Commit Graph

3 Commits

Author SHA1 Message Date
Matt Mitchell dbb10e4beb Make deployment names more unique (to avoid more collisions)
Only create the vnet and subnets in deployment if they are not specifically specified (to avoid overwriting vnet settings)
2016-10-21 07:38:30 -07:00
Claudiu Guiman c8713852a1 Updated the ARM templates so the storage account name is unique per resource group 2016-10-18 12:42:43 -07:00
Matt Mitchell 6b53a4e3a3 Azure RM model updates
* Make the resource group configurable
* Update SDK to 0.9.4
* Fix: Token expiration was incorrectly calculated (API is named a bit oddly).
* Fix: Token expiration is in seconds.
* Change VM/deployment names to ones that are valid
* Fix location + VM sizes
  It appears that that the location and VM size APIs do not currently support the AAD based authentication.  The certificate is required.  Rather than re-introduce the cert for just this limited UI scenario, I have decided to hard-code the list based on the returned info from current Azure.  This is a decent short term solution, since a move to the 1.0.0 API (when released) will require that this code be changed anyway and presumably this problem should be fixed for good at that point.
* Proper handling for custom image URIs
* Asynchronous verification of the subscription info
* Asynchronous verification of the azure templates
* Asynchronous provisioning
* Clean up resources after unsuccessful provisioning
* Lots of logging updates
* Fix: SSH launcher - Get channels before connection for exec channels
  Getting the channels after connection introduces a race where we could potentially fail to read from the input streams if they were obtained from the channel after the connection had completed.
* Add diagnostics to image verification messages
* Asynchronous deletion via UI
* Update version of Jsch
* Temporarily disable image verification for reference images
* Allow for initialization scripts to be run as root
* Cleanup of slaves now works properly
    * Doesn't clean up when node is taken offline by user
    * Post build tasks won't kill other runs or show errors in the job logging
    * Retention strategy now for online nodes, cleanup for offline nodes
* Nodes that are marked to shut down on idle can now be restarted properly (before too many could be started)
* Add option to treat failures of the initialization script as a reason to discard the VM (linux only currently)
* Reenable Windows custom script extension for startup
* Update documentation with new sample startup scripts

Add setAcceptingTasks appropriately

doc fixup
2016-10-06 15:06:03 -07:00