fix: propagate numCores and extraCores from wekacluster to wekacontainers per role (OP-260)

[kristina-solovyova]

No description provided.

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Snapshot Warnings

⚠️: No snapshots were found for the head SHA ff0a024.

Ensure that dependencies are being submitted on PR branches and consider enabling retry-on-snapshot-warnings. See the documentation for more information and troubleshooting advice.

Scanned Files

None

[kristina-solovyova]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• fix: align container memory to per-core 2MiB hugepage boundary (OP-283) #2444
• fix: propagate numCores and extraCores from wekacluster to wekacontainers per role (OP-260) #2395 👈 (View in Graphite)
• fix: ensure hugepages spec hash reflects DPDK base memory values for correct change detection #2439
• main

---

How to use the Graphite Merge Queue

Add the label main-merge-queue to this PR to add it to the merge queue.

You must have a Graphite account in order to use the merge queue. Sign up using this link.

An organization admin has required the Graphite Merge Queue in this repository.

Please do not merge from GitHub as this will restart CI on PRs being processed by the merge queue.

This stack of pull requests is managed by Graphite. Learn more about stacking.

[claude]

Claude finished @kristina-solovyova's task in 1m 35s —— View job

---

Code Review

• Gather context (parent PR diff, changed files)
• Review the changed file
• Post review feedback

---

Summary

This PR does two things: (1) fixes a real bug where NumCores was never propagated from WekaCluster to WekaContainer specs, and (2) refactors the repeated switch role statements into a unified forRole() helper. The changes are clean and well-motivated.

---

Observations

Bug fix: NumCores propagation (funcs_upgrade.go:399)

NumCores was simply absent from HandleSpecUpdates. The fix is correct — GetNonZeroOrDefault preserves existing container values when the cluster spec has 0 (unset), which is the right semantic.

container.Spec.NumCores = util.GetNonZeroOrDefault(rv.NumCores, container.Spec.NumCores)

One implicit design decision worth documenting: once NumCores is set on a container, it can never be reduced to 0 via the cluster spec (since 0 is treated as "not set"). This appears intentional but is worth a comment if not already covered elsewhere.

---

Overrides refactor (funcs_upgrade.go:296–346)

The old pattern fetched overrides multiple times and conditionally wrote back. The new pattern fetches once, mutates, and unconditionally writes back at the end:

overrides := container.Spec.GetOverrides()
// ... modify fields ...
container.Spec.Overrides = overrides  // always assigned

This relies on client.MergeFrom to suppress no-op API calls, which is correct and idiomatic. No functional regression.

One subtle behavior change: UpgradePreventEviction on compute containers is now always written (even when value didn't change). This is fine since the patch mechanism handles it, but the old code's explicit guards made the intent clearer.

---

forRole() helper (funcs_upgrade.go:45–61)

Clean consolidation. The fallthrough to return containerRoleSpec{} for unknown roles is correct — unknown roles will get zero values and GetNonZeroOrDefault will preserve existing container fields.

---

dynamicTemplate nil-guard removal (funcs_upgrade.go:127–144)

The old code had a manual nil check:

// old
if spec.Dynamic != nil {
    computeExtraCores = spec.Dynamic.ComputeExtraCores
}

The new code reads from dynamicTemplate, which is always non-nil (falls back to the default template from GetWekaClusterTemplate). When spec.Dynamic == nil, the default template returns 0 for ExtraCores fields, so behavior is preserved.

---

Potential issue: SmbwHugepages default uses tmpl.Cores.Smbw (funcs_upgrade.go:224)

SmbwHugepages: util.GetNonZeroOrDefault(dynamicTemplate.SmbwFrontendHugepages, 1400*tmpl.Cores.Smbw),

Note tmpl (the default template from GetWekaClusterTemplate(spec.Dynamic)) is used rather than dynamicTemplate for the Smbw core count. This is intentional — you want the core count from the resolved template, not the raw config — but it's subtle. If spec.Dynamic specifies SmbwCores but SmbwFrontendHugepages is unset, the default is computed from the resolved core count, which is correct. No issue here, just worth noting.

---

Verdict

The refactoring is solid — no regressions visible, the NumCores propagation fix is the key value, and Smbw support is properly wired in throughout. The PR is ready to merge after the stack is resolved.

