Changes to `cost` strategy:
- In every `ServerHolder`, track the number of segments per datasource per interval
- Perform cost computations for a given interval just once, and then multiply by a constant
factor to account for the total segment count in that interval
- Do not perform joint cost computations with segments that are outside the compute interval
(± 45 days) for the segment being considered for move
- Remove metrics `segment/cost/*` as they were coordinator killers! Turning on these metrics
(by setting `emitBalancingStats` to true) has often caused the coordinator to be stuck for hours.
Moreover, they are too complicated to decipher and do not provide any meaningful insight into
a Druid cluster.
- Add new simpler metrics `segment/balancer/compute/*` to track cost computation time,
count and errors.
Other changes:
- Remove flaky test from `CostBalancerStrategyTest`.
- Add tests to verify that computed cost has remained unchanged
- Remove usages of mock `BalancerStrategy` from `LoadRuleTest`, `BalanceSegmentsTest`
- Clean up `BalancerStrategy` interface
Changes:
- Throw an `InsertCannotAllocateSegmentFault` if the allocated segment is not aligned with
the requested granularity.
- Tests to verify new behaviour
* Updates: use the target table directly, sanitized replace time chunks and clustered by cols.
* Add DruidSqlParserUtil and tests.
* minor refactor
* Use SqlUtil.isLiteral
* Throw ValidationException if CLUSTERED BY column descending order is specified.
- Fails query planning
* Some more tests.
* fixup existing comment
* Update comment
* checkstyle fix: remove unused imports
* Remove InsertCannotOrderByDescendingFault and deprecate the fault in readme.
* minor naming
* move deprecated field to the bottom
* update docs.
* add one more example.
* Collapsible query and result
* checkstyle fixes
* Code cleanup
* order by changes
* conditionally set attributes only for explain queries.
* Cleaner ordinal check.
* Add limit test and update javadoc.
* Commentary and minor adjustments.
* Checkstyle fixes.
* One more checkArg.
* add unexpected kind to exception.
* MSQ: Change default clusterStatisticsMergeMode to SEQUENTIAL.
This is an undocumented parameter that controls how cluster-by statistics
are merged. In PARALLEL mode, statistics are gathered from workers all
at once. In SEQUENTIAL mode, statistics are gathered time chunk by time
chunk. This improves accuracy for jobs with many time chunks, and reduces
memory usage.
The main downside of SEQUENTIAL is that it can take longer, but in most
situations I've seen, PARALLEL is only really usable in cases where the
sketches are small enough that SEQUENTIAL would also run relatively
quickly. So it seems like SEQUENTIAL is a better default.
* Switch off-test from SEQUENTIAL to PARALLEL.
* Fix sequential merge for situations where there are no time chunks at all.
* Add a couple more tests.
* Add OverlordStatusMonitor and CoordinatorStatusMonitor to monitor service leader status
* make the monitor more general
* resolve conflict
* use Supplier pattern to provide metrics
* reformat code and doc
* move service specific tag to dimension
* minor refine
* update doc
* reformat code
* address comments
* remove declared exception
* bind HeartbeatSupplier conditionally in Coordinator
Users can now add a guardrail to prevent subquery’s results from exceeding the set number of bytes by setting druid.server.http.maxSubqueryRows in Broker's config or maxSubqueryRows in the query context. This feature is experimental for now and would default back to row-based limiting in case it fails to get the accurate size of the results consumed by the query.
* SqlResults: Coerce arrays to lists for VARCHAR.
Useful for STRING_TO_MV, which returns VARCHAR at the SQL layer and an
ExprEval with String[] at the native layer.
* Fix style.
* Improve test coverage.
* Remove unnecessary throws.
Recently, we have seen flakiness in these two tests, apparently due to
computations based on Runtime.getRuntime().maxMemory() differing during
static initialization and in the actual tests. I can't think of a reason
why this would be happening, but anyway, this patch switches the tests to
use the statics instead of recomputing Runtime.getRuntime().maxMemory().
* SQL OperatorConversions: Introduce.aggregatorBuilder, allow CAST-as-literal.
Four main changes:
1) Provide aggregatorBuilder, a more consistent way of defining the
SqlAggFunction we need for all of our SQL aggregators. The mechanism
is analogous to the one we already use for SQL functions
(OperatorConversions.operatorBuilder).
2) Allow CASTs of constants to be considered as "literalOperands". This
fixes an issue where various of our operators are defined with
OperandTypes.LITERAL as part of their checkers, which doesn't allow
casts. However, in these cases we generally _do_ want to allow casts.
The important piece is that the value must be reducible to a constant,
not that the SQL text is literally a literal.
3) Update DataSketches SQL aggregators to use the new aggregatorBuilder
functionality. The main user-visible effect here is [2]: the aggregators
would now accept, for example, "CAST(0.99 AS DOUBLE)" as a literal
argument. Other aggregators could be updated in a future patch.
4) Rename "requiredOperands" to "requiredOperandCount", because the
old name was confusing. (It rhymes with "literalOperands" but the
arguments mean different things.)
* Adjust method calls.
* Clarify compaction docs.
The prior wording made it sound like segmentGranularity, queryGranularity,
and rollup are always required for granularitySpec. They are not required,
but they are strongly recommended. The adjusted wording hopefully does
a better job of making that clear.
* Fix link.
* Wording adjustments.
* S3: Attach SSE key to doesObjectExist calls.
We did not previously attach the SSE key to the doesObjectExist request,
leading to an inconsistency that may cause problems on "S3-compatible"
implementations. This patch implements doesObjectExist using similar
logic to the S3 client itself, but calls our implementation of
getObjectMetadata rather than the S3 client's, ensuring the request
is decorated with the SSE key.
