on what to do
- Run the download in parallel
cat urls.txt | xargs -n 1 -P 4 curl -O- Run the installation of RPMs in parallel - does not work, rpm has lock file
cat rpms.txt | xargs -n 1 -P 4 sudo rpm -ivh-
Individual Installation:
time for rpm in /path/to/rpms/*.rpm; do sudo rpm -Uvh "$rpm"; done
-
Bulk Installation:
time sudo rpm -Uvh /path/to/rpms/*.rpm
- Use generator instead of custom shell scripts per bundle
Create the environment via generator and not via custom shell scripts
- Look into generator for all actions and update them accordingly
- Use something like dependabot to update all actions versions
version: 2
updates:
- package-ecosystem: 'github-actions'
directory: '/'
schedule:
interval: 'weekly'
commit-message:
prefix: 'chore'
open-pull-requests-limit: 10
ignore:
- dependency-name: '*'
versions: ['<breaking-change-version']We have more than 1200 repos, we can't update them manually
- Use dependabot to update all node.js dependencies
- Run npm bot and doctor
- Automerge all pull requests that are green
version: 2
updates:
- package-ecosystem: 'npm'
directory: '/'
schedule:
interval: 'weekly'
commit-message:
prefix: 'chore'
open-pull-requests-limit: 10
automerge:
- match:
- dependency-type: 'all'
automerge: true- FTPX and owncloud combination are very slow and cumbersome
- Move to S3 and use caboose to handle everything
- caboose is only available via github action. No one will get a user/api key
- S3 will be a void bucket with no public access. You can retrieve files but not change them
- Get rid of netlify (120+ dollars a month) and move to github pages
- Make release of documentation easier
- Use github actions to build and deploy the documentation
- Make use of textblocks for cross-referencing
- Maybe a monorepo?
- Make the repo public for a good apperance on the github profile of sealsystems
- safe money on azure vms and storage
- use the existing vSphere infrastructure (already paid for)
- use bigger and faster machines
- introduce long running tests (e.g. after EA stage)
- use branch releases for bundles
- build docker images in service branches if they are tagged with e.g
pi-team-1or something along those lines - use the branch name as a tag for the docker image
- only merge to master like it is currently. Branch releases are forbidden in release branches of a bundle
- get rid of third party bundles
- instead, install directly from vendor
- move custom parts (configuration/patches) into the product bundle itself
- add those custom parts to the generator
- maybe one sub generator per product (out-ngn / dpf / etc.)
"seal": {
"auxiliaries": {
"clients": {
"seal-island-checkout": "^1.10.0"
},
"apps": {
"mongodb": "^6.0.0",
"nats": "^2.10.0",
"keycloak": "^23.0.0",
"elasticsearch": "^7.17.5",
"easyprima": "^1.11.0".
}
},
"packaging": {
"dependencies": {
"seal-bodyscanner": "6.18.3",
"seal-bos2gyros": "1.2.3",
"seal-checkin": "8.24.1",
"seal-checkout": "8.21.2",
"seal-co-notifier": "4.0.1",
...- use semver for all product bundles, dont care for release planning versions
- this might need version mapping. so a technical version and a version'name' which is used by product management
- maybe like ubuntu/node.js naming e.g. iron / xenial / bionic / focal / etc.
- product management/planning need to work with placeholders like
nextorlatest
- new release needs to have updated generator
- new release needs to have an update path with it.
- breaking change might need custom scripts for that
- if generator run (install/update/performance test) fails, no release
- get rid of potpurri, streamline with generator
- current signing is on single vCloud VM (bottleneck)
- Create PLOSSYS OUTPUT ENGINE for icepanel
- Create all used 3rd party
- Create all possible deployment methods
- Create Build Pipeline as a System for show and tell where things need to be changed
Noch als Ergänzung so was wie api.sealsystems.de mit Beispielen. Alle Schnittstellen unter einer Doku zusammengefasst. Ähnlich wie: https://operator.docs.sealsystems.de/reference/api/use_cases/create_document