Conversation
|
Since I have encountered initialisation order problems in the past (nice to see you've removed some duct tape introduced in #1812), I think we are better off with an inversion of control model than singletons, which will lead to deterministic initialisation order. I think this would be a nice direction for this PR to take. |
bantonsson
left a comment
There was a problem hiding this comment.
Great that you untangle this!
|
As discussed offline, the question of deterministic initialisation is another matter and can be addressed independently of this. |
066b8a2 to
c4e4d67
Compare
|
Can we prioritise getting this merged? |
c4e4d67 to
9b1cca9
Compare
… own executors where appropriate.
2b8aaa2 to
7271fc4
Compare
stale review from when PR was a draft
richardstartin
left a comment
There was a problem hiding this comment.
Looks really good. Not related to this PR but I think it's probably a mistake to be using the CommonTaskExecutor in DDAgentApi. We can fix that in another PR.
|
re. |
Changes:
CommonTaskExecutortoAgentTaskSchedulerto better reflect this is the agent's own task schedulerDelayQueue