* Move all intrinsic tag lookup to the query-frontend and prioritize them in the results
* lint, error handling, fix tests
* Fix some tests
* Revert unintended change to search/tags v1 behavior, update tests
* Reduce diff
* Revert unintended change to 'none' scope
* reduce diff
* changelog
* Update test to test intrinsic handling at the limit
* todos
* make block metas for search take a time.Time to force the caller to interpret their time correctly
Signed-off-by: Joe Elliott <number101010@gmail.com>
* Applied the same fix to cache key
Signed-off-by: Joe Elliott <number101010@gmail.com>
* linter i love you, but you you're bringing me down
Signed-off-by: Joe Elliott <number101010@gmail.com>
* maybe
Signed-off-by: Joe Elliott <number101010@gmail.com>
* only calculate start/end time once in tag sharder
Signed-off-by: Joe Elliott <number101010@gmail.com>
* push time.Time out one level
Signed-off-by: Joe Elliott <number101010@gmail.com>
---------
Signed-off-by: Joe Elliott <number101010@gmail.com>
docker / manifest (tempo) (push) Blocked by required conditions
docker / manifest (tempo-cli) (push) Blocked by required conditions
docker / manifest (tempo-query) (push) Blocked by required conditions
docker / manifest (tempo-vulture) (push) Blocked by required conditions
docker / cd-to-dev-env (push) Blocked by required conditions
Snyk Monitor Scanning / snyk-scan-ci (push) Waiting to run
* fix cache
Signed-off-by: Joe Elliott <number101010@gmail.com>
* changelog
Signed-off-by: Joe Elliott <number101010@gmail.com>
* remove the thing that lint doesn't like
Signed-off-by: Joe Elliott <number101010@gmail.com>
* fix changelog
Signed-off-by: Joe Elliott <number101010@gmail.com>
---------
Signed-off-by: Joe Elliott <number101010@gmail.com>
* Consolidate block builder on common lag_seconds metric, clean out metrics for revoked partitions, cleanup
* Cleanup, lint
* utc
* Reuse existing test data functions
* e2e tests: min_ready_duration set to 1
It reduces time for ingester to be ready
That reduces the slowest step in e2e tests
from 17 seconds to a couple of seconds
* Refactoring: Move block flush sleep time to global
* e2e test: Reduce complete_block_timeout
* e2e tests: reduce block flush wait time
* Parallelise slowest unit tests
* Improve the slowest test by making it less heavy
* Parallelise slow e2e test
To achieve that, each test will have
its own unique network name
* Parallelise slow unit test
* Revert "Parallelise slow unit test"
This reverts commit b8266a642ba7fa9f620b2f81d00363c92ab14130.
The optimisation does not have effect on CI
* this is garbage
Signed-off-by: Joe Elliott <number101010@gmail.com>
* filtery stuff
Signed-off-by: Joe Elliott <number101010@gmail.com>
* fix
Signed-off-by: Joe Elliott <number101010@gmail.com>
* max def everywhere
Signed-off-by: Joe Elliott <number101010@gmail.com>
* clean up benches
Signed-off-by: Joe Elliott <number101010@gmail.com>
* clean up
Signed-off-by: Joe Elliott <number101010@gmail.com>
* remove vendor chagnes
Signed-off-by: Joe Elliott <number101010@gmail.com>
* changelog
Signed-off-by: Joe Elliott <number101010@gmail.com>
* add details about bench env vars
Signed-off-by: Joe Elliott <number101010@gmail.com>
---------
Signed-off-by: Joe Elliott <number101010@gmail.com>
* [Metrics Generator] Allow running on a different source of data
* fix lint errors
* Address review feedback
* regenerate manifest
* do not start ring lifecycler
* make partition ring watcher a separate module
* do not join ring when gRPC is disabled
* changelog
* Set CPU architecture API version for amd64 and arm64
This enables the compiler to use instructions from more recent API versions
* Strip debug symbols from production binary
Reduces Tempo binary size by ~20%
* Remove GO111MODULE=on environment variable
GO111MODULE defaults to 'on' since Go 1.16
* CHANGELOG.md
* rythm: fail safe when no partitions are assigned to a blockbuilder instance
* returns an error if no partitions assigned
* include also pending partitions
* improve test to include another test case
* return the wait time when no partitions assigned
* remove not needed assertion
* Add new concurrency config options and validate
* Reduce allocations of IDmap when replaying wal blocks since we know the number of traces
* Generator read from kafka concurrency, add shared ingest lag metric, option to uniqify strings
* Memoize spanmetrics sanitizelabelname, move to better location
* Moved to shared queue for localblocks wal completion, allow concurrency within a tenant
* Honor max live traces in non-flushing local blocks processor
* Add metric for enqueue time
* Add missing mutex lock for enqueuing on replay
* Fix mutex lock while reloading blocks
* Increase default concurrency
* Simplify local blocks complete queue
* Remove uniqify, fix test
* Switch completequeue to reference counting and shut it down when last local blocks processor is stopped
* cache rename/cleanup
* Cleanup/denoising PR
* Lint/cleanup
* lint
* Update config manifest
* fix race condition
* cleanup
* review feedback
* Update config manifest
* changelog
* rhythm: weighted partition consumption
* remove printf
* make it simpler
* better logging
* fix typo
* update the loop to priorize the laggiest partition based in time not in offset
* remove log line
* address comments
* fix bug and remove test statements
* changelog
* remove unused variable
* Implement span basic rebatching and 4to4 converter
Signed-off-by: Joe Elliott <number101010@gmail.com>
* Generated trace data with unique scope and resource spans
The functions test.MakeBatch and test.MakeTrace* now create
more unique ResourceSpans and ScopeSpans. This avoids unwanted
rebatching in tests
* Fix tests in spanmetrics processor
Test data now has additional attributes on resource level
that caused these tests to fail
* Move rebatch code out of nested set function
* Make parquetTraceToTempopbTrace() private
* Add benchmark for testing traceToParquet()
* make check-fmt
* rebatchTrace() without allocating new slices for ResourceSpans and ScopeSpans
* Fix bug with duplicated rs ans ss in rebatchTrace()
* CHANGELOG.md
* Add finalizeTrace() function
* Call finalizeTrace() in combiner
---------
Signed-off-by: Joe Elliott <number101010@gmail.com>
Co-authored-by: Joe Elliott <number101010@gmail.com>
