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 1.0.26Documentation: 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 (spoofableX-SSL-Client-Verify: SUCCESSwith 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 (raisesettings.ui.sessionLifetimeHours/sessionAbsoluteHoursinstead); 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 malformedDATABASE_URI. Also OWASP CRS (GHSA-6jp8-c2w2-x7wr, GHSA-f5qm-3h4p-8qhg), Mbed TLS 4.2.0, and Alpinepython33.14.7-r0 (CVE-2026-7210). - [FEATURE]
mariadb: newmariadb.securityContext— pod-level security context, needed to setfsGroupwhen running a non-root MariaDB image. (#94, thanks @martinbfrey) - [FEATURE]
scheduler.features.metrics: newmetricsRedisTtl(METRICS_REDIS_TTL) — TTL in seconds for metrics keys in Redis (default2592000;0keeps them permanent), refreshed on every sync so abandoned metrics stay evictable undervolatile-lru. - [BUGFIX] upstream
scheduler: pod restarts no longer strand an instance on its loading configuration (stale.bw-appliedmarker), 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 themtlsplugin; reach these viascheduler.extraEnvs. - [DOCS] upstream:
CROWDSEC_*settings are nowmultisite(the chart'sscheduler.features.crowdSeckeys still apply globally);KEEP_CONFIG_ON_RESTARTis no longer a setting — set it viabunkerweb.extraEnvs;QUERYis allowed by default inALLOWED_METHODS.