Byzantine Reality

Searching for Byzantine failures in the world around us

AppScale 1.12.0 Released!

We’re happy to announce the release of AppScale 1.12.0! After three weeks of development and one week of QA, we’re happy to have fixed the following bugs and added the following major features:

  • Now using monit instead of god to monitor processes, and killing App Engine apps that take up too much memory
  • Load is rebalanced after scaling events
  • Upgraded Cassandra to 2.0.1
  • Upgraded pycassa to 1.9.1

Here’s a quick YouTube video that also summarizes this up:

In addition, we’re also happy to have fixed the following bugs and added the following features:

  • Upgraded GCE image to Ubuntu Precise
  • Upgraded KVM image to Ubuntu Precise
  • Upgraded VirtualBox image to Ubuntu Precise
  • Upgraded Eucalyptus image to Ubuntu Precise
  • Upgraded EC2 image to Ubuntu Precise
  • Now using monit to revive nginx if it fails
  • Now using monit to revive rabbitmq if it fails
  • Now using monit to revive ZooKeeper if it fails
  • Now using monit to revive haproxy if it fails
  • Now using monit to revive ejabberd if it fails
  • Now using monit to revive Cassandra if it fails
  • Improved Blobstore API fidelity for Python 2.7 apps
  • XMPP now works correctly on relocated apps
  • AppScale can now run in other AWS regions
  • No longer displaying oauth2client warnings on GCE
  • Not failing GCE image creation due to modprobe not being supported
  • Logging levels off by one in AppDashboard for Python apps
  • Fixed broken ImageLoadTest on Python 2.7 Hawkeye
  • Java AppServer now reads nginx port from filesystem, instead of as an arg
  • Protect access to flower
  • Nginx shouldn’t expose Blobstore port to users in their apps
  • Prevent deprecated Python 2.5 apps from being uploaded
  • Allowing users to set keys longer than 250 characters in Java Memcache
  • AppServer ports are now automatically reused
  • Nginx and haproxy ports are now automatically reused
  • If there is a version mismatch between tools and main, tools now report each version to user and what they can do to fix it
  • AppController now gets IP correctly on non-English locales
  • Users can now bootstrap AppScale VMs with a list of commands in their AppScalefile
  • kill_all_instances_for_app now works every time, instead of just the first time
  • Celery works on rebooted VMs in one node deployments
  • Cleaned up AppDashboard interface to AppController
  • Fixed uploading apps via the AppDashboard
  • Removed link to Admin Console in AppDashboard
  • Uploading apps via the Dashboard now shows their information there
  • AppDashboard now allows users to upload zip files
  • Remove app reliably clears apps off the Dashboard
  • InfrastructureManager no longer prints users’ AWS secret key
  • Removing apps then uploading them no longer results in 500s
  • Dashboard now shows updated ports for apps after relocation
  • Uploading apps via the Dashboard that were removed no longer results in 500s
  • Command-line interface now allows users to upload zipped apps

As usual, we have both the source and public AMI available. Here are some links:

AppScale source code, on GitHub

AppScale Tools source code, on GitHub

A public Amazon EC2 ami, ami-335b005a for the US East Region.

Links for KVM, VirtualBox, Eucalyptus, and Google Compute Engine images can be found on our Downloads page.

We’ll be doing three weeks of development and one week of QA for AppScale 1.13.0. We’ll send out an email once we get a Trello board ready for 1.13.0, which should be in the next 24 hours.

Of course, file any bugs you run into on our AppScale GitHub Issue Tracker and our AppScale Tools GitHub Issue Tracker so that we can prioritize them and address them. You can also find us on #appscale on freenode if you run into problems or the like.

And as always, thanks for using AppScale!