This repository has been archived by the owner on Jul 6, 2022. It is now read-only.
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Provision and Deprovision enhancements to use new async features (#204)
* Enhancement of asynch engine This PR enhances the asynch engine to enable the concenpt of delayed or paused tasks. Tasks can be added to the engine in a queue separate from the main work queue. As delayed tasks are added, a resumer is notified and will periodically start the tasks up. Delayed tasks can also be started by engine clients. Using this, pkg/api/provision and deprovision were enhanced to determine if a provision or deprovision action needs to wait for another operation to finish. If so, they submit delayed tasks, either wait_for_parent or wait_for_children, whichever is appropriate. Those new jobs will make the same determination, and either readd themselves as delayed tasks, or initiate the task that would otherwise originally have been created. * Added bad parent response message * Merge master into feature branch (#209) * refactor event function signatures (#147) * pass around instances and bindings instead of pointers * Refactor instance encryption/decryption Previously, encryption and decryption of certain instance fields was accomplished through special getters and setters. To use these, you had to pass in a codec. This meant that any bit of code needing to access these fields needs access to the codec. Since I'm entertaining changes to the service.Module interface that will result in modules having access to instance objects, I wanted a way for modules to be able to extract information from encrypted instance fields _without_ needing them to be burdened with additional moving parts-- like codecs. So this commit refactors instances and the storage layer so that encryption occurs during instance marshaling and decryption occurs during instance unmarshaling. * Refactor binding encryption/decryption Previously, encryption and decryption of certain binding fields was accomplished through special getters and setters. To use these, you had to pass in a codec. This meant that any bit of code needing to access these fields needs access to the codec. Since I'm entertaining changes to the service.Module interface that will result in modules having access to binding objects, I wanted a way for modules to be able to extract information from encrypted binding fields _without_ needing them to be burdened with additional moving parts-- like codecs. So this commit refactors bindings and the storage layer so that encryption occurs during bindings marshaling and decryption occurs during bindings unmarshaling. * clean up a few errant lint errors * pass instance to provisioning steps * pass instance to deprovisioning steps * make bind and unbind accept instances as arguments * get defensive about possible unanticipated modifications to instances by module code * pass instance to updating steps * minor lint fixes * refactor lifecycle tests * Firewall rule parameterization (#152) * Refactor to support allowing firewall rules. Fixes: #146 Changes to suport optional parameters to allow specifying start and end IP address for firewall rules on MySQL, PostgreSQL and MSSQL. Previously, this defaulted to allow any connection. This isn't ecure in practice, so allowing the customer to specify desired rules. Also changed the default to 0.0.0.0 for both, which results in only Azure internal addresses. Added some parameter validation for the firewall values as well: * Valid IPV4 addresses * Start <= End Extracted a method to build the arm template parameter map. Providing empty strings is invalid for the ARM template, so only including when non-empty. Added parameters to the lifecycle tests to allow running them from local machines (will need ao enhance the charts in helm-charts) Added some tests. * Fixing merge conflicts * Test cleanup * json field cleanup * Resolving code review comments * error message cleanup * error message cleanup * specify firewall rules for relational databased in example manifests * Few minor corrections to the Quickstart (#136) * Correcting the az ad sp delete command * Update quickstart-minikube.md Remove a leading $ from a few bash commands. copy paste on mac is messed up :-) * Update quickstart-minikube.md Clarification around logging in to WordPress * restoring prompts in quickstart * minor punctuation fix * removing distracting echo command * make prompts used consistently * kick of redis tests sooner (#155) * s/module lifecycle tests/service lifecycle tests/g (#156) * make storage know how to deal with service-specific types (#153) * don't use codec where we don't need it anymore (#161) * Updating CircleCI to define DOCKER_REPO env var (#167) * Updating CircleCI to define REGISTRY env var We previously set the REGISTRY environment variable in a deploy script. When that was removed to streamline the release process, we need to define it in Circle. Added two environment blocks to the publish-rc-images and publish-release-images jobs. Fixes: 145 * Qualified registry with docker.io * Changed REGISTRY to DOCKER_REPO * s/context/details/g (#163) * make fake bind/unbind functionality not rely on standard context * collapse standard provisioning context onto instance * do away with (sort of) redundant standard provisioning parameters * s/provisioning context/instance details/ * s/binding context/binding details/ * fix error handling when standard params aren't expected types * add comment that adds some clarity to instance comparisons * don't store credentials as their own field (#169) * quickstart fixes (#174) * adding missing docker pushes to release process * Adding build flags to inject the version and commit (#173) * Adding build flags to inject the version and commit This commit injects the version and commit to the OSBA binary. If there is no version set, then sets the version to “devel” * Using new variable for version * Removing redundant setting for main.commit * Printing version and commit on broker startup * Adding fields to the log message on startup * Removing build flags on the CLI builds * broker doesn't need to hang on to codec (#180) * remove quotes from devel pseudo-version string (#184) Merging without CI. * fix bug decoding tags (#183) * move version info to avoid import cycle later (#185) * Adding user-agent string to requests to ARM (#172) * Adding user-agent string to requests to ARM * get version for user agent string from version package * use separate redis clients for storage and async engine (#182) * Adding ability to filter lifecycle tests (#181) * Adding ability to filter lifecycle tests This adds the ability to filter lifecycle tests using an environment variable (TEST_MODULES). Via this variable, the developer can provide a comma delimited list of module names and only the test cases matching that module will be enabled. If the variable is empty, all tests will run. If an invalid module name is provided, no tests will run. * fix typo and lint error * remove unnecessary funtion * use redis namespaces (#188) * use redis namespaces for instances and bindings * use redis namespaces in async engine * start using lightweight-docker-go image for development tasks (#190) * simplify ci config (#194) * README work (#196) * README formatting + one cmdline fix * Link fixes * Revert heading to title case * append instead of replacing user-agent (#198) * append instead of replacing user-agent * fix lint errors * add deferred task execution to async package (#208) achieving this involved a significant refactor of the entire package. the good news is that the tests are improved beyond what we had before, so i think we can be more confident in the refactored package than the original. * Enhancement of asynch engine This PR enhances the asynch engine to enable the concenpt of delayed or paused tasks. Tasks can be added to the engine in a queue separate from the main work queue. As delayed tasks are added, a resumer is notified and will periodically start the tasks up. Delayed tasks can also be started by engine clients. Using this, pkg/api/provision and deprovision were enhanced to determine if a provision or deprovision action needs to wait for another operation to finish. If so, they submit delayed tasks, either wait_for_parent or wait_for_children, whichever is appropriate. Those new jobs will make the same determination, and either readd themselves as delayed tasks, or initiate the task that would otherwise originally have been created. * Remove resumer changes * removed references to submitdelayed task * Fixed debug statements that got clobbered in merge * Debug message * Fixed response code * Whitespace * Code review comments * Code review comments * Refactor based on code-review comments. * Code review comments * Lint fixes
- Loading branch information