There is a couple of confusing points in the documentation that make me struggle to understand how exactly distribution across the cluster happens in Orleans. Hence, the questions.
Question #1
Orleans claims to have a built-in distribution capabilities to distribute across multiple servers. To me it sounds that Orleans can act as a load balancer itself and can scale out automatically. Thus, if I deploy Orleans app to several servers, then service discovery and load management should happen automatically, correct?
In this case, why some docs and articles suggest using other tools, like Ocelot or Consul, as a single entry point to Orleans cluster?
Question #2
I would like to use simple but distributed in-memory storage across several servers, like Redis or Apache Ignite, and I would like to know if it's possible to use a simple grain as this kind of a data storage?
Let's say, one grain will store a collection of restaurants and some other grain will keep track of the last 1000 visitors for selected restaurant. Can I activate these 2 grains only once as a singleton collection, add or remove records to each collection, and use these 2 grains as in-memory storage evenly available to all nodes in the cluster? Also, if answer is yes, do I need to add locks to these collections or each grain always exists in a single thread?
Service discovery and load management happen automatically indeed.
Consul is not a strong required. The only external requirement is a Membership table provider - something that is used internally by Orleans Clustering. There are many build in Membership table providers that come already built-in with Orleans. For example, Azure table storage. all you need is to configure Orleans to use it and of course have Azure storage account. Consul is another alternative to Membership table provider and there are more.
Another thing that does not come built-in is infrastructure scaling. If your service demand increases, something need to ask the infrastructure provider (Cloud Provider) to add more Servers. Once servers are added, Orleans will automatically adjust the workload and load balance across the new servers as well. But figuring out that more servers are needed and adding them is not done by Orleans itself (there likely some externally contributed tools to do that. maybe K8 can be configured to do that? I am not completely sure about that).
Yes, you can use those 2 grains as in-memory storage, just like you wrote. And no, you do not need to use locks. All grains are single threaded.
Is it possible to have a centralized storage/volume that can be shared between two pods/instances of an application that exist in different worker nodes in Kubernetes?
So to explain my case:
I have a Kubernetes cluster with 2 worker nodes. In each one of these I have 1 instance of app X running. This means I have 2 instances of app X running totally at the same time.
Both instances subscribe on the topic topicX, that has 2 partitions, and are part of a consumer group in Apache Kafka called groupX.
As I understand it the message load will be split among the partitions, but also among the consumers in the consumer group. So far so good, right?
So to my problem:
In my whole solution I have a hierarchy division with the unique constraint by country and ID. Each combination of country and ID has a pickle model (python Machine Learning Model), which is stored in a directory accessed by the application. For each combination of a country and ID I receive one message per minute.
At the moment I have 2 countries, so to be able to scale properly I wanted to split the load between two instances of app X, each one handling its own country.
The problem is that with Kafka the messages can be balanced between the different instances, and to access the pickle-files in each instance without know what country the message belongs to, I have to store the pickle-files in both instances.
Is there a way to solve this? I would rather keep the setup as simple as possible so it is easy to scale and add a third, fourth and fifth country later.
Keep in mind that this is an overly simplified way of explaining the problem. The number of instances is much higher in reality etc.
Yes. It's possible if you look at this table any PV (Physical Volume) that supports ReadWriteMany will help you accomplish having the same data store for your Kafka workers. So in summary these:
AzureFile
CephFS
Glusterfs
Quobyte
NFS
VsphereVolume - (works when pods are collocated)
PortworxVolume
In my opinion, NFS is the easiest to implement. Note that Azurefile, Quobyte, and Portworx are paid solutions.
Our cloud application consists of 3 tightly coupled Docker containers, Nginx, Web and Mongo. Currently we run these containers on a single machine. However as our users are increasing we are looking for a solution to scale. Using Kubernetes we would form a multi container pod. If we are to replicate we need to replicate all 3 containers as a unit. Our cloud application is consumed by mobile app users. Our app can only handle approx 30000 users per Worker node and we intend to place a single pod on a single worker node. Once a mobile device is connected to worker node it must continue to only use that machine ( unique IP address )
We plan on using Kubernetes to manage the containers. Load balancing doesn't work for our use case as a mobile device needs to be tied to a single machine once assigned and each Pod works independently with its own persistent volume. However we need a way of spinning up new Pods on worker nodes if the number of users goes over 30000 and so on.
The idea is we have some sort of custom scheduler which assigns a mobile device a Worker Node ( domain/ IPaddress) depending on the number of users on that node.
Is Kubernetes a good fit for this design and how could we implement a custom pod scale algorithm.
Thanks
Piggy-Backing on the answer of Jonah Benton:
While this is technically possible - your problem is not with Kubernetes it's with your Application! Let me point you the problem:
Our cloud application consists of 3 tightly coupled Docker containers, Nginx, Web, and Mongo.
