Skip to content

README.md

Latest commit

 

History

History
94 lines (66 loc) · 3.29 KB

README.md

File metadata and controls

94 lines (66 loc) · 3.29 KB

Containerized postgresql

The file Dockerfile.pg-hll builds a postgresql 13 container with HyperLogLog extension.

docker build -f Dockerfile.pg-hll -t pg-hll .

Then run the container like this :

docker run -e POSTGRES_PASSWORD=password -p 5432:5432 pg-hll

Database schema

Migrations

Migrations are managed with yoyo

pip install yoyo-migration psycopg2
yoyo apply --database postgresql://user:password@host:port/database migrations

table nodes

Columns definition :

  • id: uniquely identify an entry in this table
  • name: the usual acronym of the node
  • contact: an email contact

table tokens

Describe all the tokens :

  • id: uniquely identify a token
  • node_id: refernce of the node(id) column
  • value: the value of the token
  • valid_from : timestamp when the token starts to be valid
  • valid_until: when the token expires
  • created_at: timestamp when the token has been created

table payloads

This table keeps track of all the payloads received. In order to prevent the server to ingest twice the same statistics

  • node_id: reference of the node(id) column
  • hash: mmh3 hash of the received payload
  • version: the version of the aggregation schema
  • generated_at: the date when the aggregation has been computed
  • first_stat_at: the first date for the statistic (not the month, but the real event)
  • last_stat_at: the last date for the statistic (not the month, but the real event)
  • created_at: timestamp when the payload has been received

Uniqueness is defined on node_id + hash

table dataselect_stats

Columns definition :

  • node_id: reference of the node(id) column
  • date: the date of the statistics. It's the first day of a month, as the stats are aggregated monthly.
  • network: extended seismic network code (eg Z32015)
  • station, location, channel: usual seismological stream identification
  • country: iso code for the country
  • bytes: quantity of data delivered
  • nb_reqs: number of total requests
  • nb_successful_reqs: number of successful requests (http rc 200)
  • nb_failedrequests: number of failed requests TODO: clarify this definition
  • clients: unique IPs as a HyperLogLog hash
  • created_at: timestamp when the statistic has been submitted
  • updated_at: timestamp when the statistic has been updated

Uniqueness is defined on node_id + date + network + station + location + channel + country

view coverage

This view is used as a helper to consult statistics coverage for each node.

CREATE OR REPLACE VIEW coverage AS 
    SELECT a.name AS node, unnest(b.stats) AS stat_day FROM
        (SELECT nodes.id,nodes.name FROM nodes) a 
        LEFT JOIN 
        (SELECT node_id, array_agg(splitup) as stats FROM 
            (select node_id, unnest(coverage) splitup FROM payloads) splitup 
            group by 1) b 
        ON a.id=b.node_id;

For instance, in order to get the monthly coverage for a given node :

    select node, a.month, round(count(*)/(SELECT extract(days FROM a.month + interval '1 month - 1 day'))*100) as percent from
    (select node, date_trunc('month',stat_day) as month, stat_day from coverage where node='RESIF' and
  date_part('year', stat_day)=2021) as a group by 1,2 order by 1;