Uploaded lttoolbox/apertium/cg3 to Debian experimental.
Started packaging Apertium-APY and language paris.
Fixing git review issues with operations/debs/ repo with Alex.
Santhosh pointed that we need more CPU/RAM than current Labs instance for Apertium service. Asked Alex to provide large instance at deployment-apertium01.eqiad.wmflabs
Current status in Beta cluster with Content Translation.
Why we need Apertium as service?
Questions from Alex about:
What will be situation with Google/Bing in future?
Google/Bing will have agreement on API uptime and reliability with us if we opt for such service in future. This will make sure that we're not affected.
How frequent Apertium service will be updated?
We will have mostly update on Apertium language pairs once in a month, which are data packages. Core Apertium packages will be only updated upon critical bugs found in upstream bug tracker. Any bug from Wikimedia dev/user will be track at Apertium upstream bug tracker.
Blockers from LE team:
Apertium need to run as service before we go ahead with deployment.
Blockers from Ops:
Ops running tight with server allocation.
Action items:
Kartik will start working with Alex for setting up Debian repositories in WMF.
Kartik will start work on packages and pushing them to Debian repositories.
Alex will provide machine in Beta cluster when Packaging is done (in a week).
Kartik/Niklas will start working on Apertium Puppetizing with Alex.
Alex will talk to Mark about production server estimate date.
Alolita to setup meeting with Mark/Ops.
Kartik will document everything that comes under the hood of Apertium/Puppet/Repository and CX deployment process!
ETAs:
Beta cluster machine for Apertium service: 1 week after Packaging.
Production server: 3 to 4 weeks.
Update Deployment plan document with daily updates.