Repository navigation
druid: fix historical Guice startup failure and load's COUNT(*) parsing - #2444
Open
KazukiKandaKK wants to merge 1 commit into
Open
KazukiKandaKK wants to merge 1 commit into
KazukiKandaKK wants to merge 1 commit into
Conversation
1. The historical process fails to start: install doubles
druid.processing.buffer.sizeBytes (500MiB -> 1000MiB) on
single-server-medium but leaves historical's
-XX:MaxDirectMemorySize at the stock 13g in single-server-medium's
jvm.config. Per Druid's own direct memory formula,
(numThreads + numMergeBuffers + 1) * buffer.sizeBytes, that's
20 * 1000MiB = 20,971,520,000 bytes (~19.53 GiB), which exceeds the
13g cap and throws a Guice ProvisionException before historical can
start. This is independent of dataset size. 20g (21,474,836,480
bytes) already clears that; this raises MaxDirectMemorySize to 21g
for a small margin.
2. load's COUNT(*) check parses the SQL response
[{"EXPR$0":<count>}] with `grep -oE '[0-9]+' | head -n1`, which matches
the 0 in the key name "EXPR$0" instead of the actual count. This makes
the check always see 0 and poll for its full 6h budget even when the
table is already queryable, failing with "hits datasource still not
queryable" afterward. load now parses the response as JSON.
Verified end-to-end on a fresh c6a.4xlarge: with both fixes applied,
the official benchmark.sh no longer aborts and runs through all 43
queries with the real 100M-row dataset. 12 of those return null
(8 pre-existing SQL-dialect limits; 4 fail only at 100M rows, not
addressed here); see follow-up.
KazukiKandaKK
had a problem deploying
to
benchmark-approval
October 7, 2026 23:06 — with 
GitHub Actions
Error
Contributor
|
The run of Logs:
|
Contributor
|
The run of Logs:
|
Contributor
|
The run of Logs:
|
Contributor
|
The run of Logs:
|
Contributor
|
The run of Logs:
|
Member
|
@KazukiKandaKK, it didn't work. |
This branch was successfully deployed
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
2 participants
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.
druid/currently fails to start onsingle-server-medium, andload's completion check never detects success.1.
installdoubles historicaldruid.processing.buffer.sizeBytesfrom 500MiB to 1000MiB but leaves-XX:MaxDirectMemorySizeat the stock 13g. Druid requires(numThreads + numMergeBuffers + 1) * buffer.sizeBytes; for this profile (numThreads=15,numMergeBuffers=4) that is 20 × 1000MiB ≈ 19.53 GiB, which exceeds 13g and fails Guice provisioning. RaisedMaxDirectMemorySizeto 21g.2. Druid returns
[{"EXPR$0":<count>}]. The check usedgrep -oE '[0-9]+' | head -n1, which matches the0inEXPR$0, socntis always 0 and the loop polls for 6h then exits with "hits datasource still not queryable".loadnow parses the response as JSON.The run completed all 43 queries on a fresh
c6a.4xlargewith the real 100M-row dataset. 12/43 still returned[null,null,null]: 8 are a pre-existing SQL-dialect limitation independent of this PR; 4 fail only at 100M rows. Out of scope here.