Thinking some more about the problem, I think your solution of reserving some of the deadline to handle the spanner log in the error case sounds like the best answer. However it does lead to questions like, if you reserve n seconds to log to spanner in the error path, and it takes longer than n, what happens to the error, is it dropped?
On Thursday, 16 August 2018 14:44:08 UTC+10, Dave Cheney wrote: > > What would happen if you write the error to spanner with a setting > context.Backgrond(), ie, no deadline? > > On 16 August 2018 at 13:57, Robert Bartoszynski <r...@google.com> wrote: > > Thanks. Perhaps an alternative would be for me to create a child context > > with a deadline of (context - x seconds) and pass that to OtherService, > with > > the expectation that there should be x seconds left over for the write > to > > spanner. > > > > On Wed, Aug 15, 2018 at 8:10 PM Dave Cheney <d...@cheney.net> wrote: > >> > >> > >> > >> On Thursday, 16 August 2018 12:15:57 UTC+10, r...@google.com wrote: > >>> > >>> As an example: > >>> Client calls MyService with a deadline of 10 seconds. > >>> MyService calls OtherService as part of responding. However, the call > to > >>> OtherService times out due to the deadline in 10 seconds. > >>> MyService tries to log the error to Spanner; but it's still using that > >>> context deadline which expired. > >>> > >>> Is there a way to get a new context with an extended deadline? Are > there > >>> any issues with that approach? > >> > >> > >> Sure, just call context.WithDeadline(context.Background()) and use that > >> instead > >> > >> The difficulty is you want the deadline of this new context to live > >> beyond its parent. Logically it feels like this new context is > subordinate > >> to the previous, but by design we've said that the new context is _not_ > >> subordinate -- the deadline does not apply to it. I think you need to > >> address this incongruousness before proceeding. > >> > >>> > >>> > >>> > >>> On Tuesday, August 14, 2018 at 1:25:47 PM UTC-7, r...@google.com > wrote: > >>>> > >>>> I'm working on a service that write some log info to spanner when > it's > >>>> done responding to a request. > >>>> However, the service uses the context's deadline to write to spanner, > so > >>>> if the deadline expires due to some long running RPC, the write to > spanner > >>>> fails (because the deadline expired), and we don't get any log info. > >>>> > >>>> What's the best practice for dealing with this situation? > >> > >> -- > >> You received this message because you are subscribed to a topic in the > >> Google Groups "golang-nuts" group. > >> To unsubscribe from this topic, visit > >> https://groups.google.com/d/topic/golang-nuts/TUicHyvYNX0/unsubscribe. > >> To unsubscribe from this group and all its topics, send an email to > >> golang-nuts+unsubscr...@googlegroups.com. > >> For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