Here is your first problem: Is you can only deploy these three containers together and not independently - you cannot scale one or the other!
While MongoDB can be scaled to insane loads - if it's bundled with your web server and web application it won't be able to...
So the first step for you is to break up these three components so they can be managed independently of each other. Next:
Currently we run these containers on a single machine.
While not strictly a problem - I have serious doubt's what it would mean to scale your application and what the challenges that come with scalability!
Once a mobile device is connected to worker node it must continue to only use that machine ( unique IP address )
Now, this IS a problem. You're looking to run an application on Kubernetes but I do not think you understand the consequences of doing that: Kubernetes orchestrates your resources. This means it will move pods (by killing and recreating) between nodes (and if necessary to the same node). It does this fully autonomous (which is awesome and gives you a good night sleep) If you're relying on clients sticking to a single nodes IP, you're going to get up in the middle of the night because Kubernetes tried to correct for a node failure and moved your pod which is now gone and your users can't connect anymore. You need to leverage the load-balancing features (services) in Kubernetes. Only they are able to handle the dynamic changes that happen in Kubernetes clusters.
Using Kubernetes we would form a multi container pod.
And we have another winner - No! You're trying to treat Kubernetes as if it were your on-premise infrastructure! If you keep doing so you're going to fail and curse Kubernetes in the process!
Now that I told you some of the things you're thinking wrong - what a person would I be if I did not offer some advice on how to make this work:
In Kubernetes your three applications should not run in one pod! They should run in separate pods:
your webservers work should be done by Ingress and since you're already familiar with nginx, this is probably the ingress you are looking for!
Your web application should be a simple Deployment and be exposed to ingress through a Service
your database should be a separate deployment which you can either do manually through a statefullset or (more advanced) through an operator and also exposed to the web application trough a Service
Feel free to ask if you have any more questions!
Building a custom scheduler and running multiple schedulers at the same time is supported:
https://kubernetes.io/docs/tasks/administer-cluster/configure-multiple-schedulers/
That said, to the question of whether kubernetes is a good fit for this design- my answer is: not really.
K8s can be difficult to operate, with the payoff being the level of automation and resiliency that it provides out of the box for whole classes of workloads.
This workload is not one of those. In order to gain any benefit you would have to write a scheduler to handle the edge failure and error cases this application has (what happens when you lose a node for a short period of time...) in a way that makes sense for k8s. And you would have to come up to speed with normal k8s operations.
With the information provided, hard pressed to see why one would use k8s for this workload over just running docker on some VMs and scripting some of the automation.
I would like to request some expert opinion from SMEs on:
What would it take to build a CAAS using physical/bare-metal servers.
a) Suggested options to implement concerns like : Resource Monitoring, Utilization etc
b) Any suggested approaches to make a robust, scalable, resilient underlying infrastructure which marries well kubernetes/docker/etcd/flannel based CAAS layer
Kubernetes as a solution likes to leverage underlying "Cloud Provider" ; any suggestion to create it from scratch? (of course I am not looking at elaborate solutions like OpenStack, since the idea is to create a lean/low-cost/light-weight solution) The solution I am looking for is to create an IAAS out of lets say 10 prod servers. Not looking for Data Center type of solutions.
For your second question, you should check out the Getting started from Scratch guide in the Kubernetes docs.
i have implemented similar setup on baremetal.
you dont need etcd on each machine
setup 3 node etcd cluster for high availability
if you have space machines then keep etcd cluster separated from controle plane
setup 3 node control plane
use nginx for load balancing the masters/control plane
Run keepalived service on each control plane.
The virtual ip should be routable in the across the nodes in cluster. chose VIP from same subnet
keepalived provide failover in case primary control plance/master is down
I'm trying to implement a cluster using Erlang as the glue that holds it all together. I like the idea that it creates a fully connected graph of nodes, but upon reading different articles online, it seems as though this doesn't scale well (having a max of 50 - 100 nodes). Did the developers of OTP impose this limitation on purpose? I do know that you can setup nodes to have explicit connections only as well as have hidden nodes, etc. But, it seems as though the default out-of-the-box setup isn't very scalable.
So to the questions:
If you had 5 nodes (A, B, C, D, E) that all had explicit connections such that A-B-C-D-E. Does Erlang/OTP allow A to talk directly to E or does A have to pass messages from B through D to get to E, and thus that's the reason for the fully connected graph? Again, it makes sense but it doesn't scale well from what I've seen.
If one was to try and go for a scalable and fault-tolerant system, what are your options? It seems as though, if you can't create a fully connected graph because you have too many nodes, the next best thing would be to create a tree of some kind. But, this doesn't seem very fault-tolerant because if the root or any parent of children nodes dies, you would lose a significant portion of your cluster.
