This project aims to help with taking care of (rather expensive) resources, for example several ephemeral virtual machines for the purposes of your CI.
Server side allows you to:
- automatically allocate resources
- periodically check that the resources are working properly
- and once the resource is not needed anymore, dispose it
The client side let's your users:
- request particular resource type
- wait till the resource is available
- release the resource
Such allocation of resource might be time consuming, so to not let your
users wait too much -- the server side is able to pre-allocate several
instances in advance. For more info, have a look at ./config/pools.yaml
configuration example.
Resalloc's concept of resources is intentionally vague and general. Whatever your allocation scripts spawn and return to resalloc, those are your resources. It can be virtual machines, docker containers, IP addresses, disk volumes, or in an extreme case even people (imagine a ticket system at the post office).
There already are allocation scripts for various kinds of resources:
- resalloc-aws - Allocates VMs in the Amazon EC2 cloud
- resalloc-ibm-cloud - Allocates VMs in the IBM Cloud
- resalloc-openstack - Allocates VMs in your OpenStack instance
- resalloc-kubernetes - Allocates pods in your Kubernetes cluster
If you can, please share your own.
At the first sight, it may look like resource allocation is a simple task and using something like resalloc is an overkill. But please consider what resalloc can do for you and what you would eventually need to implement on your own when starting from scratch.
- Multiple resource pools - Resalloc allows to define multiple pools of resources. They can either provide identical resources or different kinds of resources (e.g. x86_64 VMs from Amazon AWS and x86_64 VMs from OpenStack that can be used interchangeably, and ppc64le VMs from IBM Cloud that are used for a different purpose).
- Client-server architecture - Allocating many new resources in parallel can be surprisingly heavy on CPU and memory, so it can be useful to dedicate a separate machine to resource allocation.
- Gradual allocation - Allocating resources on self-hosted hardware is for free, so it makes sense to use it to the full capacity at all times. However, somebody will have to pay for every allocated resource in the cloud, so we want to allocate only as much resources as we need at the moment.
- Preallocation - Allocating new resources can take some time (e.g. spawning a new VM in the cloud and running Ansible playbooks to provision it can take few minutes), and users don't want to wait. It is a good idea to preallocate a small number of resources that are ready to be used immediately.
- On demand allocation - In special "on demand" pool mode, resources are not preallocated in advance but started on demand, only upon a ticket requesting the resources.
- Livechecks - Clouds are unreliable. VMs can break while starting or become unresponsive for various reasons. Resalloc periodically checks the liveness of all resources and makes sure money doesn't leak out of our pockets.
- Resource prioritization - Multiple pools can provide the same resources but they might not cost the same. There is a price for running VMs in Amazon AWS. The same VMs as Spot instances are a bit cheaper, and running them on a self-hosted sever is for free. Resalloc prioritizes to be wallet-friendly.
- Web interface - It is possible to optionally run a web UI and let users see what, and how many resources are available. For example, see Copr resources.
- Reusing/Security/Sandboxing - In certain cases, you want to give the user a full access to the allocated resource, even with the ability to break the resource (poison the VM so the subsequent users are affected, e.g.). Therefore Resalloc by default doesn't re-assign the same allocated resource to other users (other tickets), and simply deletes the resource. Since this might be waste of resources (starting/stopping resources takes users the time and thus money), as opt-in, resource might be placed into a "sandbox" (string determining the sandbox) and later taken by subsequent ticket, but only if the ticket is also in the same "sandbox".
- Hooks - Most of the configuration just tells the Resalloc server server what to do with the resource (to start, stop, release the resource, etc.). There are hook points to customize some actions (e.g. when VM is being shut-down => remove local caches, when machine is released => remove some temporary directory, ...).
Resalloc was created to accommodate the ever growing farm of Copr builders.
- Ask for a resource, and wait till it is ready
$ ticket=$(resalloc ticket --tag x86_64 --tag jenkins_vm)
$ output=$(resalloc ticket-wait $ticket)
- get the resource, and periodically check till it is available
$ ticket=$(resalloc ticket --tag x86_64 --tag jenkins_vm)
$ while ! resalloc ticket-check $ticket; do true; done
$ output=$(resalloc ticket-check $ticket)
Then, you can work with the resource:
$ ip=$output
$ ssh root@"$ip" -c "do something"
$ resalloc ticket-close "$ticket"
The $output variable will contain important info from the cmd_new
command run by resalloc server. If you request VMs, you typically want
cmd_new command which outputs an IP of the allocated virtual machine.
The released versions are installable from Fedora and Fedora EPEL repositories, just do
$ sudo dnf install -y resalloc # clients
$ sudo dnf install -y resalloc-server # server
Pre-release RPMs are available in testing Copr repositories: https://copr.fedorainfracloud.org/coprs/praiskup/resalloc/