cockroachdb / cockroachdb/cockroach
Discrepancy in 30min interval metric visualisation
- Dominant language
- Go
- Stars
- 32.5k
- Forks
- 4.1k
- PR merge metrics
- PR metrics pending
Description
**Describe the problem**
We have collected the tsdump with 30min metric interval for single node cluster. We compared the metrics dashboard widgets in both DB Console and Datadog uploaded tsdump. We observed that there is a discrepancy in the gauge and histogram values. We tried to capture the tsdump with text format and the values are matching with the Datadog uploaded values. We need to bring the parity between the metrics visualised in DB console and Datadog.
**To Reproduce**
- spin up the single node cluster with command `COCKROACH_DEBUG_TS_IMPORT_FILE=./cockroach/crl_30_min.raw ./cockroach/cockroach start-single-node --insecure --advertise-addr 127.0.0.1`
- upload the tsdump to Datadog with command `./cockroach debug tsdump --format datadog --dd-api-key $DD_API_KEY --cluster-label="crl-30-min" --dd-site=us5 crl_30_min.raw`
- Datadog dashboard will take couple of hours to load the data.
- Compare gauge and histogram metrics (DB console metric name vs Datadog metric name):
- `cr.node.sql.conns` vs `cockroachdb.sql.conns`
- `cr.node.exec.latency-p99` vs `cockroachdb.exec.latency_p99`
**Expected behavior**
The graphs should match for gauge and histogram metrics.
**Additional data / screenshots**
- [crl_30_min.raw.zip](https://github.com/user-attachments/files/22924840/crl_30_min.raw.zip)
- [crl_30_min.txt](https://github.com/user-attachments/files/22924848/crl_30_min.txt)
- `cr.node.sql.conns` vs `cockroachdb.sql.conns`
- `cr.node.exec.latency-p99` vs `cockroachdb.exec.latency_p99`
Jira issue: CRDB-55433
Contributor guide
Assessment
This issue has not been assessed yet.