[mssonicbld]

Why I did it

database_config.json is read concurrently during database startup. PR
#28343 made the initial publication
atomic, but three later publication paths remained non-atomic and could truncate a live JSON file
before writing its replacement. This PR extends atomic publication to those remaining known paths.

Instrumentation on DUT measured a 285 us truncate-and-rewrite window in the live-file
update_chassisdb_config path. The controlled 202605 A/B test reproduced the empty-read failure
with the installed writer and eliminated it with the atomic writer.

Work item tracking

• Microsoft ADO (number only): N/A

How I did it

• Copy /etc/sonic/database_config*.json to a sibling temporary file and rename it onto the live
  path.
• Rename the already-generated modified configuration onto the live path instead of copying over
  it, and tolerate the consumed temporary file during cleanup.
• Make update_chassisdb_config write a sibling temporary file and publish it with
  os.replace().

How to verify it

Run concurrent JSON readers while repeatedly invoking the installed in-place writer and the
patched atomic writer against the same scratch configuration. Use identical writer iterations and
reader counts for both arms, and verify that the live database configuration remains untouched.

Which release branch to backport (provide reason below if selected)

• 202305
• 202311
• 202405
• 202411
• 202505
• 202511
• 202512
• 202605
• 202608

Tracking issue/work item for backport/cherry-pick request (GitHub issue or Microsoft ADO):
N/A (internal tracking is not disclosed)

Failure type: other — existing non-atomic database configuration publication race

Tested branch

• master
• 202305
• 202311
• 202405
• 202411
• 202505
• 202511
• 202512
• 202605
• 202608
• N/A

Test result

• master: — HEAD was exercised in three patched boots as part
  of a broader startup-path test and produced zero parser errors. Because that arm contained
  additional startup-path changes, it is integration coverage rather than isolated proof for this
  PR. Restoring the control produced three errors.
• 202605: — controlled A/B with 150 writes and four concurrent readers per arm:
  • Installed in-place writer: 118,554 successful reads and 701 empty reads.
  • Patched temp-file plus os.replace() writer: 132,972 successful reads and zero empty reads.
  • The experiment used a scratch copy; the live database configuration was not modified.

Description for the changelog

Publish database_config.json atomically to prevent readers from observing empty JSON during
normal updates.

Signed-off-by: Sonic Build Admin sonicbld@microsoft.com

[mssonicbld]

Original PR: #29464

[azure-pipelines]

Azure Pipelines:
There may be pipelines that require an authorized user to comment /azp run to run.

[mssonicbld]

/azp run Azure.sonic-buildimage

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).

[chartsai-nvidia]

/azpw retry

[mssonicbld]

Retrying failed(or canceled) jobs...

[mssonicbld]

Retrying failed(or canceled) stages in build 1227190:

✅Stage Test:

• Job impacted-area-kvmtest-t1-lag-vpp by Elastictest: retried.