Hi,
Would like to know the behavior of Prometheus (in newer versions) with respect to ingestion of old data. Is there a hard limit on the time-window such that any samples outside this window are dropped ?
I have been looking at Prometheus test this behavior. I have seen Prometheus ingesting 25-30 mins old data without any complaint. However, if data is 4 hrs old, the messages are not accepted and are dropped. The error messages contain something like:
msg="Error on ingesting samples that are too old or are too far into the future" num_dropped=47190
Regarding this requirement on Prometheus roadmap page,
https://prometheus.io/docs/introduction/roadmap/, it says
“Backfilling will permit bulk loads of data in the past. This will allow for retroactive rule evaluations, and transferring old data from other monitoring systems.”
Are any of these functionalities implemented in Prometheus today ?
I am testing Prometheus behavior/performance by making Prometheus consume data from a Kafka topic. If we make Prometheus not scrape REST end points,
then we also need to handle situations when Prometheus has to read and catch up reading old data from Kafka before consuming new data. So, knowing Prometheus’s
behavior becomes important. Is there any alternative or better way to back-fill old data into Prometheus ? Are there any APIs to push old bulk data into Prometheus TSDB ?
Thanks,
Dhiman