Open euroelessar opened 5 months ago
Pinging code owners:
processor/deltatocumulative: @sh0rez @RichieSams @jpkrohling
See Adding Labels via Comments if you do not have permissions to add labels yourself.
This issue has been inactive for 60 days. It will be closed in 60 days if there is no activity. To ping code owners by adding a component label, see Adding Labels via Comments, or if you are unsure of which component this issue relates to, please ping @open-telemetry/collector-contrib-triagers
. If this issue is still relevant, please ping the code owners or leave a comment explaining why it is still relevant. Otherwise, please close it.
Pinging code owners:
See Adding Labels via Comments if you do not have permissions to add labels yourself.
Hitting this issue as well
https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/34157 should address it, we've been using it in production for over a month. Can someone please review?
This issue has been inactive for 60 days. It will be closed in 60 days if there is no activity. To ping code owners by adding a component label, see Adding Labels via Comments, or if you are unsure of which component this issue relates to, please ping @open-telemetry/collector-contrib-triagers
. If this issue is still relevant, please ping the code owners or leave a comment explaining why it is still relevant. Otherwise, please close it.
Pinging code owners:
See Adding Labels via Comments if you do not have permissions to add labels yourself.
This problem is still relevant and we would not be able to use deltatocumulative in production without https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/34157.
@sh0rez @RichieSams @jpkrohling please review 😄
Component(s)
processor/deltatocumulative
What happened?
Description
Currently size of an individual histogram is unbounded and can grow until OOM is reached. This is prominent if an application has large distribution of data overall but only relatively small in an individual delta datapoint. In this case cumulative aggregation keeps the scale but keeps growing number of buckets to fit all the data.
Steps to Reproduce
Send two histogram datapoints with the same scale but drastically different offsets.
Expected Result
Cumulative exponential histogram is downscaled to keep around ≈160 buckets.
Actual Result
Exponential histogram grows number of buckets instead, leading to OOM.
Collector version
v0.101.0
Environment information
Environment
OS: Linux Compiler(if manually compiled): go 1.22.2
OpenTelemetry Collector configuration
No response
Log output
No response
Additional context
No response