[graphite-app]

Graphite Automations

"Add anton/matt/sergey/kristina as reviwers on operator PRs" took an action on this PR • (03/25/26)

2 reviewers were added to this PR based on Anton Bykov's automation.

[claude]

Claude finished @kristina-solovyova's task in 31s —— View job

---

No changes since the last review — the PR still contains the same single commit (716e1611). See the previous review above for full feedback.

[claude]

Claude finished @kristina-solovyova's task in 1m 39s —— View job

---

• Check for new changes since last review
• Review changes
• Post review feedback

---

Code Review (updated commit 97c60007)

The commit was amended since the last review. Key changes from the prior version:

---

New: explicit NumCores increase-only semantics (funcs_upgrade.go:399–402)

// NumCores can only be increased — decreasing requires manual intervention (container restart/reconfiguration).
if rv.NumCores > container.Spec.NumCores {
    container.Spec.NumCores = rv.NumCores
}

The previous version used GetNonZeroOrDefault (which would silently preserve existing values when cluster spec has 0). This version uses a > guard, which is more explicit and correct: it documents intent, allows reducing to a positive but lower value by hand, and still won't trigger on unset (0) values in the cluster spec. Good improvement.

---

New: ExtraCores vs NumCores semantic comment (funcs_upgrade.go:394–395)

// ExtraCores allows zero (explicit removal), so uses equality guard.
// NumCores and Hugepages treat zero as "not set" — only update when non-zero.

This directly addresses the implicit design decision flagged in the previous review. No changes needed.

---

New: Smbw fully wired throughout

SmbwCores, SmbwExtraCores, SmbwHugepages, SmbwHugepagesOffset are now added to UpdatableClusterSpec, populated from dynamicTemplate in NewUpdatableClusterSpec, and included in forRole(). The SmbwHugepages default (1400 * tmpl.Cores.Smbw) uses the resolved template core count, which is correct.

One edge case: if tmpl.Cores.Smbw == 0 (no Smbw cores configured), the default hugepages will be 0. GetNonZeroOrDefault will then fall back to the container's existing value, so no regression — but it's worth confirming this is intended when Smbw is not configured.

---

DriversBuildId comparison simplified (funcs_upgrade.go:410–412)

oldId, newId := container.Spec.DriversBuildId, updatableSpec.DriversBuildId
if (oldId == nil) != (newId == nil) || (oldId != nil && *oldId != *newId) {

Cleaner than the previous triple-condition. Logically equivalent.

---

Verdict

All issues from the previous review are addressed. The Smbw propagation is complete, NumCores semantics are correctly documented and implemented, and the refactoring is clean. Ready to merge once the stack is resolved.

[claude]

Claude finished @kristina-solovyova's task in 2m 7s —— View job

---

Code Review

• Check for new changes since last review
• Review parent PR fix: ensure hugepages spec hash reflects DPDK base memory values for correct change detection #2439 for context
• Review current diff
• Post review feedback

---

Summary

This synchronization event was triggered by a rebase onto the updated parent PR (#2439). The commit was force-pushed (amended from 97c60007 to ff0a0246), but the functional content is unchanged from the previous review.

No new code changes to review — the previous review (2026-03-25) covers this PR in full and concluded it was ready to merge.

---

Verified: No Regressions Against Updated Parent

Parent PR #2439 added hugepagesIsUserSet/hugepagesOffsetIsUserSet tracking in GetContainerHugepages — user-set values now represent the total (weka + DPDK), so DPDK is NOT added on top. This PR propagates those computed hugepages values via rv.Hugepages (from UpdatableClusterSpec.*Hugepages) without adding any DPDK on top again. The two PRs compose correctly.

---

Verdict

Still ready to merge once the stack is resolved.

[graphite-app]

Merge activity

• Apr 11, 7:13 PM UTC: kristina-solovyova added this pull request to the Graphite merge queue.
• Apr 11, 7:14 PM UTC: CI is running for this pull request on a draft pull request (#2446) due to your merge queue CI optimization settings.
• Apr 11, 7:51 PM UTC: Merged by the Graphite merge queue via draft PR: #2446.