You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Apr 3, 2019. It is now read-only.
While testing the final fix for #2496, I noticed that the email metrics rely on a flowId and flowBeginTime that are manually copied from the request payload instead of via metricsContext.gather, which checks both the payload and memcached. This means that, with the content server stopping sending metrics context data to some endpoints in train 123, there would be a blip in the metrics.
I am working on a fix now, but paused to open this issue because the fix has led me to a wider refactoring of metricsContext.gather. Essentially, I intend to change it so we have a lazy getter for metrics context data on the request object. This has the mild benefit of stopping us from hitting gather multiple times per request as we do now (mild because memcached is fast) and the more significant benefit of getting rid of the slightly awkward interface for gather, where it mutates its input rather than returns fresh state.
It's not a hugely messy change, but it's big enough that I thought there should be a proper issue detailing the rationale, for us to refer back to in the future.
The text was updated successfully, but these errors were encountered:
While testing the final fix for #2496, I noticed that the email metrics rely on a
flowId
andflowBeginTime
that are manually copied from the request payload instead of viametricsContext.gather
, which checks both the payload and memcached. This means that, with the content server stopping sending metrics context data to some endpoints in train 123, there would be a blip in the metrics.I am working on a fix now, but paused to open this issue because the fix has led me to a wider refactoring of
metricsContext.gather
. Essentially, I intend to change it so we have a lazy getter for metrics context data on therequest
object. This has the mild benefit of stopping us from hittinggather
multiple times per request as we do now (mild because memcached is fast) and the more significant benefit of getting rid of the slightly awkward interface forgather
, where it mutates its input rather than returns fresh state.It's not a hugely messy change, but it's big enough that I thought there should be a proper issue detailing the rationale, for us to refer back to in the future.
The text was updated successfully, but these errors were encountered: