing time you use?
The reason for our inefficient use of CPU resources is that our speech generation app uses lost of 3rd party voices but can only run single threaded when calling into SAPI because the voice engine is prone to crashing when multithreading. We have no control over this voice engine. It must have access to a system registry and Windows SAPI so the ideal solution is to somehow wrap all dependencies is a package and deploy this onto Azure and then kick off multiple instances of this. What "this" is I have no Idea
Microsoft just announced a new serverless compute service as an alternative to AWS Lambda, called Azure Functions:
https://azure.microsoft.com/en-us/services/functions/
http://www.zdnet.com/article/microsoft-releases-preview-of-new-azure-serverless-compute-service-to-take-on-aws-lambda/
With Azure Functions you only pay for what you use with compute metered to the nearest 100ms at Per/GB price based on the time your function runs and the memory size of the function space you choose. Function space size can range from 128mb to 1536mb. With the first 400k GB/Sec free.
Azure Function requests are charged per million requests, with the first 1 million requests free.
Based on the documentation on Azure website here: https://azure.microsoft.com/en-in/campaigns/azure-vs-aws/mapping/, the services equivalent to AWS Lambda are Web Jobs and Logic Apps.
The most direct equivalent of Lambda on Azure is Azure Automation which does a lot of what Lambda does except it runs Powershell instead of Node etc. It isn't as tightly integrated into other services like Lambda is, but it has the same model. i.e. you write a script, and it is executed on demand.
I presume by SAPI you are refering to the speech API? If so you can create Powershell modules for Azure, and they can include dll files. In which case you could create a module to wrap around the SAPI dll, and that should do what you are looking for.
If you want a full compute environment, without the complexity of multiple machines when you run. You could use Azure Batch which would be the Azure recommended way of running what you are looking for.
The cost benefit you need to evaluate would be how much quicker your solution would run against a native .net stack (in batch), and if performance is significantly degraded when run from Powershell.
Personally I would give Automation a try, it is surprisingly powerful.
There is something called "Cloud Service" in azure which allows you to run code on a pure VM. Scaling options on these include such things as CPU%, queue size, etc. If you can schedule your needs, Azure allows you to easily set up a scheduled scaler, i.e. 4 VM's from 8AM until 08:10AM, and of course, in Azure, you pay by the minute, so it could be a feasible solution.
I'd say more, but the documentation in Azure is really so great that I'd be offending them by offering my "translation" here. Checkout azure.com for more info :)
I've had a few outages of 10 to 15 minutes, because apparently Microsoft had a 'blip' on their storages. They told me that it is because of a shared file system between the instances (making it a single point of failure?)
I didn't understand it and asked how file share is involved, because I would assume a really dumb stateless IIS app that communicates with SQL Azure for its data.
I would assume the situation below:
This is their reply to my question (I didn't include the drawing)
The file shares are not necessarily for your web app to communicate to
another resources but they are on our end where the app content
resides on. That is what we meant when we suggested that about storage
being unavailable on our file servers. The reason the restarts would
be triggered for your app that is on both the instances is because the
resources are shared, the underlying storage would be the same for
both the instances. That’s the reason if it goes down on one, the
other would also follow eventually. If you really want the
availability of the app to be improved, you can always use a traffic
manager. However, there is no guarantee that even with traffic manager
in place, the app doesn’t go down but it improves overall availability
of your app. Also we have recently rolled out an update to production
that should take care of restarts caused by storage blips ideally, but
for this feature to be kicked it you need to make sure that there is
ample amount of memory needs to be available in the cases where this
feature needs to kick in. We have couple of options that you can have
set up in order to avoid any unexpected restarts of the app because of
a storage blip on our end:
You can evaluate if you want to move to a bigger instance so that
we might have enough memory for the overlap recycling feature to be
kicked in.
If you don’t want to move to a bigger instance, you can always use
local cache feature as outlined by us in our earlier email.
Because of the time differences the communication takes ages. Can anyone tell me what is wrong in my thinking?
The only thing that I think of is that when you've enabled two instances, they run on the same physical server. But that makes really little sense to me.
I have two instances one core, 1.75 GB memory.
My presumption for App Service Plans was that they were automatically split into availability sets (see below for a brief description) Largely based on Web Apps sales spiel which states
App Service provides availability and automatic scale on a global data centre infrastructure. Easily scale applications up or down on demand, and get high availability within and across different geographical regions.
Following on from David Ebbo's answer and comments, the underlying architecture of Web apps appears to be that the VM's themselves are separated into availability sets. However all of the instances use the same fileserver to share the underlying disk space. This file server being a significant single point of failure.
To mitigate this Azure have created the WEBSITE_LOCAL_CACHE_OPTION which will cache the contents of the file server onto the individual Web App instances. Using caching in lieu of solid, high availability engineering principles.
The problem here is that as a customer we have no visibility into this issue, we've no idea if there is a plan to fix it, or if or when it will ever be fixed since it seems unlikely that Azure is going to issue a document that admits to how badly this has been engineered, even if it is to say that it is fixed.
I also can't imagine that this issue would be any different between ASM and ARM. It seems exceptionally unlikely that there was originally a high availability solution at the backend that they scrapped when ARM came along. So it is very likely that cloud services would suffer the exact same issue.
