mirror of
https://github.com/fnproject/fn.git
synced 2022-10-28 21:29:17 +03:00
88635fcf08
* remove usage of time.After these little guys leak in the background, it's easy enough to remove usage of them altogether. this reduces cpu load on fn, the main culprit was the one in waitHot which polls every 200ms. if you run a load test where functions get a slot and exec a lot faster than 200ms then these things pile up kinda quickly. * common.NewTimer this covers up the hole left by the stdlib timer, see: https://golang.org/pkg/time/#Timer.Reset and changes all usage of time.NewTimer to use common.NewTimer for posterity, should anyone decide to start calling reset on any timers where we previously were not, this should make them safe. we'll have to catch new timers in review. see comment by rsc https://github.com/golang/go/issues/14038#issuecomment-219909704 for details about the thorny side effects of the stdlib handling this behavior (namely, that the returned value of Reset's behavior could change). in any case, this is desirable for our use case from what I can tell in all cases, as we're not particularly interested in the return value from Stop nor Reset. * replace the timer instead of resetting since the Reset contract for the timer is messed up, if we wrap the timer and implement reset by just making a new timer we can get the behavior we want without having to dance around in the caller making sure we stop timers in any of our loops before resetting them, since Reset is supposed to behave this way anyway. and with this, so long as we defer stop it will stop the last assigned timer. so I think this checks out