perf(etcd): refactor some code to improve performance by xuruidong · Pull Request #12011 · apache/apisix

xuruidong · 2025-02-28T11:53:13Z

Description

Fixes # (issue)

Before:
The original code performed a cleanup of self.values and rebuilt self.values_hash by:

Made a full copy of self.values
Cleared the original and reinserted non-nil values
Rebuilt values_hash in a separate step

Problems:

Wasted memory (copied entire table)
Slow (processed data twice)

Now:

In-place cleanup:
- Removes nil values while keeping the same table
- Updates values_hash at the same time
Trims empty space:
- Deletes unused slots at the end

Why better?

Faster: Does everything in one pass
Uses less memory: No extra table copies
Works better for big datasets

Before:
The original code performed a cleanup of self.values and rebuilt self.values_hash by:

Made a full copy of self.values
Cleared the original and reinserted non-nil values
Rebuilt values_hash in a separate step

Problems:

Wasted memory (copied entire table)
Slow (processed data twice)

Now:

In-place cleanup:
- Removes nil values while keeping the same table
- Updates values_hash at the same time
Trims empty space:
- Deletes unused slots at the end

Why better?

Faster: Does everything in one pass
Uses less memory: No extra table copies
Works better for big datasets

Checklist

I have explained the need for this PR and the problem it solves
I have explained the changes or the new features added to this PR
I have added tests corresponding to this change
I have updated the documentation to reflect this change
I have verified that this change is backward compatible (If not, please discuss on the APISIX mailing list first)

Signed-off-by: xuruidong <xuruidong@gmail.com>

juzhiyuan · 2025-03-04T06:06:11Z

@xuruidong Just approved to run the CI

xuruidong · 2025-03-07T10:51:50Z

@xuruidong Just approved to run the CI

please run the CI again @juzhiyuan

juzhiyuan · 2025-03-10T03:33:37Z

@xuruidong Just approved to run the CI

please run the CI again @juzhiyuan

Hi @xuruidong, I couldn't find the Re-run button. Can you confirm the failed test cases?

xuruidong · 2025-03-17T03:06:14Z

I don't think the CI failure is related to the code changes.

Baoyuantop · 2025-03-26T08:35:51Z

Hi @xuruidong, can you add some descriptive information for PR?

Baoyuantop · 2025-04-28T14:55:50Z

Hi @xuruidong, please add a PR description so that others can review it better ~

xuruidong · 2025-05-10T13:36:36Z

Hi @xuruidong, please add a PR description so that others can review it better ~

Hi @Baoyuantop , the PR description has been added, please help review it. Thanks.

Baoyuantop · 2025-06-16T08:05:00Z

Please fix the failed ci

xuruidong · 2025-06-17T03:22:48Z

Please fix the failed ci

Hi @Baoyuantop ,please trigger the ci

xuruidong · 2025-06-19T03:34:21Z

Hi @Baoyuantop , the ci is not stable, the failed test should be rerun

Baoyuantop · 2025-07-21T10:04:38Z

Hi @xuruidong, there are still bad CIs that need to be fixed.

Baoyuantop · 2025-07-28T07:27:18Z

Hi @xuruidong, any updates?

xuruidong · 2025-07-29T02:27:51Z

Hi @xuruidong, any updates?

Hi @Baoyuantop ,please trigger the ci

github-actions · 2025-09-27T10:04:52Z

This pull request has been marked as stale due to 60 days of inactivity. It will be closed in 4 weeks if no further activity occurs. If you think that's incorrect or this pull request should instead be reviewed, please simply write any comment. Even if closed, you can still revive the PR at any time or discuss it on the dev@apisix.apache.org list. Thank you for your contributions.

Baoyuantop · 2025-09-29T02:30:18Z

Hi @xuruidong, are you still working on this pull request? There are some failed tests that need to be addressed.

xuruidong · 2025-10-10T09:47:11Z

Hi @xuruidong, are you still working on this pull request? There are some failed tests that need to be addressed.

Hi @Baoyuantop ，please trigger the CI

github-actions · 2025-12-09T10:07:04Z

This pull request has been marked as stale due to 60 days of inactivity. It will be closed in 4 weeks if no further activity occurs. If you think that's incorrect or this pull request should instead be reviewed, please simply write any comment. Even if closed, you can still revive the PR at any time or discuss it on the dev@apisix.apache.org list. Thank you for your contributions.

github-actions · 2026-02-22T10:07:09Z

This pull request has been marked as stale due to 60 days of inactivity. It will be closed in 4 weeks if no further activity occurs. If you think that's incorrect or this pull request should instead be reviewed, please simply write any comment. Even if closed, you can still revive the PR at any time or discuss it on the dev@apisix.apache.org list. Thank you for your contributions.

Copilot

Pull request overview

This pull request optimizes the array compaction logic in the etcd configuration synchronization code to improve memory usage and performance. The optimization targets the cleanup routine that runs after 100+ deletions to remove stale entries from the values array.

Changes:

Added table.remove as a localized function remove_tab following codebase conventions
Refactored the array compaction algorithm to perform in-place compaction instead of copy-clear-rebuild
Combined values_hash rebuilding with array compaction into a single pass

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-24T02:15:34Z

apisix/core/config_etcd.lua

+            local pre = 1
+            local cur = 1


The variable names pre and cur are unclear and could be more descriptive. Consider renaming to write_idx and read_idx respectively to better convey their purpose in the compaction algorithm. write_idx represents the next position to write a valid element, while read_idx tracks the current position being read from the original array.

