fix: pullsync make offer timeout #3520
Conversation
istae
requested review from
a team,
mrekucci and
aloknerurkar
and removed request for
a team
| @@ -29,7 +29,7 @@ const loggerName = "puller" | |||
|
|
|||
| var errCursorsLength = errors.New("cursors length mismatch") | |||
|
|
|||
| const DefaultSyncErrorSleepDur = time.Second * 5 | |||
| const DefaultSyncErrorSleepDur = time.Second * 30 | |||
There was a problem hiding this comment.
I'm just curious as to why the length of sleep needs to be increased.
There was a problem hiding this comment.
it's an arbitrary value, 5 seconds seemed a bit aggressive to me after some thinking
| @@ -50,6 +50,7 @@ var ( | |||
|
|
|||
| const ( | |||
| storagePutTimeout = 5 * time.Second | |||
| makeOfferTimeout = 5 * time.Minute | |||
There was a problem hiding this comment.
My nodes are (well, were before the stampes expired) routinely taking longer than 5 minutes to construct the offer. makeOffer's context is automatically cancelled at the end of the commit round by the agent, so why have a local timeout specified here?
There was a problem hiding this comment.
this is puller/pullsync stuff, not reserve sampler :)
There was a problem hiding this comment.
Duh. makeOffer, not makeSample! That's what I get for commenting before I'm fully awake. I'll blame it on Nicole, the hurricane that passed by us last night!
There was a problem hiding this comment.
jeez, a bad year for florida this year! so many hurricanes
There was a problem hiding this comment.
At least Nicole was barely a category 1, just above a tropical storm. Amazingly, we didn't even lose power; a few blinks overnight, but no outright loss.
Checklist
Description
Should alleviate goroutine pile up for intervals that never receive a chunk.
Open API Spec Version Changes (if applicable)
Motivation and Context (Optional)
Related Issue (Optional)
Screenshots (if appropriate):
This change is