TensorHive 1.1.0

What's Changed

Main changes:

• Extended resource protection mechanism (#347)
  • configurable protection levels, with new mode (level 2) which does not allow using GPUs without reservations, even if there are no other reservations violated
  • violations are detected for all available resources, not only for current reservations
  • configurable process killing levels (0 - no killing, 1 - try to kill as the process owner, 2 - kill using sudo)
  • gather multiple violated resources in one message in both mailer and message sending handlers
  • send violation messages to user TTYs across multiple hosts
• Reverse proxy support in (#329)
• Persisting selected resources in calendar view (#353)

Minor changes:

• dependency version upgrades
• bugfixes

Configuration Changes

Standard upgrade process using pip or git pull will work. However, the following changes in main_config.ini should be taken into account:

• enabled option under [protection_service] has been deleted and is no longer relevant
• level option has been introduced under [protection_service] see description
• kill_processes option has been introduced under [protection_service] see description
• url_schema and url_port options have been introduced under [api] see description

New Contributors

• @tboinski made their first contribution in #329

Full Changelog: 1.0.0...1.1.0

TensorHive 1.0.0

What's Changed

Main improvements:

• expansion of the Task Execution module to the new Job execution module:
  • tasks can grouped into jobs and managed jointly
  • tasks can be cloned
  • tasks can be edited considering the individual command segments
• job queue for non reserved time - automatic scheduler can run jobs when the required resources are not reserved

Minor improvements:

• e-mail queuing and multiple admin addresses
• dependency fixes
• bugfixes

Full Changelog: 0.3.6...1.0.0

TensorHive 0.3.6

What's Changed

• Hotfix/r0.3.6 (#320)

Full Changelog: 0.3.5...0.3.6

TensorHive 0.3.5

What's Changed

Main improvements:

• access control - users and groups can be granted privileges to reserve, monitor and utilize specific GPUs
• security improvements
• functional API tests, increased test coverage

Minor improvements:

• redesigned configuration and DB initialization CLI
• automatic DB migrations
• OpenAPI upgrade to 3.0.3
• upgraded webapp dependencies
• removed unnecessary versioning from API URL
• improved tail mode in tasks overview
• several bugfixes

New Contributors

• @jszemplinski made their first contribution in #243
• @matpiotrowski made their first contribution in #247
• @Acrobot made their first contribution in #260
• @martyole made their first contribution in #307

Full Changelog: 0.3.3...0.3.5

TensorHive 0.3.3

Full Changelog: 0.3.2...0.3.3

TensorHive 0.3.2

Full Changelog: 0.3.1...0.3.2

TensorHive 0.3.1

What's Changed

• UX improvements in webapp (redesigned reservation schedule component, calendar refreshing, highlighting of current users reservations, saving watches in local storage)
• commandline UX improvements
• showing GPU IDs in violation messages
• using local time in reservation dates
• CLI commands for connection testing and displaying TH public SSH key
• handling missing SSH keys
• updated configuration guides
• allowing custom API URLs
• vulnerable dependency version bumps
• improved README with new screenshots
• various bugfixes

Full Changelog: 0.3...0.3.1

TensorHive 0.3

What's Changed

The 0.3 version introduces task scheduling module, which allows for:

• easy defining computational tasks in the web application, using templates with variable placeholders
• automatic spawning tasks on the assigned computational devices
• viewing task logs in the web application
• easy terminating and killing a task or group of tasks
• attaching terminal to a running task using screen
• scheduling tasks to a given time slot
• linking tasks to a specific reservation

Full Changelog: 0.2.4...0.3

TensorHive 0.2.4

What's Changed

• added user registration through adding generated RSA key to authorized_keys
• added separate intruder/admin timer in mailbot, improved mailbot logic
• fixed bug in reservation statistics which resulted in no statistics shown and not hiding statistic files
• fixed lack of calendar refreshing after reservation cancelling
• fixed refresh token handling in webapp
• UX improvements: button sizes, datetime string conversion, default values to inputs, calendar refreshing, confirmation button in reservation datetime picker, add version information
• added mypy code checks and flake8, also for test code

Full Changelog: 0.2.3...0.2.4

TensorHive 0.2.3

What's Changed

• mailbot handler for reservation protection service
• basic reservation usage statistics
• vertical calendar view in webapp
• reservation and user edition dialogs in webapp
• bugfixes and minor improvements

Full Changelog: 0.2.2...0.2.3