SWATCH-654: bugfix: only a single set of measurements kept per day

[kahowell]

This corrects a bug where we were clearing hourly measurements on processing each successive hour in a given day. This was happening because we adjust the range of fetched existing snapshots to include enough to aggregate the next granularity (hourly -> daily), and we were treating all fetched snapshots as affected by the gathered usage (rather than just the given hour).

Testing

Run through these steps once on hotfix or main, and then on this branch.

Run mock prometheus:

bin/prometheus-mock-data.sh

Clear state:

psql -h localhost -U rhsm-subscriptions <<EOF
delete from tally_snapshots cascade;
delete from hosts cascade;
EOF

Run the application:

DEV_MODE=true \
  USER_USE_STUB=true \
  PROM_URL=http://localhost:9090/api/v1 \
  ./gradlew :bootRun

Gather metrics (adjust date/time as needed; the mock prometheus generates data for the past 24 hours):

http :9000/hawtio/jolokia \
  type=exec \
  mbean=org.candlepin.subscriptions.metering.jmx:name=meteringJmxBean,type=MeteringJmxBean \
  operation='performCustomMetering(java.lang.String,java.lang.String,java.lang.Integer)' \
  arguments:='["rhosak","2022-10-25T12:00:00Z",240]'

Run tally against several successive hours:

http :9000/hawtio/jolokia \
  type=exec \
  mbean=org.candlepin.subscriptions.jmx:name=tallyJmxBean,type=TallyJmxBean \
  operation='tallyAllAccountsByHourly(java.lang.String,java.lang.String)' \
  arguments:='["2022-10-25T09:00:00Z","2022-10-25T10:00Z"]'
sleep 2
http :9000/hawtio/jolokia \
  type=exec \
  mbean=org.candlepin.subscriptions.jmx:name=tallyJmxBean,type=TallyJmxBean \
  operation='tallyAllAccountsByHourly(java.lang.String,java.lang.String)' \
  arguments:='["2022-10-25T10:00:00Z","2022-10-25T11:00Z"]'
sleep 2
http :9000/hawtio/jolokia \
  type=exec \
  mbean=org.candlepin.subscriptions.jmx:name=tallyJmxBean,type=TallyJmxBean \
  operation='tallyAllAccountsByHourly(java.lang.String,java.lang.String)' \
  arguments:='["2022-10-25T11:00:00Z","2022-10-25T12:00Z"]'

Check the tally snapshots measurements:

• on main/hotfix notice that the measurements were cleared
• on this branch, notice they are intact

psql -h localhost -U rhsm-subscriptions <<EOF
select id, value, snapshot_date
from tally_snapshots s
         left join tally_measurements m on m.snapshot_id = s.id
where sla = '_ANY'
  and usage ='_ANY'
  and billing_account_id='_ANY'
  and granularity='HOURLY'
  and billing_provider='_ANY'
  and (measurement_type='PHYSICAL'
   or measurement_type is null)
order by snapshot_date asc
EOF