Join GitHub today
GitHub is home to over 40 million developers working together to host and review code, manage projects, and build software together.Sign up
Minimum viable speculating command runner #7992
Thanks, this looks sane.
In terms of how to break up #7949, I think that because this ended up being fairly small, you might consider doing a little bit of design for steps 3 and 4 (perhaps as a comment on #7949), and then including the absolute most basic subset of the options in this PR in order to enable actually implementing basic speculation as one or two commits atop the existing one in this PR.
Thanks for putting this together, it looks good :)
If we can make it work, I think I'd rather be using
So ideally the Speculating CommandRunner will just have two
@illicitonion After discussion with @stuhood we decided that for now, and potentially for the long run, we won't extend the speculation runner into a high level retrier. That retry should be handled by lower level components as it is currently done, and we will be opaque about it for now at the speculation level. That is for a given speculation call. The actual RPC may be retried at the command level in a way that the speculating command runner won't know about. Curious to get your thoughts on this potential design, and if you are happy with the current retry capabilities.
On retrying, there are a few use cases for retrying:
Jul 10, 2019
illicitonion left a comment
Looks good, thanks!
I'd probably revert the
I don't know much about cancellation - would it be possible to add a test which shows that cancellation works properly? I imagine something like the