Copilot · 2026-02-24T02:15:34Z

apisix/core/config_etcd.lua

+            local pre = 1
+            local cur = 1
+            table.clear(self.values_hash)
+            log.info("clear stale data in `values_hash` for key: ", key)
+            for _, val in ipairs(self.values) do
                if val then
-                    table.insert(self.values, val)
+                    self.values[pre] = val
+                    key = short_key(self, val.key)
+                    self.values_hash[key] = pre
+                    pre = pre + 1
                end
-            end

-            table.clear(self.values_hash)
-            log.info("clear stale data in `values_hash` for key: ", key)
+                cur = cur + 1
+            end

-            for i = 1, #self.values do
-                key = short_key(self, self.values[i].key)
-                self.values_hash[key] = i
+            for i = cur - 1, pre, -1 do
+                remove_tab(self.values, i)
            end


The refactored array compaction logic (lines 811-828) lacks test coverage. This is a critical code path that triggers after 100 deletions to clean up the values array. Consider adding a test that creates multiple items, deletes them to increment sync_times beyond 100, and verifies the array is correctly compacted with no data corruption.

Copilot · 2026-02-24T02:15:35Z

apisix/core/config_etcd.lua

+            local cur = 1
+            table.clear(self.values_hash)
+            log.info("clear stale data in `values_hash` for key: ", key)
+            for _, val in ipairs(self.values) do
                if val then
-                    table.insert(self.values, val)
+                    self.values[pre] = val
+                    key = short_key(self, val.key)
+                    self.values_hash[key] = pre
+                    pre = pre + 1
                end
-            end

-            table.clear(self.values_hash)
-            log.info("clear stale data in `values_hash` for key: ", key)
+                cur = cur + 1
+            end

-            for i = 1, #self.values do
-                key = short_key(self, self.values[i].key)
-                self.values_hash[key] = i
+            for i = cur - 1, pre, -1 do


The cur variable is only used to track the iteration count and could be replaced with #self.values captured before the loop. For example: local original_len = #self.values before line 815, then use original_len instead of cur - 1 in line 826. This would eliminate the need for incrementing cur on every iteration and make the code slightly clearer.

Copilot · 2026-02-24T02:15:35Z

apisix/core/config_etcd.lua

+                    key = short_key(self, val.key)
+                    self.values_hash[key] = pre


The variable key shadows the outer loop variable declared at line 726. While this doesn't cause incorrect behavior since key is reassigned at the start of each outer loop iteration, it's poor practice and reduces code clarity. Consider using a different variable name such as item_key or short_k for the inner usage.

Suggested change

key = short_key(self, val.key)

self.values_hash[key] = pre

local item_key = short_key(self, val.key)

self.values_hash[item_key] = pre

perf(etcd): refactor some code to improve performance

fccc82b

Signed-off-by: xuruidong <xuruidong@gmail.com>

dosubot bot added size:S This PR changes 10-29 lines, ignoring generated files. performance generate flamegraph for the current PR labels Feb 28, 2025

Baoyuantop added the wait for update wait for the author's response in this issue/PR label Apr 29, 2025

github-actions bot added the user responded label May 10, 2025

Baoyuantop removed wait for update wait for the author's response in this issue/PR user responded labels Jun 16, 2025

Merge branch 'master' into config-etcd

69c26ef

xuruidong force-pushed the config-etcd branch from 6eafd11 to 69c26ef Compare June 19, 2025 05:20

Merge branch 'master' into config-etcd

4cdd0ed

github-actions bot added the stale label Sep 27, 2025

github-actions bot removed the stale label Sep 29, 2025

Merge branch 'master' into config-etcd

b9971cd

github-actions bot added the stale label Dec 9, 2025

Baoyuantop requested a review from membphis December 17, 2025 08:14

github-actions bot removed the stale label Dec 17, 2025

Baoyuantop added the awaiting review label Dec 24, 2025

github-actions bot added the stale label Feb 22, 2026

Baoyuantop requested a review from Copilot February 24, 2026 02:07

Copilot started reviewing on behalf of Baoyuantop February 24, 2026 02:07 View session

Copilot AI reviewed Feb 24, 2026

View reviewed changes

Baoyuantop removed the awaiting review label Feb 24, 2026

github-actions bot removed the stale label Feb 24, 2026

Conversation

xuruidong commented Feb 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Checklist

Uh oh!

juzhiyuan commented Mar 4, 2025

Uh oh!

xuruidong commented Mar 7, 2025

Uh oh!

juzhiyuan commented Mar 10, 2025

Uh oh!

xuruidong commented Mar 17, 2025

Uh oh!

Baoyuantop commented Mar 26, 2025

Uh oh!

Baoyuantop commented Apr 28, 2025

Uh oh!

xuruidong commented May 10, 2025

Uh oh!

Baoyuantop commented Jun 16, 2025

Uh oh!

xuruidong commented Jun 17, 2025

Uh oh!

xuruidong commented Jun 19, 2025

Uh oh!

Baoyuantop commented Jul 21, 2025

Uh oh!

Baoyuantop commented Jul 28, 2025

Uh oh!

xuruidong commented Jul 29, 2025

Uh oh!

github-actions bot commented Sep 27, 2025

Uh oh!

Baoyuantop commented Sep 29, 2025

Uh oh!

xuruidong commented Oct 10, 2025

Uh oh!

github-actions bot commented Dec 9, 2025

Uh oh!

github-actions bot commented Feb 22, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

xuruidong commented Feb 28, 2025 •

edited

Loading