* Fix tests.
New metrics:
- `segment/metadatacache/refresh/time`: time taken to refresh segments per datasource
- `segment/metadatacache/refresh/count`: number of segments being refreshed per datasource
* Fix compatibility issue with SqlTaskResource
The DruidException changes broke the response format
for errors coming back from the SqlTaskResource, so fix those
Changes:
- Add property `useDefaultTierForNull` for all load rules. This property determines the default
value of `tieredReplicants` if it is not specified. When true, the default is `_default_tier => 2 replicas`.
When false, the default is empty, i.e. no replicas on any tier.
- Fix validation to allow empty replicants map, so that the segment is used but not loaded anywhere.
Our CI system has a lot of tests. And much of this testing is really unnecessary for most of the PRs. This PR adds some checks so we can skip these expensive tests when we know they are not necessary.
Added a new monitor SysMonitorOshi to replace SysMonitor. The new monitor has a wider support for different machine architectures including ARM instances. Please switch to SysMonitorOshi as SysMonitor is now deprecated and will be removed in future releases.
Add a new planning strategy that explicitly decouples the DAG from building the native query.
With this mode, it is Calcite's job to generate a "logical DAG" which is all of the various
DruidProject, DruidFilter, etc. nodes. We then take those nodes and use them to build a native
query. The current commit doesn't pass all tests, but it does work for some things and is a
decent starting baseline.
This commit does a complete revamp of the coordinator to address problem areas:
- Stability: Fix several bugs, add capabilities to prioritize and cancel load queue items
- Visibility: Add new metrics, improve logs, revamp `CoordinatorRunStats`
- Configuration: Add dynamic config `smartSegmentLoading` to automatically set
optimal values for all segment loading configs such as `maxSegmentsToMove`,
`replicationThrottleLimit` and `maxSegmentsInNodeLoadingQueue`.
Changed classes:
- Add `StrategicSegmentAssigner` to make assignment decisions for load, replicate and move
- Add `SegmentAction` to distinguish between load, replicate, drop and move operations
- Add `SegmentReplicationStatus` to capture current state of replication of all used segments
- Add `SegmentLoadingConfig` to contain recomputed dynamic config values
- Simplify classes `LoadRule`, `BroadcastRule`
- Simplify the `BalancerStrategy` and `CostBalancerStrategy`
- Add several new methods to `ServerHolder` to track loaded and queued segments
- Refactor `DruidCoordinator`
Impact:
- Enable `smartSegmentLoading` by default. With this enabled, none of the following
dynamic configs need to be set: `maxSegmentsToMove`, `replicationThrottleLimit`,
`maxSegmentsInNodeLoadingQueue`, `useRoundRobinSegmentAssignment`,
`emitBalancingStats` and `replicantLifetime`.
- Coordinator reports richer metrics and produces cleaner and more informative logs
- Coordinator uses an unlimited load queue for all serves, and makes better assignment decisions
Introduce DruidException, an exception whose goal in life is to be delivered to a user.
DruidException itself has javadoc on it to describe how it should be used. This commit both introduces the Exception and adjusts some of the places that are generating exceptions to generate DruidException objects instead, as a way to show how the Exception should be used.
This work was a 3rd iteration on top of work that was started by Paul Rogers. I don't know if his name will survive the squash-and-merge, so I'm calling it out here and thanking him for starting on this.
Description:
Druid allows a configuration of load rules that may cause a used segment to not be loaded
on any historical. This status is not tracked in the sys.segments table on the broker, which
makes it difficult to determine if the unavailability of a segment is expected and if we should
not wait for it to be loaded on a server after ingestion has finished.
Changes:
- Track replication factor in `SegmentReplicantLookup` during evaluation of load rules
- Update API `/druid/coordinator/v1metadata/segments` to return replication factor
- Add column `replication_factor` to the sys.segments virtual table and populate it in
`MetadataSegmentView`
- If this column is 0, the segment is not assigned to any historical and will not be loaded.
* Throw ValidationException if CLUSTERED BY column descending order is specified.
- Fails query planning
* Some more tests.
* fixup existing comment
* Update comment
* checkstyle fix: remove unused imports
* Remove InsertCannotOrderByDescendingFault and deprecate the fault in readme.
* move deprecated field to the bottom
They were not previously loaded because supportsQueries was false.
This patch sets supportsQueries to true, and clarifies in Task
javadocs that supportsQueries can be true for tasks that aren't
directly queryable over HTTP.
* Limit select results in MSQ
* reduce number of files in test
* add truncated flag
* avoid materializing select results to list, use iterable instead
* javadocs
* fix kafka input format reader schema discovery and partial schema discovery to actually work right, by re-using dimension filtering logic of MapInputRowParser
changes:
* auto columns no longer participate in generic 'null column' handling, this was a mistake to try to support and caused ingestion failures due to mismatched ColumnFormat, and will be replaced in the future with nested common format constant column functionality (not in this PR)
* fix bugs with auto columns which contain empty objects, empty arrays, or primitive types mixed with either of these empty constructs
* fix bug with bound filter when upper is null equivalent but is strict
Kashif Faraz
imply-cheddar
Clint Wylie
Laksh Singla
Abhishek Radhakrishnan
Nhi Pham
Katya Macedo
Gian Merlino
YongGang
Tejaswini Bandlamudi
Rishabh Singh
Peter Marshall
Adarsh Sanjeev
Abhishek Agarwal
Sergio Ferragut
Hardik Bajaj
Rohan Garg
George Shiqi Wu
Maytas Monsereenusorn
Pranav
Alexander Saydakov
Vadim Ogievetsky