Skip to content

v1.0.26

Latest

Choose a tag to compare

@TheophileDiot TheophileDiot released this 24 Aug 09:39
444f0ae

BunkerWeb Helm chart v1.0.26 (app version 1.6.14).

helm repo add bunkerweb https://repo.bunkerweb.io/charts
helm repo update
helm install bunkerweb bunkerweb/bunkerweb --version 1.0.26

Documentation: https://docs.bunkerweb.io/

Changelog:
App version: 1.6.14.

  • [DEPS] upgrade to BunkerWeb 1.6.14.
  • [SECURITY] upstream: X-SSL-* request headers from clients are stripped before reaching an upstream (spoofable X-SSL-Client-Verify: SUCCESS with mTLS header forwarding off, or on any PHP-FPM service); UI session revocations are stored in Redis instead of a file outside the persistent volume, so they survive pod recreation and reach every replica; the year-long "remember me" token is removed (raise settings.ui.sessionLifetimeHours / sessionAbsoluteHours instead); TOTP codes can no longer be replayed; API config pushes no longer empty the target directory on a live instance mid-copy; Biscuit tokens with appended blocks are rejected; the database password is no longer logged on a malformed DATABASE_URI. Also OWASP CRS (GHSA-6jp8-c2w2-x7wr, GHSA-f5qm-3h4p-8qhg), Mbed TLS 4.2.0, and Alpine python3 3.14.7-r0 (CVE-2026-7210).
  • [FEATURE] mariadb: new mariadb.securityContext — pod-level security context, needed to set fsGroup when running a non-root MariaDB image. (#94, thanks @martinbfrey)
  • [FEATURE] scheduler.features.metrics: new metricsRedisTtl (METRICS_REDIS_TTL) — TTL in seconds for metrics keys in Redis (default 2592000; 0 keeps them permanent), refreshed on every sync so abandoned metrics stay evictable under volatile-lru.
  • [BUGFIX] upstream scheduler: pod restarts no longer strand an instance on its loading configuration (stale .bw-applied marker), the initial push waits for the instance to answer, and the job environment is refreshed on reload — a DNS-01 service created in the web UI previously needed a container restart to get its certificate.
  • [BUGFIX] upstream core: an empty setting no longer renders an argument-less directive (gzip_proxied ;) that NGINX refuses to start on, and deleting a plugin cache entry no longer wipes that plugin's cache directory (which destroyed Let's Encrypt ACME accounts and live certificate links).
  • [DOCS] MTLS_CA_CERTIFICATE_DATA / MTLS_CRL_DATA (new in 1.6.14) supply the client CA bundle and CRL inline as base64 or PEM instead of a path every instance must read — the usual Kubernetes case. As in v1.0.23, the chart does not expose the mtls plugin; reach these via scheduler.extraEnvs.
  • [DOCS] upstream: CROWDSEC_* settings are now multisite (the chart's scheduler.features.crowdSec keys still apply globally); KEEP_CONFIG_ON_RESTART is no longer a setting — set it via bunkerweb.extraEnvs; QUERY is allowed by default in ALLOWED_METHODS.