fix: set allowExitOnIdle on source pool to prevent hang#62
Merged
Conversation
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
Found New Recommended Indexes
The following indexes are likely to make the most impact to your queries. They're ordered by how many queries seen in your tests
| Index Definition | Usage Count |
|---|---|
assets(event_id, uploader_id, inserted_at desc) |
1 |
guest_ip_addresses(ip_address) |
1 |
Statistics Mode
When generating recommendations, we made the following changes to your database statistics:
fixed rows per table
- Rows per table: 10000
- Pages per table: 1
Instead of assuming a fixed number of rows per table (which can cause unnecessary recommendations), you can export statistics from your production database and import it using the STATISTICS_PATH environment variable. You can read more about how to sync stats here.
- name: Analyze
uses: query-doctor/analyzer@v0
env:
STATISTICS_PATH: ./statistics.jsonOptimization Overview
| Query | Base Cost | Optimized Cost | Improvement |
|---|---|---|---|
| 3527900520088515000 | 15922.03 | 1638.53 | 9.72x |
| 7119592456364862000 | 126 | 9.04 | 13.94x |
Query 3527900520088515000
New indexes improve cost by 9.72x:assets(event_id, uploader_id, inserted_at desc)
View Query (too long to display inline)
SELECT
"guests"."id",
"guests"."session_id",
"guests"."username",
"guests"."avatar_path",
"guests"."color",
"guests"."side",
"guests"."audio_recording_path",
"guests"."audio_recording_public",
"guests"."memo",
"guests"."memo_public",
"guests"."setup_at",
"guests"."last_upload",
"guests"."inserted_at",
"guests"."updated_at",
"userAssets"."id",
"userAssets"."kind",
"userAssets"."event_id",
"userAssets"."uploader_id",
"userAssets"."uploader_ip",
"userAssets"."path",
"userAssets"."file_size",
"userAssets"."width",
"userAssets"."height",
"userAssets"."visible_at",
"userAssets"."deleted_at",
"userAssets"."inserted_at",
"userAssets"."updated_at"
FROM
(
SELECT
"id",
"session_id",
"username",
"avatar_path",
"color",
"side",
"audio_recording_path",
"audio_recording_public",
"memo",
"memo_public",
"setup_at",
"last_upload",
"inserted_at",
"updated_at"
FROM
"guests"
ORDER BY
"guests"."last_upload" DESC,
"guests"."id" DESC
LIMIT
100
) "guests"
CROSS JOIN LATERAL (
SELECT
"id",
"kind",
"event_id",
"uploader_id",
"uploader_ip",
"path",
"file_size",
"width",
"height",
"visible_at",
"deleted_at",
"inserted_at",
"updated_at"
FROM
"assets"
WHERE
(
"assets"."event_id" = (
SELECT
"id"
FROM
"events"
WHERE
"events"."event_key" = '01JKCVP4M2CH34SVTQGHSW4Y5G'
)
AND "assets"."uploader_id" = "guests"."id"
)
ORDER BY
"assets"."inserted_at" DESC
LIMIT
100
) "userAssets";
View Explain Plan (before optimization)
{
"Node Type": "Nested Loop",
"Parallel Aware": false,
"Async Capable": false,
"Join Type": "Inner",
"Startup Cost": 159.35,
"Total Cost": 15922.03,
"Plan Rows": 100,
"Plan Width": 370,
"Inner Unique": false,
"Plans": [
{
"Node Type": "Limit",
"Parent Relationship": "Outer",
"Parallel Aware": false,
"Async Capable": false,
"Startup Cost": 0.16,
"Total Cost": 1.78,
"Plan Rows": 100,
"Plan Width": 238,
"Plans": [
{
"Node Type": "Index Scan",
"Parent Relationship": "Outer",
"Parallel Aware": false,
"Async Capable": false,
"Scan Direction": "Forward",
"Index Name": "guests_last_upload_desc_id_desc_index",
"Relation Name": "guests",
"Alias": "guests",
"Startup Cost": 0.16,
"Total Cost": 162.16,
"Plan Rows": 10000,
"Plan Width": 238
}
]
},
{
"Node Type": "Limit",
"Parent Relationship": "Inner",
"Parallel Aware": false,
"Async Capable": false,
"Startup Cost": 159.19,
"Total Cost": 159.19,
"Plan Rows": 1,
"Plan Width": 132,
"Plans": [
{
"Node Type": "Index Scan",
"Parent Relationship": "InitPlan",
"Subplan Name": "InitPlan 1 (returns $0)",
"Parallel Aware": false,
"Async Capable": false,
"Scan Direction": "Forward",
"Index Name": "events_event_key_index",
"Relation Name": "events",
"Alias": "events",
"Startup Cost": 0.16,
"Total Cost": 8.18,
"Plan Rows": 1,
"Plan Width": 8,
"Index Cond": "(event_key = '01JKCVP4M2CH34SVTQGHSW4Y5G'::text)"
},
{
"Node Type": "Sort",
"Parent Relationship": "Outer",
"Parallel Aware": false,
"Async Capable": false,
"Startup Cost": 151.01,
"Total Cost": 151.01,
"Plan Rows": 1,
"Plan Width": 132,
"Sort Key": [
"assets.inserted_at DESC"
],
"Plans": [
{
"Node Type": "Seq Scan",
"Parent Relationship": "Outer",
"Parallel Aware": false,
"Async Capable": false,
"Relation Name": "assets",
"Alias": "assets",
"Startup Cost": 0,
"Total Cost": 151,
"Plan Rows": 1,
"Plan Width": 132,
"Filter": "((event_id = $0) AND (uploader_id = guests.id))"
}
]
}
]
}
]
}View Explain Plan (after optimization)
{
"Node Type": "Nested Loop",
"Parallel Aware": false,
"Async Capable": false,
"Join Type": "Inner",
"Startup Cost": 8.5,
"Total Cost": 1638.53,
"Plan Rows": 100,
"Plan Width": 370,
"Inner Unique": false,
"Plans": [
{
"Node Type": "Limit",
"Parent Relationship": "Outer",
"Parallel Aware": false,
"Async Capable": false,
"Startup Cost": 0.16,
"Total Cost": 1.78,
"Plan Rows": 100,
"Plan Width": 238,
"Plans": [
{
"Node Type": "Index Scan",
"Parent Relationship": "Outer",
"Parallel Aware": false,
"Async Capable": false,
"Scan Direction": "Forward",
"Index Name": "guests_last_upload_desc_id_desc_index",
"Relation Name": "guests",
"Alias": "guests",
"Startup Cost": 0.16,
"Total Cost": 162.16,
"Plan Rows": 10000,
"Plan Width": 238
}
]
},
{
"Node Type": "Limit",
"Parent Relationship": "Inner",
"Parallel Aware": false,
"Async Capable": false,
"Startup Cost": 8.34,
"Total Cost": 16.36,
"Plan Rows": 1,
"Plan Width": 132,
"Plans": [
{
"Node Type": "Index Scan",
"Parent Relationship": "InitPlan",
"Subplan Name": "InitPlan 1 (returns $0)",
"Parallel Aware": false,
"Async Capable": false,
"Scan Direction": "Forward",
"Index Name": "events_event_key_index",
"Relation Name": "events",
"Alias": "events",
"Startup Cost": 0.16,
"Total Cost": 8.18,
"Plan Rows": 1,
"Plan Width": 8,
"Index Cond": "(event_key = '01JKCVP4M2CH34SVTQGHSW4Y5G'::text)"
},
{
"Node Type": "Index Scan",
"Parent Relationship": "Outer",
"Parallel Aware": false,
"Async Capable": false,
"Scan Direction": "Forward",
"Index Name": "assets(event_id, uploader_id, inserted_at desc)",
"Relation Name": "assets",
"Alias": "assets",
"Startup Cost": 0.16,
"Total Cost": 8.18,
"Plan Rows": 1,
"Plan Width": 132,
"Index Cond": "((event_id = $0) AND (uploader_id = guests.id))"
}
]
}
]
}Query 7119592456364862000
New indexes improve cost by 13.94x:guest_ip_addresses(ip_address)
SELECT
*
FROM
guest_ip_addresses
WHERE
ip_address = '127.0.0.1';
View Explain Plan (before optimization)
{
"Node Type": "Seq Scan",
"Parallel Aware": false,
"Async Capable": false,
"Relation Name": "guest_ip_addresses",
"Alias": "guest_ip_addresses",
"Startup Cost": 0,
"Total Cost": 126,
"Plan Rows": 50,
"Plan Width": 56,
"Filter": "(ip_address = '127.0.0.1'::text)"
}View Explain Plan (after optimization)
{
"Node Type": "Index Scan",
"Parallel Aware": false,
"Async Capable": false,
"Scan Direction": "Forward",
"Index Name": "guest_ip_addresses(ip_address)",
"Relation Name": "guest_ip_addresses",
"Alias": "guest_ip_addresses",
"Startup Cost": 0.16,
"Total Cost": 9.04,
"Plan Rows": 50,
"Plan Width": 56,
"Index Cond": "(ip_address = '127.0.0.1'::text)"
}What are the numbers next to the query?
The numbers are a fingerprint uniquely identifying the query. Let us know in the Discord if you'd like to be able to assign unique names to your queries.What is cost?
Cost is an arbitrary value representing the amount of work postgres decided it needs to do to execute a query based on what it knows about the database and the query itself.We use cost to look for improvements when checking if an index helps optimize a query in CI as the full production dataset is simply not available to work with.
Execution metadata
- Log size
- 212123 bytes
- Time elapsed
- 2245ms
- Queries Seen
- 28
- Queries matched
- 10
- Queries optimized
- 2
- Queries errored
- 1
pg-pool keeps idle connections and their timers ref'd by default, which prevents Node from exiting even after pool.end() resolves. The underlying TCP sockets continue draining for ~15s (matching idleTimeoutMillis), keeping the event loop alive. allowExitOnIdle unrefs idle connections so they don't block exit. Adds integration tests that spawn child processes to verify: - Idle connections don't block process exit - Explicit pool close exits 0, not SIGPIPE (13) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
veksen
force-pushed
the
veksen/hat-yai-v1
branch
from
77effda to
5327deb
Compare
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.
Summary
process.exit()revealed the sourcepg.Poolkeeps the event loop alive via ref'd idle connections and timersDEFAULT_IDLE_TIMEOUT_MS = 15_000exactly:pool.end()resolves but the underlying TCP sockets continue draining, keeping the event loop aliveallowExitOnIdle: trueon the source pool so idle connections areunref()'d and don't block Node from exitingTest plan
🤖 Generated with Claude Code