-
Notifications
You must be signed in to change notification settings - Fork 9
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
multiple exemplars per sample #34
base: main
Are you sure you want to change the base?
multiple exemplars per sample #34
Conversation
Signed-off-by: György Krajcsovits <[email protected]>
In the current solution we have a circular buffer for all exemplars. The exemplars are organized into two lists. One list links all exemplars from oldest received to newest so that we can make space when maximum number of exemplars is hit on insert. The other link | ||
is per time series, linking exemplars from oldest to newest. Also there is a pointer to the latest exemplar per time series to be able to check for duplicates, out of order. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do you mean "the other list" rather than "the other link"? Would think this refers to the second list.
In the current solution we have a circular buffer for all exemplars. The exemplars are organized into two lists. One list links all exemplars from oldest received to newest so that we can make space when maximum number of exemplars is hit on insert. The other link | |
is per time series, linking exemplars from oldest to newest. Also there is a pointer to the latest exemplar per time series to be able to check for duplicates, out of order. | |
In the current solution we have a circular buffer for all exemplars. The exemplars are organized into two lists. One list links all exemplars from oldest received to newest so that we can make space when maximum number of exemplars is hit on insert. The other list | |
is per time series, listing exemplars from oldest to newest. Also there is a pointer to the latest exemplar per time series to be able to check for duplicates, out of order. |
|
||
* There is no interface to pass multiple exemplars to head.AppendExemplar, so multiple calls need to be made, increasing the call count and lock contention. | ||
* If a native histogram has 10 exemplars per scrape/sample and all are received in the same remote write request, we currently try appending all in a loop. We consider them out of order if all are out of order, but if there's a duplicate of the last stored exemplar or some are newer than the last stored exemplar, then we just ignore the old exemplars. | ||
* However the logic above fails if exemplars are received in multiple remote write requests, so the logic is a workaround at best. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
[Nit]
* However the logic above fails if exemplars are received in multiple remote write requests, so the logic is a workaround at best. | |
* However, the logic above fails if exemplars are received in multiple remote write requests, so the logic is a workaround at best. |
|
||
Currently the remote write protocol (1.0 and 2.0) processes exemplars individually, which means that if a native histogram has multiple exemplars, they may arrive in separate remote write requests, making it hard to decide if the exemplar is out of order it belonged to a previous set of exemplars and should be considered a duplicate. | ||
|
||
Currently there is no mapping of the "set of exemplars" per sample concept that is in Open Telemetry specification. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nit for consistency :)
Currently there is no mapping of the "set of exemplars" per sample concept that is in Open Telemetry specification. | |
Currently there is no mapping of the "set of exemplars" per sample concept that is in OpenTelemetry specification. |
No description provided.