I'm having trouble understanding the config definitions of a task.
I want to understand the resources. There are a few options (if we talk only about memory):
memory
containerDefinitions.memory
containerDefinitions.memoryReservation
There are a few things I'm not sure about.
First of all, the docs say that when the hard limit is exceeded, the container will stop running. Isn't the goal of a container orchestration service to keep the service alive?
Root level memory must be greater than all containers memory. In theory I would imagine once there aren't enough containers deployed, new containers are created for the image. I wouldn't like to use more resources than I need, but if I reserve the memory on root level, first, I do reserve much more than needed, and second, if my application receives a huge load, the whole cluster will shut down if the memory limit is exceeded or what?
I want to implement a system that auto-scales, and I would imagine that this way I don't have to define resources allocated, it just uses the amount needed, and deploys/kills new containers if the load increases/decreases.
For me there are a lot of confusion around ECS, and Fargate, and how it works, how it scales, and the more I read about it, the more confusing it gets.
I would like to set the minimum amount of resources per container, at how much load to create a new container, and at how much load to kill one (because it's not needed anymore).
P.S. not experienced in devops in general, I used kubernetes at my company, and there are things I'm not clear about, just learning this ECS world.
First of all, the docs say that when the hard limit is exceeded, the container will stop running. Isn't the goal of a container orchestration service to keep the service alive?
I would say the goal of a container orchestration service is to deploy your containers, and restart them if they fail for some reason. A container orchestration service can't magically add RAM to a server as needed.
I want to implement a system that auto-scales, and I would imagine that this way I don't have to define resources allocated, it just uses the amount needed, and deploys/kills new containers if the load increases/decreases.
No, you always have to define the amount of RAM and CPU that you want to reserve for each of your Fargate tasks. Amazon charges you by the amount of RAM and CPU you reserve for your Fargate tasks, regardless of what your application actually uses, because Amazon is having to allocate physical hardware resources to your ECS Fargate task to ensure that much RAM and CPU are always available to your task.
Amazon can't add extra RAM or CPU to a running Fargate task just because it suddenly needs more. There will be other processes, of other AWS customers, running on the same physical server, and there is no guarantee that extra RAM or CPU are available on that server when you need it. That is why you have to allocate/reserve all the CPU and RAM resources your task will need at the time it is deployed.
You can configure autoscaling to trigger on the amount of RAM your tasks are using, to start more instances of your task, thus spreading the load across more tasks which should hopefully reduce the amount of RAM being used by each of your individual tasks. You have to realize each of those new Fargate task instances created by autoscaling are spinning up on different physical servers, and each one is reserving a specific amount of RAM on the server they are on.
I would like to set the minimum amount of resources per container, at how much load to create a new container, and at how much load to kill one (because it's not needed anymore).
You need to allocate the maximum amount of resources all the containers in your task will need, not the minimum. Because more physical resources can't be allocated to a single task at run time.
You would configure autoscaling with the target value, of for example 60% RAM usage, and it would automatically add more task instances if the average of the current instances exceeds 60%, and automatically start removing instances if the average of the current instances is well below 60%.
Related
I have a question about Cloud Run: If I setup my service with 4GB of RAM and 2 vCPUs, for example, differing from the standard 256MB and 1 vCPU, will I have to pay much more even if I never consume all the resources I have made available? For example again, let's say that I set --memory to 6GB and no request ever consumes more than 2GB, will I pay for 6GB of RAM or 2GB, considering that the peak of usage was 2GB?
I am asking because I want to be sure that my application will never die out of memory, since I think that the default 256MB of Cloud Run isn't enough for me, but I want to be sure of how Google charges and scales.
Here's a quote from the docs:
You are billed only for the CPU and memory allocated during billable time, rounded up to the nearest 100 milliseconds.
Meaning, if you allocated 4GB of memory on your Cloud Run service, you're still billed with 4GB whether it's underused or not.
On your case, since you want to make sure that requests don't run out of memory, then you can dedicate an instance resource to each request. With this, you just find out the cheapest memory setting that can run your requests and limit the concurrency setting to one.
Or, take advantage of concurrency (and allocate higher memory) because Cloud Run allows concurrent requests so you have control on how many requests can share resources before it starts a new instance, which can be a good thing as it helps drive down costs and minimize starting many instances (see cold starts). This can be a better option if you are confident that certain amount of requests can share an instance without running out of memory.
When more container instances are processing requests, more CPU and memory will be used, resulting in higher costs. When new container instances need to be started, requests might take more time to be processed, decreasing the performances of your service.
Note that each approach have different advantages and drawbacks. You should take note of each before making a decision. Experimenting with the costs by using GCP Calculator can help (the calculator includes Free Tier on the computation).
I am running the ECS cluster that currently has 3 services running on T3 medium instance. Each of those services is running only one task which has a soft memory limit of 1GB, the hard limit is different for each (but that should not be the problem). I will always have enough memory to run one, new deployed task (new one will also take 1GB, and T3 medium will be able to handle it since it has 4GB total). After the new task is up and running, the old one will be stopped and I will have again 1GB free for the new deployment. I did similar to the CPU (2048 CPU, each task has 512, and 512 free for new deployments).
So everything runs fine now, but I am not completely satisfied with this setup for the future. What will happen if I need to add another service with another task? I need to deploy all existing tasks and to modify their task definitions to use less CPU and memory in order to run this new task (and new deployments). I am planning to get a reserved EC2 instance, so it will not be easy to swap the current EC2 instance with the larger one.
Is there a way to spin up another EC2 instance for the same ECS cluster to handle bursts in my tasks? Also deployments, it's not a perfect scenario to have the ability to deploy only one task, and then wait for old to be killed in order to deploy the next one, without downtimes.
And biggest concern, what if I need new service and task, I need again to adjust all others in order to run a new one and deploy others, which is not very maintainable and what if I cannot lower CPU and memory more because I already reached the lowest point in order to run the task smoothly.
I was thinking about having another EC2 instance for the same cluster, that will handle bursts, deployments, and new services/tasks. But not sure if that's possible and if that's the best way of doing this. I was also thinking about Fargate, but this is much more expensive and I cannot afford it for now. What do you think? Any ideas, suggestions, and hints will be helpful since I am desperate to find the best way to avoid the problems mentioned above.
Thanks in advance!
So unfortunately, there is no out of the box solution to ensure that all your tasks run on min possible (i.e. one) instance. You can use our new feature called Capacity Providers (CP), which will allow you to ensure the minimum number of ec2 instances required to run all your tasks. The major difference between CP vs ASG is that CP gives more weight to task placement (where as ASG will scale in/out based on resource utilization which isn't ideal in your case).
However, it's not an ideal solution. Just as you said in your comment, when the service needs to scale out during a deployment, CP will spin up another instance, the new task will be placed on it and once it gets to Running state, the old task will be stopped.
But now you have an "extra" EC2 instance because there is no way to replace a running task. The only way I can think of would be to use a lambda function that drains the new instance, which will move all the service tasks to the other instance. CP will, after about 15 minutes, terminate this instance as there are no tasks are running on it.
A couple caveats:
CP are new, a little rough around the edges, and you can't
delete/modify them. You can only create or deactivate them.
CP needs an underlying ASG and they must have a 1-1 relationship
Make sure to enable managed scaling when creating CP
Choose 100% capacity target
Don't forget to add a default capacity strategy for the cluster
Minimizing EC2 instances used:
If you're using a capacity provider, the 'binpack' placement strategy minimises the number of EC2 hosts that are used.
However, there are some scale-in scenarios where you can end up with a single task running on its own EC2 instance. As Ali mentions in their answer; ECS will not replace this running task, but depending on your setup, it may be fairly easy for you to replace it yourself by configuring your task to voluntarily 'quit'.
In my case; I always have at least 2 tasks running per service. So I just added some logic to my tasks' healthchecks, so they report as unhealthy after ~6 hours. ECS will spot the 'unhealthy' task, remove it from the load balancer, and spin up a replacement (according to the binpack strategy).
Note: If you take this approach; add some variation to your timeout so you're less likely to have all of your tasks expire at the same time. Something like: expiry = now + timedelta(hours=random.uniform(5.5,6.5))
Sharing memory 'headspace' with soft-limits:
If you set both soft and hard memory limits; ECS will place your tasks based on the soft limit. If your tasks' memory usage varies with usage, it's fairly easy to get your EC2 instance to start swapping.
