Skip to content

Fixes TopologyConstraints for scaled PodGangs#340

Merged
unmarshall merged 3 commits into
ai-dynamo:mainfrom
unmarshall:podgang-tas-fix
Jan 16, 2026
Merged

Fixes TopologyConstraints for scaled PodGangs#340
unmarshall merged 3 commits into
ai-dynamo:mainfrom
unmarshall:podgang-tas-fix

Conversation

@unmarshall

@unmarshall unmarshall commented Jan 16, 2026

Copy link
Copy Markdown
Collaborator

What type of PR is this?

/kind bug

What this PR does / why we need it:

  • Corrects godoc for PodGang to clearly specify the meaning of TopologyConstraint and TopologyConstraintGroupConfigs fields in PodGangSpec.
  • Fixes computation of expected PodGangs w.r.t topology constraints.
  • Adds unit test for computeExpectedPodGangs for different combinations.

Which issue(s) this PR fixes:

Fixes #339

Special notes for your reviewer:

Does this PR introduce a API change?

There is no change to scheduler `PodGang` API but interpretation of `PodGangSpec.TopologyConstraint` changes when it is set on a base or scaled pod gang.

Additional documentation e.g., enhancement proposals, usage docs, etc.:


  TopologyConstraint and TopologyConstraintGroupConfigs fields in
  PodGangSpec.
* Fixed computation of expected PodGangs w.r.t topology constraints.
* Added unit test for computeExpectedPodGangs for different combinations.

Signed-off-by: Madhav Bhargava <madhav.bhargava@sap.com>
@unmarshall unmarshall changed the title * Corrected godoc for PodGang to clearly specify the meaning of Fixed TopologyConstraints for scaled pod gang Jan 16, 2026
@unmarshall unmarshall changed the title Fixed TopologyConstraints for scaled pod gang Fixes TopologyConstraints for scaled PodGangs Jan 16, 2026
Optimized on conditions in syncflow when computing expected podgangs

Signed-off-by: Madhav Bhargava <madhav.bhargava@sap.com>
Signed-off-by: Madhav Bhargava <madhav.bhargava@sap.com>
@unmarshall unmarshall merged commit 095d529 into ai-dynamo:main Jan 16, 2026
6 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

TopologyConstraint for a scaled PodGang should have the PCSG topology constraint and not the PCS topology constraint

3 participants