XS✔ ◾ feat: HA & Raft quorum safety guardrails by dimoschi · Pull Request #2 · hackthebox/typesense-helm

dimoschi · 2026-03-20T07:33:00Z

Summary

PDB enabled by default for multi-node clusters (replicaCount > 1), automatically skipped for single-replica deployments
Auto-calculated maxUnavailable for both PDB and StatefulSet updateStrategy based on Raft fault tolerance (floor(replicaCount/2)), with user override support and quorum validation
Explicit updateStrategy on StatefulSet with RollingUpdate default and OnDelete support, with schema-level enum and if/then/else conditional validation
Built-in soft pod anti-affinity (preferredDuringSchedulingIgnoredDuringExecution) on kubernetes.io/hostname for multi-node clusters
Template-level validation that rejects maxUnavailable values exceeding Raft fault tolerance and invalid updateStrategy.type values (defense in depth)
Remove hardcoded chart version from test suites
Bump chart version to 1.1.0

Quorum safety

replicaCount	Raft fault tolerance	Auto `maxUnavailable`
1	0 (no HA)	N/A (skipped)
3	1	1
5	2	2
7	3	3

Both pdb.maxUnavailable and updateStrategy.rollingUpdate.maxUnavailable default to "auto" (auto-calculated as floor(replicaCount/2)). Users can set an integer value: 0 to block all voluntary disruptions, or any positive value up to floor(replicaCount/2). The template fails with a descriptive error if the override exceeds the Raft fault tolerance.

Breaking changes

pdb.enabled default changed from false to true
pdb.maxUnavailable default changed from 1 to "auto" (auto-calculate as floor(replicaCount/2))
updateStrategy block added with type: RollingUpdate and rollingUpdate.maxUnavailable: "auto"
Built-in soft pod anti-affinity is now injected when affinity is unset or empty ({}) and replicaCount > 1

Schema validation

updateStrategy.type enforced as enum: RollingUpdate or OnDelete
maxUnavailable validated as anyOf: [integer >= 0, "auto"]
rollingUpdate conditionally required only when type is RollingUpdate (via if/then)
affinity typed as object

Test plan

91 helm-unittest tests pass (12 PDB + 48 StatefulSet + 31 others)
helm lint . passes
Schema and docs regenerated
CI passes on PR

…xUnavailable

…with quorum validation

…axUnavailable

… clusters

github-actions · 2026-03-20T07:33:11Z

PR Metrics

✔ Thanks for keeping your pull request small.
✔ Thanks for adding tests.

	Lines
Product Code	123
Test Code	252
Subtotal	375
Ignored Code	115
Total	490

Metrics computed by PR Metrics. Add it to your Azure DevOps and GitHub PRs!

… null in schema annotations

…ffinity default to {} maxUnavailable: 0 means auto-calculate as floor(replicaCount/2). This gives proper integer typing in schema and docs. affinity: {} enables correct object type inference by helm-docs.

…ateStrategy.type - Change maxUnavailable default from 0 to -1 so users can explicitly set 0 to block all voluntary disruptions - Add template validation for updateStrategy.type (must be RollingUpdate or OnDelete) - Fix changelog to reflect actual defaults (was incorrectly referencing null)

vlasopoulos

Just for discussion,
since we are in a flimsy dynamically typed template world that becomes strict at render (don't get me started), we could set a trend of values making sense on their own.

vlasopoulos · 2026-03-20T09:41:05Z

templates/pdb.yaml

+{{- $maxUnavailable := int .Values.pdb.maxUnavailable -}}
+{{- if lt $maxUnavailable 0 }}
+  {{- $maxUnavailable = $maxFault -}}
+{{- end }}


Suggested change

{{- $maxUnavailable := int .Values.pdb.maxUnavailable -}}

{{- if lt $maxUnavailable 0 }}

{{- $maxUnavailable = $maxFault -}}

{{- end }}

{{- $maxUnavailable := .Values.pdb.maxUnavailable }}

{{- if and (kindIs "string" .Values.pdb.maxUnavailable) (eq .Values.pdb.maxUnavailable "auto") }}

{{- $maxUnavailable = $maxFault }}

{{- end }}

vlasopoulos · 2026-03-20T09:42:44Z

values.yaml

  # type: integer
-  # minimum: 1
+  # minimum: -1
  # @schema
-  # -- Maximum number of pods that can be unavailable during disruption
-  maxUnavailable: 1
+  # -- Maximum number of pods that can be unavailable during disruption.
+  # Set to -1 (default) to auto-calculate as floor(replicaCount/2), preserving Raft quorum.
+  # Set to 0 to block all voluntary disruptions. Any positive value is used directly
+  # and must not exceed floor(replicaCount/2).
+  maxUnavailable: -1


Suggested change

# type: integer

# minimum: 1

# minimum: -1

# @schema

# -- Maximum number of pods that can be unavailable during disruption

maxUnavailable: 1

# -- Maximum number of pods that can be unavailable during disruption.

# Set to -1 (default) to auto-calculate as floor(replicaCount/2), preserving Raft quorum.

# Set to 0 to block all voluntary disruptions. Any positive value is used directly

# and must not exceed floor(replicaCount/2).

maxUnavailable: -1

# type: integer, string

# minimum: -1

# @schema

# -- Maximum number of pods that can be unavailable during disruption.

# Set to "auto" (default) to auto-calculate as floor(replicaCount/2), preserving Raft quorum.

# Set to 0 to block all voluntary disruptions. Any positive value is used directly

# and must not exceed floor(replicaCount/2).

maxUnavailable: auto

…else for updateStrategy - maxUnavailable defaults to -1 (auto-calculate), 0 blocks all disruptions - Schema enforces updateStrategy.type enum (RollingUpdate/OnDelete) - Schema conditionally requires rollingUpdate only for RollingUpdate type - Template validation kept as defense in depth - Fix changelog to reflect actual defaults

dimoschi · 2026-03-20T10:13:37Z

Just for discussion, since we are in a flimsy dynamically typed template world that becomes strict at render (don't get me started), we could set a trend of values making sense on their own.

I've missed your comment and went on a different way. Do you think that makes more sense using -1 for auto? Regardless, I went with your way, even though it's not so Helm friendly, it looks more intuitive.

…ma validation - maxUnavailable defaults to "auto" (auto-calculate as floor(replicaCount/2)) - Integer values are used directly, 0 blocks all voluntary disruptions - Schema enforces anyOf [integer >= 0, "auto"] for maxUnavailable - Schema enforces updateStrategy.type enum (RollingUpdate/OnDelete) - Schema uses if/then to conditionally require rollingUpdate for RollingUpdate - Update changelog to reflect "auto" defaults

vlasopoulos

The quorum approves

dimoschi added 7 commits March 20, 2026 08:46

feat(values): enable PDB by default, add updateStrategy, auto-calc ma…

12129cc

…xUnavailable

feat(pdb): auto-enable for multi-node, auto-calculate maxUnavailable …

f32ae71

…with quorum validation

feat(statefulset): add explicit updateStrategy with auto-calculated m…

8062508

…axUnavailable

feat(statefulset): add built-in soft pod anti-affinity for multi-node…

e6444bd

… clusters

docs: regenerate schema and documentation for HA guardrails

91ae0ca

test: remove hardcoded chart version from test suites

643bbac

chore: bump chart version to 1.1.0

055714c

github-actions bot changed the title ~~feat: HA & Raft quorum safety guardrails~~ XS✔ ◾ feat: HA & Raft quorum safety guardrails Mar 20, 2026

dimoschi requested review from Copilot and vlasopoulos March 20, 2026 07:36

Copilot started reviewing on behalf of dimoschi March 20, 2026 07:37 View session

This comment was marked as outdated.

Sign in to view

fix(schema): add type annotation for affinity as nullable object

fbe1f41

dimoschi requested a review from a team as a code owner March 20, 2026 08:00

dimoschi added 3 commits March 20, 2026 10:03

fix(schema): use unquoted null in @Schema type annotations

24e1986

fix(schema): set affinity default to {}, add type annotation, unquote…

d733e33

… null in schema annotations

fix(schema): use 0 sentinel for auto-calculated maxUnavailable, set a…

f85ade8

…ffinity default to {} maxUnavailable: 0 means auto-calculate as floor(replicaCount/2). This gives proper integer typing in schema and docs. affinity: {} enables correct object type inference by helm-docs.

dimoschi requested a review from Copilot March 20, 2026 08:51

Copilot started reviewing on behalf of dimoschi March 20, 2026 08:51 View session

This comment was marked as duplicate.

Sign in to view

dimoschi added 2 commits March 20, 2026 11:18

test: add coverage for maxUnavailable -1, 0, and positive values

2781846

vlasopoulos reviewed Mar 20, 2026

View reviewed changes

dimoschi added 2 commits March 20, 2026 12:27

docs: set maxUnavailable type to string/int in helm-docs

22accbb

dimoschi requested review from Copilot and vlasopoulos March 20, 2026 10:32

Copilot started reviewing on behalf of dimoschi March 20, 2026 10:32 View session

This comment was marked as outdated.

Sign in to view

dimoschi requested a review from Copilot March 20, 2026 10:38

Copilot started reviewing on behalf of dimoschi March 20, 2026 10:38 View session

This comment was marked as outdated.

Sign in to view

docs: clarify affinity comment to say "when unset or empty"

51875fa

vlasopoulos approved these changes Mar 20, 2026

View reviewed changes

dimoschi merged commit eb58901 into main Mar 20, 2026
13 checks passed

dimoschi deleted the feat/ha-quorum-safety branch March 20, 2026 15:36

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

XS✔ ◾ feat: HA & Raft quorum safety guardrails#2

XS✔ ◾ feat: HA & Raft quorum safety guardrails#2
dimoschi merged 17 commits intomainfrom
feat/ha-quorum-safety

dimoschi commented Mar 20, 2026 •

edited

Loading

Uh oh!

github-actions bot commented Mar 20, 2026 •

edited

Loading

Uh oh!

This comment was marked as outdated.

Uh oh!

This comment was marked as duplicate.

Uh oh!

vlasopoulos left a comment •

edited

Loading

Uh oh!

vlasopoulos Mar 20, 2026

Uh oh!

vlasopoulos Mar 20, 2026

Uh oh!

dimoschi commented Mar 20, 2026 •

edited

Loading

Uh oh!

This comment was marked as outdated.

Uh oh!

This comment was marked as outdated.

Uh oh!

vlasopoulos left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

dimoschi commented Mar 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Quorum safety

Breaking changes

Schema validation

Test plan

Uh oh!

github-actions bot commented Mar 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Metrics

Uh oh!

This comment was marked as outdated.

Uh oh!

This comment was marked as duplicate.

Uh oh!

vlasopoulos left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

vlasopoulos Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

vlasopoulos Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

dimoschi commented Mar 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as outdated.

Uh oh!

This comment was marked as outdated.

Uh oh!

vlasopoulos left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

dimoschi commented Mar 20, 2026 •

edited

Loading

github-actions bot commented Mar 20, 2026 •

edited

Loading

vlasopoulos left a comment •

edited

Loading

dimoschi commented Mar 20, 2026 •

edited

Loading