Page MenuHomePhabricator

EPIC: Trino/MinIO/Hive-Standalone-Metaserver/Dagster/Metabase/Superset Implementation
Open, In Progress, MediumPublic

Description

Epic designed to hold all tasks related to implementation of Trino/MinIO/Hive-Standalone-Metastore, as part of a Fundraising analytics platform overhaul.

Significant Fundraising Analytics Platform Changes

At the start of the project the Fundraising analytics platform had the following on-premises hardware/software components :

  • MariaDB analytics database (frdb)
  • banner impression logs collected by kafkatee (frban)
  • data cube creation with with python scripts and Django Banner Stats (fran)
  • process scheduling with process-control (fran)
  • Superset frontend (fran)

This project makes the following on-premises platform changes:

  • network upgrade from 1Gbps to 10Gbps to accomodate anticipated Trino/MinIO network activity
  • replace MariaDB data store with a Trino (fransc/fransw) and MinIO (franio) data lake
  • replace process-control scheduling with Dagster (fran)
  • migrate cube generation code and scheduling to Dagster/DBT (fran)
  • migrate frontend from Superset to Metabase (fran)
  • replace banner impression pipeline with direct ingestion to Trino/MinIO from kafka (deprecates frban)

Project implementation has been broken into these general phases:

FY24-25

  • migrate active analytics services to codfw
  • deploy codfw Trino cluster of 1 coordinator (fransc), 3 workers (fransw)
  • deploy codfw MinIO cluster of 3 nodes (franio)
  • deploy minimal eqiad Trino cluster of 1 coordinator (fransc), 1 worker (fransw)
  • deploy eqiad MinIO cluster of 3 nodes (franio)
  • deploy services in parallel to existing Fundraising analytics services
  • limited reliability expectations on new platform during development
  • no expectation of availability of failover to eqiad rundant cluster

FY25-26

  • additional MinIO node at each datacenter for production availability
  • adjust eqiad Trino worker node count based on codfw cluster performance
  • bring eqiad redundant cluster fully online
  • phase out old banner impression pipeline (deprecate frban)
  • phase out MariaDB+Superset platform

Network Notes:

  • no change to analytics platform location, all hosts remain within frack-fundraising vlan
  • the most significant traffic is expected between Trino and MinIO nodes
  • cross-vlan and cross-firewall traffic are not expected to change dramatically
  • there will some cross-datacenter traffic for data replication to the redundant MinIO cluster
  • there should be a reduction in cross-datacenter traffic from analytics MariaDB replication
  • analytics platform and services are outside of PCI scope

Related Objects

StatusSubtypeAssignedTask
In Progressgreg
DuplicateJgreen
ResolvedJgreen
ResolvedJgreen
ResolvedJgreen
ResolvedJgreen
ResolvedArnis.dizdari
ResolvedArnis.dizdari
ResolvedArnis.dizdari
OpenAStein-WMF
ResolvedAStein-WMF
OpenNone
ResolvedAStein-WMF
ResolvedIAckerman-WMF
ResolvedDwisehaupt
ResolvedArnis.dizdari
ResolvedJgreen
ResolvedIrisbdc
ResolvedIrisbdc
ResolvedArnis.dizdari
Resolvedyuki_velir
ResolvedJgreen
ResolvedDwisehaupt
ResolvedJgreen
ResolvedJgreen
ResolvedJgreen
ResolvedJgreen
ResolvedJgreen
OpenAStein-WMF
OpenJgreen
ResolvedDwisehaupt
ResolvedDwisehaupt
ResolvedDwisehaupt
ResolvedJgreen
ResolvedJgreen
ResolvedVRiley-WMF
ResolvedDwisehaupt
ResolvedDwisehaupt
ResolvedJgreen
ResolvedJgreen
OpenNone
ResolvedAStein-WMF
ResolvedJgreen
ResolvedJgreen
OpenArnis.dizdari
ResolvedJgreen
ResolvedDwisehaupt
ResolvedPapaul
ResolvedDwisehaupt
ResolvedDwisehaupt
ResolvedDwisehaupt
ResolvedDwisehaupt
OpenJMando
ResolvedDavidbooke
ResolvedAStein-WMF
ResolvedDavidbooke
OpenNone
ResolvedArnis.dizdari
OpenNone
ResolvedDwisehaupt
ResolvedDwisehaupt
OpenNone

Event Timeline

Jgreen renamed this task from EPIC: Trino/minIO/Hive Implementation to EPIC: Trino/minIO/Hive/Hadoop Implementation.Oct 18 2024, 8:44 PM
Jgreen updated the task description. (Show Details)
Jgreen changed the task status from Open to In Progress.Oct 28 2024, 5:37 PM
Jgreen triaged this task as Medium priority.
Jgreen changed the status of subtask T377364: [Trino/Hive] Setup and configuration from Open to In Progress.
Gehel moved this task from Scratch to Watching on the Data-Platform-SRE board.
Jgreen renamed this task from EPIC: Trino/minIO/Hive/Hadoop Implementation to EPIC: Trino/minIO/Hive-Standalone-Metaserver/Dagster/Metabase/Superset Implementation.Feb 6 2025, 3:28 PM
Jgreen removed Jgreen as the assignee of this task.Apr 8 2025, 3:09 PM
Jgreen subscribed.

Epic task, doesn't make sense for me to own it!

Jgreen renamed this task from EPIC: Trino/minIO/Hive-Standalone-Metaserver/Dagster/Metabase/Superset Implementation to EPIC: Trino/MinIO/Hive-Standalone-Metaserver/Dagster/Metabase/Superset Implementation.Apr 17 2025, 12:35 PM
Jgreen updated the task description. (Show Details)