For example: Say you have a task defined with a soft limit of 900mb, and a hard limit of 1800mb. You spin up a service with 4 running instances. ECS provisions all 4 of these instances on a single t3.medium. Notice here that each instance thinks it can safely use up to 1800mb, when in fact there's very little free memory on the host server. When you hit your service with some traffic; each task tries to use some more memory, and your t3.medium is incapacitated as it starts swapping memory to disk. ECS does not recover from this type of failure very well. It notices that the task instances are no longer available, and will attempt to provision replacements, but the capacity provider is very slow to replace the swapping t3.medium.
My suggestion:
Configure your service to auto-scale based on memory usage (this will be a percentage of your soft-limit), for example: a target memory usage of 70%
Configure your tasks' healthchecks so that they report as unhealthy when they are nearing their soft-limit. This way, your tasks still have some headroom for quick spikes of memory usage, while giving your load balancer a chance to drain and gracefully replace tasks that are getting greedy. This is fairly easy to do by reading the value within /sys/fs/cgroup/memory/memory.usage_in_bytes.
I've now learnt that when I start a new EC2 instance it has a certain number of CPU credits due to which it's performance is high when it starts processing but gradually reduces over time as the credits run out. Past that point, the instance runs at which appears to be the baseline CPU utilisation rate. To numerate, when I started the EC2 instance (t2.nano), Cloudwatch reported around 80% CPU utilisation gradually decreasing down to 5%.
Now I'm happy to use one of the better instance types pending the instance limit request. But whilst that is in progress, I'd like to know whether the issue of reducing performance over time will still hold even with the better instance type?
Would I require a dedicated host setup if I wish to ensure I get consistent CPU utilisation? The only problem I can see here is that I'm running a SQS worker queue and Elastic Beanstalk allows us to easily setup a worker environment which reads messages from the queue. From what I've read and from looking at the configuration options available in Elastic Beanstalk, I don't think I'll be able to launch instances into a dedicated host directly. Most of my reading has lead me to believe that I'll have to learn how to use a VPC. Would that be correct?
So I guess my questions are - would simply increasing the instance type to a more powerful instance guarantee consistent CPU utilisation performance or is a dedicated host required and if so, is it possible to set up one with Elastic Beanstalk or would it have to be setup manually and if it is set up manually can it be configured to work with an SQS queue automatically?
If you want consistent CPU performance, you should avoid the burstable performance instances (the T2 family). All other families of instances (M5, C5, etc) will have consistent CPU performance over time. You can use any instance family with Elastic Beanstalk. No need for a dedicated host.
I'd like to have multiple different services running on an ECS cluster, each service should be running on a single EC2 instance. The EC2 instances type for all services are the same. And I would like those services to use all their hosting EC2 available resources.
I have the assumption that if i use only the soft memory parameter (without using the hard one ) in the Task Configuration, this will allow my container instance to use all the available memory on the EC2 instance hosting it and that i won't be limiting. Is that correct?
As for the EC2 type (t2.micro [vCPU=1, Memory=1Gib] for example) !! is it possible to simply put:
{
...
"memory": 1024,
"cpu": 1024,
...
}
Since the EC2 should be already set up with a bunch of Container Service Requirements.
Is it correct that you're trying to have each ECS Instance handle only a single task per instance?
The short answer to your question is, no. Usually the amount of memory made available to your containers is a bit less than the amount of memory available on the machine itself. This is so that the operating system has enough memory to keep running. From my experience, a T2.Small, which has 2048 MB of memory will end up with 2004 MB available for containers.
When it comes to your task definition, there are two ways of specifying Memory. The memory setting is a hard limit. If the containers memory usage hits this amount, the container will be terminated. If on the other hand, you specify memoryReservation, that much memory will be reserved for the task, but it can use more, up to the total amount of the machine. Check out the Task Definition documentation for further details.
An important consideration here is that only one of memory and memoryReservation are required. If both are used, memoryReservation should be less than memory. If you are only going to specify one of these, I'd recommend memoryReservation, as it will allow your task to use up to the total memory on the machine. If both are used, the memoryReservation will be used in calculating the amount of memory consumed by a task.
When placing tasks on an instance, it looks at the amount of available memory, that is the registered amount of memory for the instance, minus any tasks already placed on it. If this number is less than the amount of memory required for a task, no task will be placed on it. If no instance has enough memory for the task, it will not be placed, and the error will be logged in the Services Events log.
So it's important to look at the amount of memory actually registered by your instance type, and then ensure your memory or memoryReservation are lower than the amount registered by your instances. Otherwise, your tasks will never be placed.
