a

New Deployment Model

for

Servoy

Paul Bakker - The Orange Dots - ServoyWorld '17

High Availability
Static & versioned

Secure

No bloat

Automated

(fail-over, deployments, scaling)

.servoy export/import


WAR deployment

 

Servoy Cluster

/ single db

/ single db server

/ multi db servers

single app server

/ single db

/ single db server

/ multi db servers

multi app servers

stand-alone app server WAR stand-alone container managed containers
HA - automatic fail-over nope nope nope yup
HA - zero downtime deployment .servoy upload: possible Tomcat Parallel Deployment? Tomcat Parallel Deployment? yup
HA - scaling nope nope nope yup
static & versioned not enforced to some extend pretty much pretty much
Secure ~ ~ ~ ~
No Bloat ~ ~ ~ ~
Automated can be done can be done can be done a lot already done

Every change === new server launch (and shutting down old versions)

⚫ ⚫ ⚫ ⚫

Different solution

&

Servoy versions in parallel

⚫ ⚫ ⚫ ⚫

Managed Containers

amqpbroadcaster.hostname=localhost

servoy.properties

https://github.com/Servoy/servoy-extensions/tree/v82x_branch/com.servoy.extensions/src/com/servoy/extensions/plugins/broadcaster

GitHub

Generic database stuff

no Servoy Sequences
servoy.record.lock.lockInDB=true
databaseManager.acquireLock(...) ONLY            within transactions
databaseManager.acquireLock(...) WON'T          return until lock acquired (unless...)
no 'select * from ....'
no breaking datamodel changes

Safe operations Comments
Add column
Drop column assuming the app isn't using it anymore
Add an index concurrently
Drop a constraint (for example, non-nullable)
Add a default value to an existing column
Operations to avoid Workarounds
Add index Add index concurrently
Change column type Add new column, change code to write to both, backfill new column
Add column with default value Add column, add default in separate step, backfill column with default value
Add non-nullable column
Add column with unique constraint Add column, add unique index concurrently, add constraint 
VACUUM FULL use pgrepack instead

sources: safe-operations-for-high-volume-postgresql, zero-downtime-postgres-migrations-the-hard-parts

Overall: no long-running transactions/locks and queries

Operations to think about
Adding foreign key constraints Requires a AccessExclusive lock on both tables, potentially blocking read access

No,
not this guys this time....

The other one -->>

Sticky sessions
Session draining
Controlled shutdown

API's interact only with current server


application.getActiveClientCount
application.getClientCountForInfo
plugins.headless.getClient/getOrCreateClient
maintenance plugin
usermanager plugin
clientmanager plugin

Misc

 

No pre/postImport hooks

Cleanup

      

HSQL as Servoy Repository
Embedded Servlet Engine
Java 9 Modules (?!?)

  • plugins, beans, drivers, lafs, services, component

  • servoy.properties

  • dependancies for unused Servoy Client types

Title Text

Level I:

  • Cluster
  • Nodes
  • Pods

Level II:

  • Ingress
  • Services
  • Deployments
  • ReplicaSets
  • Jobs
selector:
  matchLabels:
    run: paragon
  matchExpressions:
    - {key: exclude, operator: NotIn, values: [true]}

label selectors: Deployment / ReplicaSet / Service

labels
    run: paragon

labels: Pods

Your database
RabbitMQ

....

Servoy

  • Separate read/write database connections
  • Table aliases
  • DB connection credentials hook

Community

  • (Postgres) datamodel migration  best practices
  • Feedback/input/....

Using

Docker

to deploy your

Servoy application

Made with Slides.com