The small upside is that now that we know this is an issue, one possible solution would be to deploy multiple web apps and have a traffic manager between them. Even if they are in the same region, different apps should have different backend file servers.
My first action would be to reply to that email, with a link to the Web Apps page, (and this question) with a copy of the quote and ask how to enable high availability within a geographic region.
After that you'll likely need to rearchitect your solution!
Availability sets
For virtual machines Azure will let you specify an availability set. An availability set will automatically split VMs into separate update and fault domains. Meaning that servers will end up in different server racks, and those server racks won't get updates at the same time. (it is a little more complex than that, but that's the basics!)
Azure Web Apps do used a shared file storage. The best way to think about it is that all the instances of your app map to the same network share that have your files. So if you modify the files by any mean (e.g. FTP, msdeploy, git, ...), all the instances instantly get the new files (since there is only one set of files).
And to answer your final question, each instance does run on a separate VM.
Is it possible to create one or several azure VMs on my local machine? I want to create a web app and load test it locally, without the need of putting it in the cloud. I'm thinking at the following scenario: I have a local VM running a IIS server with my web app; I use a tool to generate a lot of load; I need to deploy the second VM containing the same things as the first VM. The downtime of the web app should be equal to 0(hopefully).
Clarification(update):
I want to achieve the following: create a web app and a monitoring app(CPU,Memory) and deploy them on one VM. On a load test, if the VM cannot handle it(e.g. CPU goes above 80%), I want to programmatically deploy a new VM(with the same configuration, having both the web app and the monitoring app), such that no downtime occurs.
Azure has several ways for you to host sites.
Virtual Machines is just that, normal VMs. You can create them locally and upload them, but everything is up to you, including how to handle upgrades. If that is what you need to do then I don't know how you would handle upgrades with no down time; though, you can add multiple VMs to a load balancer and then upgrade them one at a time.
It sounds like what you really want to explore is Cloud Services. You can run one or more VMs locally in the emulator, upgrade with no down time once in the cloud, implement auto scaling (you will have to use a tool or write some code).
Alternatively you may want to look at Azure Web sites, but that is a completely different concept and you can't really test load and load balancing locally the same way.
Based on your statement that you essentially want to auto-scale your application you want to look at Cloud Services with Auto Scaling. However, you can't fully test this in the cloud emulator - but you can test your logic.
Background
Azure Cloud Services is designed for this kind of thing; You don't really work with VMs in the way you may be used to, instead you create a package that Azure then deploys to as many servers as you like. Once up and running, you can manually go into the management console and increase or decrease the number of active servers simply by moving a slider. Of course, you want to do this automatically, so you have a few options.
There is a management API you can use to change the number of servers. So, it would be quite simple to write a bit of code that you spin up in another thread from WebRole.Start and that simply sits and monitors the CPU on the machine and then calls the management API to spin up a new server instance if your CPU goes over a certain treshold. Okay, locally you can only test that the call to the management API is made, you won't actually see the new server coming up. But, if you grab your free trial of Azure and just try it you will see that you really don't need to test that part - it just works.
However, in practice there is an awful lot more to auto scaling. Here are some of the things you need to consider;
Even relatively idle web servers will often spike briefly to 100% so just having a simple treshold is unlikely to be good enough; You need to decide on how long the server needs to be over a certain treshold before you spin up another server instance.
What happens when you have more than one server? And, on Azure, you should always have at least two servers to ensure you have resilience. Note that the idea with Cloud Services really is to have many small servers rather than a few big servers. You pay per core, not per number of servers.
Imagine you currently have three servers and one is really busy for some reason and the other two are idle. Do you want to spin up a fourth server?
Imagine you currently have two servers and they are both quite busy. Do you really want them both to start a new server so you end up with four servers running?
There are several ways to handle these challenges. For starters, rather than having monitor programs running locally on each server, you are better of moving that monitoring outside; Azure comes with the ability to dump performance metrics to table storage at whatever interval you choose. You can then run an external program that retrieves the performance data over time from all your current servers and then reason about the overall workload before deciding to spin up or shut down additional servers. Now, you can of course host that external monitor program in a separate thread on each of your webroles to give your monitoring resilience - but the key point is that the monitoring program doesn't monitor the server it runs on, it monitors all the servers. You will, of course, still have to deal with stopping multiple monitoring program instances from all starting and stopping servers. One way to do is to place stop/start commands onto an Azure "message queue" (there are a few different types) and use the built-in "de-duper" which will automatically delete identical commands that are put on the queue within a certain time window (I am over simplyfing but you get the idea).
The actual answer
Really, though, you want to look at the Auto Scaling Application Block which will do most of this for you. I guess that is the real answer to your question, but I wanted to provide a bit of context first.
Again, I recognise you asked for how to test this locally - but I believe that that question doesn't really make sense in the context of Azure and I hope the above information helps.
I'm pretty sure you can't do that and it wouldn't make sense anyway. If you want load testing, you need to run that in an environment as similar to production as possible and that means you have to run your application is Azure cloud. How else do you know that the load will actually be processed fine on real cloud?