pyinfra-dev/pyinfra v0.7.1

pyinfra v0.7

on GitHub

This is a huge step forward for pyinfra; as well as a bunch of key new modules and a new progress bar that highlights what's pending, pyinfra now compiles deploy code so if statements work as expected (no need to use the with state.limit workaround).

• Getting started
• Documentation

Changes in 0.7:

• Add mysql module
  • Operations: mysql.sql, mysql.user, mysql.database, mysql.privileges, mysql.dump, mysql.load
  • Facts: mysql_databases, mysql_users, mysql_user_grants
• Add postgresql module
  • Operations: postgresql.sql, postgresql.role, postgresql.database, postgresql.dump, postgresql.load
  • Facts: postgresql_databases, postgresql_roles
• Add puppet module with puppet.agent operation (@tobald)
• Add server.crontab, server.modprobe and server.hostname operations
• Add git.config operation
• Add kernel_modules, crontab and git_config facts
• Add global install virtualenv support (like iPython)
• Massively improved progress bar which highlights remaining hosts and tracks progress per operation or fact
• Improved SSH config parsing, including proxyjump support (@tobald)
• Support for CONFIG variables defined in local.include files
• Fix command fact now outputs everything not just the first line

Internal changes:

• Replace --debug-state with --debug-operations and --debug-facts
• pyinfra now compiles the top-level scope of deploy code, meaning if statements no longer generate imbalanced operations
  • This means the recommendations to use state.when in place of conditional statements is invalid
  • Updated the warning shown, now once, with a link
  • Included a test deploy_branches.py which can be used to verify operations do run in order for each host when compile is disabled
  • Compiling can be disabled by setting PYINFRA_COMPILE=off environment variable
• Deprecate state.limit and replace with state.hosts(hosts) (consistency with global operation kwarg hosts not limit)
• Major internal refactor of AttrData handling to reduce operation branching:
  • Generate AttrData on creation, rather than read
  • Add nesting support for AttrData so host.data.thing['X'] will not create branching operations
  • Turn fact data into AttrData
  • Make host.name an AttrDataStr
  • Hash True, False and None constants as the same so they can change between hosts without branching operations
  • Update docs and warning on operation branching
• Better default for pool parallel size
• Show stdout if stderr is empty on command failure (surprisingly common)