Issue: #14989
The initial step in optimizing segment metadata was to centralize the construction of datasource schema in the Coordinator (#14985). Thereafter, we addressed the problem of publishing schema for realtime segments (#15475). Subsequently, our goal is to eliminate the requirement for regularly executing queries to obtain segment schema information.
This is the final change which involves publishing segment schema for finalized segments from task and periodically polling them in the Coordinator.
In the current design, brokers query both data nodes and tasks to fetch the schema of the segments they serve. The table schema is then constructed by combining the schemas of all segments within a datasource. However, this approach leads to a high number of segment metadata queries during broker startup, resulting in slow startup times and various issues outlined in the design proposal.
To address these challenges, we propose centralizing the table schema management process within the coordinator. This change is the first step in that direction. In the new arrangement, the coordinator will take on the responsibility of querying both data nodes and tasks to fetch segment schema and subsequently building the table schema. Brokers will now simply query the Coordinator to fetch table schema. Importantly, brokers will still retain the capability to build table schemas if the need arises, ensuring both flexibility and resilience.
* Use Druid's extension loading for integration test instead of maven
* fix maven command
* override config path
* load input format extensions and kafka by default; add prepopulated-data group
* all docker-composes are overridable
* fix s3 configs
* override config for all
* fix docker_compose_args
* fix security tests
* turn off debug logs for overlord api calls
* clean up stuff
* revert docker-compose.yml
* fix override config for query error test; fix circular dependency in docker compose
* add back some dependencies in docker compose
* new maven profile for integration test
* example file filter
* support using mariadb connector with mysql extensions
* cleanup and more tests
* fix test
* javadocs, more tests, etc
* style and more test
* more test more better
* missing pom
* more pom
* integration test for coordinator and overlord leadership, added sys.servers is_leader column
* docs
* remove not needed
* fix comments
* fix compile heh
* oof
* revert unintended
* fix tests, split out docker-compose file selection from starting cluster, use docker-compose down to stop cluster
* fixes
* style
* dang
* heh
* scripts are hard
* fix spelling
* fix thing that must not matter since was already wrong ip, log when test fails
* needs more heap
* fix merge
* less aggro
* Fixes and tests related to the Indexer process.
Three bugs fixed:
1) Indexers would not announce themselves as segment servers if they
did not have storage locations defined. This used to work, but was
broken in #9971. Fixed this by adding an "isSegmentServer" method
to ServerType and updating SegmentLoadDropHandler to always announce
if this method returns true.
2) Certain batch task types were written in a way that assumed "isReady"
would be called before "run", which is not guaranteed. In particular,
they relied on it in order to initialize "taskLockHelper". Fixed this
by updating AbstractBatchIndexTask to ensure "isReady" is called
before "run" for these tasks.
3) UnifiedIndexerAppenderatorsManager did not properly handle complex
datasources. Introduced DataSourceAnalysis in order to fix this.
Test changes:
1) Add a new "docker-compose.cli-indexer.yml" config that spins up an
Indexer instead of a MiddleManager.
2) Introduce a "USE_INDEXER" environment variable that determines if
docker-compose will start up an Indexer or a MiddleManager.
3) Duplicate all the jdk8 tests and run them in both MiddleManager and
Indexer mode.
4) Various adjustments to encourage fail-fast errors in the Docker
build scripts.
5) Various adjustments to speed up integration tests and reduce memory
usage.
6) Add another Mac-specific approach to determining a machine's own IP.
This was useful on my development machine.
7) Update segment-count check in ITCompactionTaskTest to eliminate a
race condition (it was looking for 6 segments, which only exist
together briefly, until the older 4 are marked unused).
Javadoc updates:
1) AbstractBatchIndexTask: Added javadocs to determineLockGranularityXXX
that make it clear when taskLockHelper will be initialized as a side
effect. (Related to the second bug above.)
2) Task: Clarified that "isReady" is not guaranteed to be called before
"run". It was already implied, but now it's explicit.
3) ZkCoordinator: Clarified deprecation message.
4) DataSegmentServerAnnouncer: Clarified deprecation message.
* Fix stop_cluster script.
* Fix sanity check in script.
* Fix hashbang lines.
* Test and doc adjustments.
* Additional tests, and adjustments for tests.
* Split ITs back out.
* Revert change to druid_coordinator_period_indexingPeriod.
* Set Indexer capacity to match MM.
* Bump up Historical memory.
* Bump down coordinator, overlord memory.
* Bump up Broker memory.
* integration test refactor
* integration test refactor
* refactor integration test
* refactor integration test
* refactor integration test
* refactor integration test
* refactor integration test
* refactor integration test
* refactor integration test
* refactor integration test
* address comments
Rishabh Singh
Jihoon Son
Clint Wylie
Maytas Monsereenusorn
Himadri Singh
zachjsh
Gian Merlino
Maytas Monsereenusorn