Whenever an RPC write to an upstream ingester fails, it is retried after
an increasing delay, until the RPC_TIMEOUT is hit. Because of this, any
RPC write error would be returned as a "timeout", masking the underling
reason the write actually failed.
This commit pushes down the timeout logic, and retains the most recently
observed RPC write error, returning it to the user instead of the
timeout error.
This PR changes the RpcWrite DmlHandler to facilitate testing using
mocked RPC client circuit breaker states, and assert the error response
for no healthy upstreams.
This simplifies the semantics of upstream RPC endpoint snapshots; an
empty snapshot will never be returned, and therefore if the caller has a
snapshot, it contains at least one endpoint to attempt an RPC call
against.
This moves the "no upstream" error handling from a write-time error, to
a pre-write error.
* feat: Display failed query
Allows a user to immediately identify the failed query.
* feat: API improvements to InfluxQL parser
* feat: Extend `SchemaProvider` trait to query for UDFs
* fix: We don't want the parser to panic on overflows
* fix: ensure `map_type` maps the timestamp data type
* feat: API to map a InfluxQL duration expression to a DataFusion interval
* chore: Copied APIs from DataFusion SQL planner
These APIs are private but useful for InfluxQL planning.
* feat: Initial aggregate query support
* feat: Add an API to fetch a field by name
* chore: Fixes to handling NULLs in aggregates
* chore: Add ability to test expected failures for InfluxQL
* chore: appease rustfmt and clippy 😬
* chore: produce same error as InfluxQL
* chore: appease clippy
* chore: Improve docs
* chore: Simplify aggregate and raw planning
* feat: Add support for GROUP BY TIME(stride, offset)
* chore: Update docs
* chore: remove redundant `is_empty` check
Co-authored-by: Christopher M. Wolff <chris.wolff@influxdata.com>
* chore: PR feedback to clarify purpose of function
* chore: The series_sort can't be empty, as `time` is always added
This was originally intended as an optimisation when executing an
aggregate query that did not group by time or tags, as it will produce
N rows, where N is the number of measurements queried.
* chore: update comment for clarity
---------
Co-authored-by: Christopher M. Wolff <chris.wolff@influxdata.com>
* chore(iox/compactor): add test cases tracking max_l0_created_at
* chore(iox/compactor): add invariant check to simulator for misordered max_l0_created_at
* chore(iox/compactor): test updates for max_l0_created_at
This commit updates the newly added tests to comply with the new invariant check.
Several test cases are deleted because the changes to comply with the invariant make the permutations pointless.
* chore(iox/simulator): Add test cases hitting new invariant check for illegal max_l0_created_at
* chore: update previous tests to comply with new invariant
* chore: address comments
---------
Co-authored-by: kodiakhq[bot] <49736102+kodiakhq[bot]@users.noreply.github.com>
Making this clone more explicit, higher up. Deleting the old files from
the catalog might only need the IDs, and the compaction levels for the
metrics, but that can be later.
* feat: add optional param to GetTables
* chore: add the third param to query plan
* feat: add table_types param
* chore: clippy
* test: add test cases with filters
* chore: update query to avoid SQL injection
* refactor: update where clause and cleanup
---------
Co-authored-by: kodiakhq[bot] <49736102+kodiakhq[bot]@users.noreply.github.com>
And define the conditions for that decision on a type responsible for
the data that goes into that condition.
Also start moving away from methods that only work on one variant of
FilesToSplitOrCompact, and add a state for when FilesToSplitOrCompact
doesn't actually contain any files.
I can't ever remember whether it's "compact or split" or "split or
compact", so now I think it's always "split or compact".
Also remove "FilesTo" from the enum variants because "FilesTo" is in the
overall enum name.
Dom
kodiakhq[bot]
Nga Tran
Carol (Nichols || Goulding)
Andrew Lamb
Fraser Savage
dependabot[bot]
CircleCI[bot]
Stuart Carnie
Joe-Blount
Chunchun Ye