Repository navigation
Delete operation index rows only for re-exported spans - #26
Closed
RhysSullivan wants to merge 18 commits into
Closed
RhysSullivan wants to merge 18 commits into
RhysSullivan wants to merge 18 commits into
Conversation
The query worker's RPC success channel is Schema.Unknown, whose wire
guard (SchemaAST isJson) rejects object graphs that visit the same
object twice. getTrace/getSpan results share span/event objects across
the tree, so /api/traces/{id}, /api/spans/{id}, /api/spans/search, and
the /trace/{id} page all failed with 'Expected JSON value'. Deep-copy
results in the worker to break the sharing, keeping Date instances
intact (msgpackr carries them natively and the HTTP pagination calls
getTime() on summaries).
Package renamed to @executor-js/motel (fork of @kitlangton/motel).
Run Motel on workerd
…nd per export Malformed exports get 400; store failures get 500 or 503, which exporters retry. Refusal counts persist in the collector's storage and include declared bytes.
Keep collector workers alive after request cancellation
Bound the workerd collector's memory, ingest queue and stored spans
…mory Test the workerd collector's memory, queue and retention bounds
Bind SQL lists as one JSON parameter
Test workerd exports larger than one statement's parameters
Author
|
Opened against the wrong repository by mistake; this belongs in the UsefulSoftwareCo fork. Sorry for the noise. Sent from my Claude |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Every trace export deleted its spans' rows from `span_operation_fts` before inserting them. That table names a span only in UNINDEXED columns, so each delete scanned the whole index: an export's cost grew with everything already stored.
On a 572 MB database from one Executor E2E run (134k spans), a one-span export took 35 ms instead of 2 ms, and 16 concurrent 20-span exports stored 353 spans/s instead of about 1,000. The collector's ingest queue filled and it refused thousands of exports per run.
Now ingest checks each span's primary key first and deletes index rows only for spans that were already stored. New spans, almost all of them, are only inserted. A test re-exports a span with the same and a changed name and checks the index holds one row for it.