CNDB-11680: Add source sstable/memtable id to vector traces #1411
Conversation
michaeljmarshall
force-pushed
the
more-tracing-info
branch
from
1d415df
to
cd8926a
Compare
michaeljmarshall
changed the title
|
VectorDistributedTest.rangeRestrictedTest passes locally |
There was a problem hiding this comment.
I think this is really useful, but as a nit, I don't love the output for memtables.
For an sstable, we see the SSTable ID, which is ULID-based in the common case, e.g.
DiskANN resume for 1/5 visited 0 nodes, reranked 0 to return 0 results from 3gl2_1sri_5vjqo2nd1q22mk2235.
For the memtables, we rely on the memtable toString implementation, which can contain lots of other information:
ANN search for 1/5 visited 1 nodes, reranked 0 to return 1 results from Memtable-tbl@1033435943(52B serialized bytes, 1 ops, 6.230KiB (0%) on-heap, 101B (0%) off-heap)
Maybe for memtable, we could just use the hashCode, as in the toString implementation.
Here are two samples of the trace at this point in commit history "ANN resume for 2/9 visited 0 nodes, reranked 0 to return 0 results from TrieMemtable@acb5508" "DiskANN search for 4/9 visited 4 nodes, reranked 0 to return 4 results from 3gl7_1sv4_58m4g2ch3lawuilnd5"
|
❌ Build ds-cassandra-pr-gate/PR-1411 rejected by Butler6 new test failure(s) in 4 builds Found 6 new test failures
Found 54 known test failures |
What is the issue
Fixes: https://github.com/riptano/cndb/issues/11680
When we resume the search for a vector query, it'd be helpful to know which sstable we're inspecting.
What does this PR fix and why was it fixed
Add the sstable/memtable to the log.
Checklist before you submit for review
NoSpamLoggerfor log lines that may appear frequently in the logs