SLAC18: PHP in the Container Cloud

License: CC BY-SA

Slides of my talk at the Secure Linux Administration Conference in Berlin at May 7th, 2018.

Transcript

1. PHP IN THE
   CONTAINER CLOUD
   MARTIN HELMICH
   MITTWALD CM SERVICE GmbH & Co. KG
   SLAC2018, BERLIN
   MAY 7th, 2018
   

   View Slide

2. This work is licensed under a Creative Commons Attribution-
   ShareAlike 4.0 International License.
   http://creativecommons.org/licenses/by-sa/4.0/
   

   View Slide

3. MARTIN HELMICH
   SOFTWARE & SYSTEMS ARCHITECT @ MITTWALD CM SERVICE
   LECTURER @ PHWT VECHTA
   http://stackoverflow.com/story/martinhelmich
   https://github.com/martin-helmich
   https://www.martin-helmich.de
   

   View Slide

4. View Slide

5. View Slide

6. View Slide

7. CC-BY-SA, Atomic Taco
   https://www.flickr.com/photos/atomictaco/25016033688
   

   View Slide

8. View Slide

9. catdoc
   pdftotext
   

   View Slide

10. catdoc
    pdftotext
    

    View Slide

11. WHY DOCKER?
    

    View Slide

12. WHY DOCKER?
    - EFFICIENCY
    - STANDARDIZED
    - STRONG COMMUNITY
    - DEVOPS SPIRIT &
    AUTOMATION
    

    View Slide

13. View Slide

14. View Slide

15. KUBERNETES
    

    View Slide

16. KUBERNETES
    - KUBER-WHAT!?
    - „GOOGLE SCALE“
    - DEVOPS‘ DREAM
    

    View Slide

17. View Slide

18. POD
    - 1(ISH) CONTAINER
    - SHORT-LIVED
    - DISPOSABLE
    KUBERNETES
    CORE CONCEPTS
    

    View Slide

19. apiVersion: v1
    kind: Pod
    metadata:
    name: some-pod
    spec:
    containers:
    - name: web
    image: nginx
    ports:
    - containerPort: 80
    

    View Slide

20. $ kubectl apply -f pod.yaml
    pod "some-pod" created
    $ kubectl get pods
    NAME READY STATUS
    some-pod 1/1 Running
    apiVersion: v1
    kind: Pod
    metadata:
    name: some-pod
    spec:
    containers:
    - name: web
    image: nginx
    ports:
    - containerPort: 80
    

    View Slide

21. KUBERNETES
    CORE CONCEPTS
    POD
    - 1(ISH) CONTAINER
    - SHORT-LIVED
    - DISPOSABLE
    REPLICA SET
    - MANAGES MULTIPLE
    PODS
    DEPLOYMENT
    - MANAGES MULTIPLE
    VERSIONS OF REPLICA SETS
    - ROLLING UPDATES/
    ROLLBACKS
    SERVICE
    - STABLE DNS/
    IP ADDRESS
    

    View Slide

22. SPACES
    CONTAINER
    ARCHITECTURE
    DATABASE PODS
    (STATEFUL SET)
    CACHING PODS
    (REDIS; STATEFUL SET)
    APPLICATION PODS
    (CUSTOM IMAGE;
    DEPLOYMENT)
    LOGGING POD
    (NXLOG;
    DEPLOYMENT)
    MANAGEMENT
    (SALTSTACK;
    DEPLOYMENT)
    NETWORK FILESYSTEM
    (PERSISTENT VOLUME)
    

    View Slide

23. $ kubectl scale deployment my-app --replicas=32
    deployment "my-app" scaled
    $ kubectl get deployments
    NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
    my-app 32 32 32 15 30m
    $ kubectl get pods
    NAME READY STATUS RESTARTS AGE
    my-app-2035384211-7ci7o 1/1 Running 0 18s
    my-app-2035384211-kzszj 1/1 Running 0 18s
    my-app-2035384211-qqcnn 1/1 Running 0 18s
    …
    

    View Slide

24. $ kubectl autoscale deployment my-app \
    --min=4 --max=128 --cpu-percent=80
    deployment "my-app" autoscaled
    

    View Slide

25. BENEFITS
    CUSTOMER + PROVIDER
    • EFFICIENT RESOURCE UTILIZATION
    (NO OVERPROVISIONING)
    • FAST SCALE-UP & SCALE-DOWN
    • „ZERO-CONF“
    

    View Slide

26. CHALLENGES
    

    View Slide

27. PHP ON
    SCALE
    SLAC 2017: „PHP ON SCALE“
    https://speakerdeck.com/martinhelmich/slac17-php-on-scale
    

    View Slide

28. SHARED
    FILESYSTEM
    /fileadmin
    /typo3conf
    /typo3temp
    /uploads
    SOLUTION #1:
    CONFIGURE APPLICATION TO USE
    OBJECT STORE (S3, CEPH, …)
    AND CENTRALIZED CACHES
    SOLUTION #2:
    USE NETWORK FILESYSTEM
    (NFS, GLUSTER, CEPHFS, …)
    

    View Slide

29. SCALING PHP APPS IS EASY… SCALING DATABASES IS NOT
    

    View Slide

30. SCALABLE
    CACHING LAYER
    (RELATIVELY) FIXED-SCALE
    DATABASE
    

    View Slide

31. DATABASE CLUSTER
    ORCHESTRATED BY VITESS
    VTGATE
    

    View Slide

32. SIZING &
    SCHEDULING
    

    View Slide

33. 8 CPUS
    64 GiB
    RAM
    KUBERNETES
    SCHEDULING
    THE NODE
    THE PODS
    REQUESTS
    LIMIT
    .5 CPU / 1 CPU
    128 MiB/
    256 MiB
    LIMITS = REQUESTS
    -> GUARANTEED POD
    LIMITS > REQUESTS
    -> OVERCOMITTMENT
    SCHEDULING
    POD REQUESTS MUST FIT NODE
    POD LIMITS MAY BE LARGER THAN
    NODE CAPACITY
    LIMITS > CAPACITY
    -> OK
    

    View Slide

34. KUBERNETES
    SETUP
    STORAGE
    ARCHITECTURE
    NETWORKING/
    INGRESS
    APPLICATION
    ARCHITECTURE
    APPLICATION
    LIFECYCLE
    THE
    KUBERNETES
    ICEBERG
    

    View Slide

35. QUESTIONS?
    

    View Slide

36. View Slide