In looking into supervisors and workers, all of the examples I've seen apply this to processes on a single node. Could it be applied to a cluster of nodes to help implement fault-tolerance?
Can nodes be part of several clusters?
Thanks for your help, if there is a semi-recent website or blogpost (roughly 1-year old) that I've missed, I'd be happy to look at those. But, I've scoured the internet pretty well.
Yes, you can send messages to a process on any remote node in a cluster, for example, by using its process identifier (pid). This is called location transparency. And yes, it scales well (see Riak, CouchDB, RabbitMQ, etc).
Note that one node can run hundred thousands of processes. Erlang has proven to be very scalable and was built for fault tolerance. There are other approaches to build bigger, e.g. SOA approach of CloudI (see comments). You also could build clusters that use hidden nodes if you really really need to.
At the node level you would take a different approach, for example, build identical nodes that are easy to replace if they fail and the work is taken over by the remaining nodes. Check out how Riak handles this (look into riak_core and check the blog post Introducing Riak Core).
Nodes can leave and enter a cluster but cannot be part of multiple clusters at the same time. Connected nodes share one cluster cookie which is used to identify connected nodes. You can set the cookie while the VM is running (see Distributed Erlang).
Read http://learnyousomeerlang.com/ for greater good.
The distribution protocol is about providing robustness, not scalability. What you want to do is to group your cluster into smaller areas and then use connections, which are not distribution in Erlang but in, say, TCP sessions. You could run 5 groups of 10 machines each. This means the 10 machines have seamless Pid distribution: you can call a pid on another machine. But distributing to another group means you can't seamlessly address the group like that.
You generally want some kind of "route reflection" as in BGP.
1) I think you need a direct connection between nodes to communicate between processes. This does, however, mean that you don't need persistent connections between all the nodes if two will never communicate (say if they're only workers, not coordinators).
2) You can create a not-fully-connected graph of erlang nodes. The documentation is hard to find, and comes with problems - you disable the global system which handles global names in the cluster, so you have to do everything by locally registered names, or locally registered names on remote nodes. Or just use Pids, as they work too. To start an erlang node like this, use erl ... -connect_all false .... I hope you know what you're up to, as I couldn't trust myself to do that.
It also turns out that a not-fully-connected graph of erlang nodes is a current research topic. The RELEASE Project is currently working on exactly that, and have come up with a concept of S-groups, which are essentially fully-connected groups. However, nodes can be members of more than one S-group and nodes in separate s-groups don't have to be fully connected but can establish the connections they need on demand to do direct node-to-node communication. It's worth finding presentations of theirs because the research is really interesting.
Another thing worth pointing out is that several people have found that you can get up to 150-200 nodes in a fully-connected cluster. Do you really have a use-case for more nodes than that? Surely 150-200 incredibly beefy computers would do most things you could throw at them, unless you have a ridiculous project to do.
3) While you can't start processes on a different node using gen_server:start_link/3,4, you can certainly call servers on a foreign node very easily. It seems that they've overlooked being able to start servers on foreign nodes, but there's probably good reason for it - such as a ridiculous number of error cases.
4) Try looking at hidden nodes, and at having a not-fully-connected cluster. They should allow you to group nodes as you see fit.
TL;DR: Scaling is hard, let's go shopping.
There are some good answers already, so I'm trying to be simple.
1) No, if A and E are not connected directly, A cannot talk to E. The distribution protocol runs on direct TCP connection - no routing included.
2) I think a tree structure is good enough - trade-offs always exist.
3) There's no 'supervisor for nodes', but erlang:monitor_node is your friend.
4) Yes. A node can talk to nodes from different 'clusters'. In the local node, use erlang:set_cookie(OtherNode, OtherCookie) to access a remote node with a different cookie.
1)
yes. they talk to each other
2) 3) and 4)
Generally speaking, when building a scalable and fault tolerant system, you would want, or more over, need to divide the work load to different "regions" or "clusters". Supervisor/Worker model has this envisioned thus the topology. What you need is a few processes coordinating work between clusters and all workers within one single cluster will talk to each other to balance out within group.
As you can see, with this topology, the "limitation" is not really a limitation as long as you divide your tasks carefully and in a balanced fashion. Personally, I believe a tree like structure for supervisor processes is not avoidable in large scale systems, and this is the practice I'm following. Reasons are vary but boils down to scalability, fault tolerance as fall back policy implementation, maintenance need and portability of the clusters.
So in conclusion,
2) use a tree-like topology for your supervisors. let workers explicitly connect to each other and talk within their own domain with the supervisors.
3) while this is the native designed environment, as I presume, I'm pretty sure a supervisor can talk to a worker on a different machine. I would not suggest this as fault tolerance can be hell in remote worker scenario.
4) you should never let a node be part of two different cluster at the same moment. You can switch it from one cluster to another though.