v2

latestOpenAPI 3.1.02026-08-07139158192.4 KB

Internal

Set one pool member's ranking weight (platform admin)

The per-replica warmup/canary dial: weight < 1 makes the member look fuller so it receives proportionally less traffic; 1 restores its full share. The member's hard capacity cap is untouched, and the reconcile loop preserves the weight across re-adds. Returns the refreshed replica view.

post/v1/admin/inference/deployments/{id}/member-weight

Path parameters

idstring required

Request body

hostnamestring required

The pool-member key (the replica's run.* hostname)

weightnumber double required

Ranking share in (0,1]; 1 restores full share

Response

Updated

affinityGuardUtilnumber double

prefix_affinity overflow bound; absent = the gateway default

policy'p2c_util' | 'prefix_affinity' required

The pool's member-selection policy (p2c_util when the pool row does not exist yet)

poolIdstring required

The deployment's endpoint pool id

strayMembersstring[]

Pool member hostnames no current replica owns (pending the reconcile loop's stray sweep). Usually empty.