As for cpu, this value is not required, and if not specified, all tasks on an instance are allowed an equal portion of the CPU available on the system. If only one task is on the instance, it can use the entire CPU of the instance by default.
When a Ec2 Container Engine cluster is created, it creates a Compute Engine managed instance group to manage the created instances. These instances are from Ec2 service, which means, they are Virtual machines.
But we know that containers represent a new way to deploy containers based on operating-system-level virtualization rather than hardware virtualization
like VMs that are heavyweight and non-portable, isn't a contradiction? correct me if I'm wrong.
We use containers because they are extremely fast (either in boot time or tasks execution) compared to VMs, and they save a lot of space storage. So if we have one node(vm) that can supports 4 containers max, our clients can rapidly lunch 4 containers, but beyond this number, Ec2 autoscaler will need to lunch a new node(vm) to support upcoming containers, which incurs some tasks delay.
Is it impossible to launch containers over physical machines?
And what do you recommend for running critical time execution tasks?
I believe you are working under an erroneous assumption that ECS scales the virtual machines ("container instances" -- the instances where containers will run) directly with task demand.
If that were true, you would have a point, because the cluster would be sluggish and unresponsive any time insufficient container instance resources were not immediately available.
ECS doesn't do that, the presence of the Auto Scaling Group notwithstanding.
Depending on the Amazon EC2 instance types you use in your clusters, and quantity of container instances you have in a cluster, your tasks have a limited amount of resources that they can use when they are run. ECS monitors the resources available in the cluster to work with the schedulers to place tasks. If your cluster runs low on any of these resources, such as memory, you will eventually be unable to launch more tasks until you add more container instances, reduce the number of desired tasks in a service, or stop some of the running tasks in your cluster to free up the constrained resource. (emphasis added)
http://docs.aws.amazon.com/AmazonECS/latest/developerguide/cloudwatch_alarm_autoscaling.html
So, no... it doesn't launch the new tasks slowly when you are out of capacity. It doesn't launch them at all.
But don't get ahead of me.
The link above explains, with examples, how scaling of the virtual machines (container instances) is designed to actually work.
Of course, you don't have to make them adaptively scalable at all. You can go with your physical server model (note: I say physical server model -- meaning a fixed, inelastic pool of resources, on always-running virtual machines, since virtual machines is what EC2 provides), and just choose how many instances you wait to have running at all times, essentially emulating physical servers. If you wanted, say, 8 container instances, the "auto scaling group" would maintain exactly 8 at all times, creating replacements if, say, one of them experienced a hardware failure. That "auto" accomplishment would be maintaining the status quo. And, of course, in this configuration, you could manually reconfigure from 8 to, say, 12 and the "auto" accomplishment would be that you'd automatically get 4 new ones to add to the existing 8.
But the idea of how the service is ideally used is that your group of virtual machines scales up and down by rules you devise, to anticipate the resources needed by future tasks -- or a future lack of tasks.
In the example given, memory reservation is the trigger:
When the memory reservation of your cluster rises above 75% (meaning that only 25% of the memory in your cluster is available to for new tasks to reserve), the alarm triggers the Auto Scaling group to add another instance and provide more resources for your tasks and services.
It triggers the addition of more container instances so that you always have whatever you have determined to be the appropriate threshold of surplus capacity already online by the time you need it.
Of course, memory is just one resource, and 75% is just an arbitrary threshold chosen for the example.
Auto Scaling Groups can scale on a variety of triggers -- the phrase of the moon, the price trends in the stock market, whatever is appropriate to anticipating your desired amount of surplus capacity and can be quantified and monitored can be used... but this service does not scale itself directly by the actual attempt to launch a new task when the task can't be launched due to insufficient resources.
Herein lies the flaw in your original argument.
Why virtual machines? Simply enough, because when you destroy a virtual machine because the capacity is not expected to be needed, you stop paying for it.
In this light, perhaps you'll agree that this is not a weakness, it's a strength. Physical servers never stop costing you when you are not using them.
You don't need to pay anything at all for capacity you will not be needing with VMs -- you only have to pay for the capacity you're using plus the amount you need to keep immediately available to handle anticipated demand.
You can have as much idle surplus immediately ready as you are willing to pay for, or you can maximize savings by allowing as little surplus capacity as you are comfortable with being able to access without delay.