• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

chideat / valkey-operator / 34866629600
36%

Build:
DEFAULT BRANCH: main
Ran 14 Sep 2026 04:11PM UTC
Jobs 1
Files 147
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

14 Sep 2026 04:07PM UTC coverage: 26.588% (+0.05%) from 26.536%
34866629600

push

github

web-flow
fix: TLS failover instances never became ready (two defects) (#127)

* fix: a TLS failover instance never became ready, its sentinel stayed plaintext

Setting access.enableTLS on a failover instance produced TLS valkey pods and a
plaintext sentinel. The valkey pods dial the sentinel using the failover's own
TLS setting, so every check hit a listener that does not speak TLS:

  check monitoring master failed
    {"node": "rfs-<name>-0.rfs-<name>.<ns>:26379", "error": "TLS handshake timeout"}

The nodes never went ready and the instance sat in Initializing indefinitely.

The defaulting webhook already propagates ServiceType and IPFamilyPrefer from
the instance to its sentinel, for the same reason both have to agree on them.
EnableTLS was simply missing from that group, along with the issuer fields.

TLS here is not a per-side preference. The two sides share one certificate and
dial each other: the valkey pods reach the sentinel using the failover's
setting, and the operator reaches it using the sentinel's -- failover_controller
reads Spec.Sentinel.Access.EnableTLS off the CR to decide whether its own
sentinel client uses TLS. Any disagreement is one side speaking TLS to a
plaintext listener, so the sentinel is now made to match rather than merely
allowed to differ. That also corrects the inverse mismatch, where a
sentinel-only EnableTLS left the valkey pods dialling a TLS listener in
plaintext.

Fixing this in the webhook rather than in NewFailoverSentinel is deliberate:
the builder only shapes the StatefulSet, so flipping it there would have given
the sentinel TLS while the CR field stayed false -- moving the mismatch from
the helper to the operator instead of removing it. Setting the field once,
where the other two are already set, makes every downstream reader agree.

Creating a Failover directly still needs both sides set by hand; that API has a
validating webhook only, and ServiceType has the same gap today.

Found by the new TLS e2e coverage runni... (continued)

12 of 12 new or added lines in 1 file covered. (100.0%)

5530 of 20799 relevant lines covered (26.59%)

0.3 hits per line

Jobs
ID Job ID Ran Files Coverage
1 34866629600.1 14 Sep 2026 04:11PM UTC 147
26.59
GitHub Action Run
Source Files on build 34866629600
  • Tree
  • List 147
  • Changed 2
  • Source Changed 0
  • Coverage Changed 2
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • 1250ae24 on github
  • Prev Build on main (#34866477612)
  • Next Build on main (#36077131083)
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc