mirror of https://github.com/prometheus/prometheus
Browse Source
When samples are committed in the head, they are also written to the WAL. The order of WAL records should be sample then exemplar, but this was not the case for native histogram samples. This PR fixes that. The problem with the wrong order is that remote write reads the WAL and sends the recorded timeseries in the WAL order, which means exemplars arrived before histogram samples. If the receiving side is Prometheus TSDB and the series has not existed before then the exemplar does not currently create the series. Which means the exemplar is rejected and lost. Signed-off-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>pull/13092/head^2
George Krajcsovits
1 year ago
committed by
GitHub
2 changed files with 69 additions and 8 deletions
Loading…
Reference in new issue