Slurm: the caterpillar turned into a butterfly
<Tl; dr>
- Slurm really allows you to enter the topic Kubernetes or pull up your knowledge.
- Participants are satisfied. Those who have not learned anything new or have not solved their problems are just a few. The unconditional Manibak of the first day (“If you feel that Slurm is not suitable for you, we will refund the full price of the ticket”) only one person used, justifying that he overestimated his strength.
- The next Slurm will be held in St. Petersburg in early September. Selectel, our permanent sponsor, provides not only a cloud for stands, but also a conference room.
- We repeat the basic Slurm (September 9-11) and present a new program: Slurm DevOps (September 4-6).
What is Slurm and how has it changed
A year ago, we had the idea to conduct courses on Kubernetes. In August '18 Slurm-1 passed: difficult, with continuous presintation (when the presentation was finished on the stage), with a bunch of everyday problems. Trials are rallied: the participants of the first Slurm, like the Fellowship of the Ring, still communicate with each other.
It looked Slurm-1
At the first Slurme, the idea was born to hold Megaslum. We asked people what topics they were interested in, and in October conducted an advanced course “At the request of the participants”. It turned out an interesting, but one-time event. By May'19 we prepared a real advanced course, with our own logic and inner history.
For the year Slurm and organizationally transformed:
- Docker and Anisble removed from the main program and made separate online courses.
- Organized technical support that helps students troubleshoot study clusters.
- The speakers have methodical support.
The team that made Slurm-4
Feedback from participants
Another record was taken: 170 participants at the base Slurm, 75 - at MegaSlreme.
Slurm-4
The feedback form was filled out by 101 out of 170 people.
Is Kubernetes clear?
41 - I don't understand k8s yet, but I see where to dig.
36 - Previously, k8s did not know, now figured out.
23 - I knew k8s before, now I know better.
1 - I learned nothing new.
0 - I did not understand anything about k8s.
How do you like the intensity of Slurm?
16 people believe that Slurm is too simple and slow, and 14 - that it is too difficult and fast. The rest is just right.
Did you solve the problem from which you went to Slurm?
90 - Yes.
11 - No.
MegaSlurm
40 people filled out the feedback form. 2 people said it was too easy and slow. 1 person did not solve the problem with which he was going to Megu. The rest is OK.
Feedback on Slurme at https://serveradmin.ru
Speaker Reviews
If at the St. Petersburg Slurme in February there were mostly newbies, then at Moscow Slurme people in the mass have already tried in Kuburnethes. There were a lot of advanced questions that made you think.
If in St. Petersburg they asked when we lay out our kubespray fork, then in Moscow they already asked why we suggest using our fork, rather than taking the original kubespray. This is the critical thinking of the seniors.
The practice was difficult, people made a lot of mistakes, and this is great: you need to fill the bumps while studying, and not in battle.
We regularly encountered certificates limits, download limits from Github, etc. This is life - we simultaneously deployed about 200 clusters in the Selectel cloud. No one is preparing their resources and limits for this.
Announcement of Slurm in Selectel
→ Registration for Slurm-5
Price: 25 000 â‚˝
Topic 1: Introduction to Kubernetes, main components
- Introduction to k8s technology. Description, application, concepts
- Pod, ReplicaSet, Deployment, Service, Ingress, PV, PVC, ConfigMap, Secret
Topic # 2: Cluster device, main components, fault tolerance, k8s network
- Cluster device, main components, fault tolerance
- k8s network
Theme number 3: Kubespray, tuning and tuning cluster Kubernetes
- Kubespray, tuning and tuning cluster Kubernetes
Topic 4: Advanced Kubernetes Abstractions
- DaemonSet, StatefulSet, RBAC, Job, CronJob, Pod Scheduling, InitContainer
Topic 5: Publication of services and applications
- Review of methods of publishing services: NodePort vs LoadBalancer vs Ingress
- Ingress controller (Nginx): we balance incoming traffic
- ert-manager: automatically get SSL / TLS certificates
Topic 6: Introduction to Helm
Topic 7: Installing cert-manager
Subject number 8: Ceph: installation in the "do as I do" mode
Topic 9: Logging and Monitoring
- Cluster Monitoring, Prometheus
- Cluster logging, Fluentd / Elastic / Kibana
Topic 10: Cluster Update
Topic 11: Practical work, application dockerization and launch into the cluster
Docker and Ansible courses on stepik.org are included in the price.
→ Registration for SlOrm DevOps
Price: 45 000 â‚˝
Topic 1: Introduction to Git
- Basic commands git init, commit, add, diff, log, status, pull, push
- Setting up a local environment: practical recommendations
- Git flow, branches and tags, merge strategies
- Work with multiple remote repo
Topic 2: Teamwork with Git
- GitHub flow
- Fork, remoute, pull request
- Conflicts, releases, once again about Gitflow and other flow applied to commands
Topic 3: CI / CD Automation Introduction
- Automation in gita (bots, introduction to CI, hooks)
- Tools (bash, make, gradle)
- Factory assembly lines for assembly and their application in IT
Topic 4: CI / CD: Working with Gitlab
- Build, test, deploy
- Stages, variables, execution control (only, when, include)
Theme number 5: Working with the application in terms of development
- We write microservice in Python (including tests)
- Docker-compose application development
Topic 6: Infrastructure as Code
- IaC: approach to infrastructure as a code
- IaC on the example of Terraform
- IaC on the example of Ansible
- Idempotency, declarative
- Practice creating ansible playbooks
- Configuration storage, collaboration, application automation
Topic 7: Infrastructure Testing
- Testing and continuous integration with Molecule and Gitlab CI
Theme number 8: Automation lifting servers
- We collect images
- PXE and DHCP
Topic 9: Infrastructure Automation
- An example of an infrastructure service for authorization on servers
- ChatOps (integration of instant messengers with pipelines)
Topic 10: Security Automation
- Signature artifacts CI / CD
- Vulnerability Scanning
Topic 11: Monitoring
- Definition of SLA, SLO, Error Budget and other terrible terms from the world of SRE
- SRE: The practice of monitoring SLI and SLO
- SRE: Practicing Error Budget
- SRE: Interrupt and Operational Load Management (apigateway, service mesh, circuit brackers)
- Monitoring Pipelines and Development Metrics
')
Source: https://habr.com/ru/post/455972/
All Articles