You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
graylog:appVersion moves from 7.0 to 7.1.8. graylog.image.tag and datanode.image.tag default to appVersion, so upgrading moves the running Graylog and Data Node images unless you pin both tags (#181) (eee5bd8)
values:global.imagePullSecrets, graylog.image.imagePullSecrets and datanode.image.imagePullSecrets now take {name: <secret>} objects instead of bare strings. Schema validation rejects the old string form (#93) (cdb719c)
mongodb: the default replica set is now 3 data-bearing members with no arbiter, replacing 2 members plus 1 arbiter. Upgrading on the defaults adds a member and drops the arbiter, so budget for a third data volume (#127) (47edf85)
mongodb: the chart now sets MongoDB container resources, which nothing could override before. A pod reserves 300m/1152Mi instead of the operator's 1000m/800M, and upgrading rolls the replica set (#176) (e49b97f)
pods and containers now run under a restricted security context: runAsNonRoot: true, seccompProfile: RuntimeDefault, allowPrivilegeEscalation: false, and all capabilities dropped except NET_BIND_SERVICE on graylog and the file-ownership set on datanode. A custom image that runs as root or needs other capabilities will not start (#100) (cb9bd63)
values: removed datanode.persistence.enabled, datanode.persistence.data.existingClaim, datanode.persistence.data.selector, datanode.persistence.data.dataSource, datanode.persistence.nativeLibs.existingClaim, datanode.persistence.nativeLibs.selector, graylog.persistence.mountPath and graylog.persistence.selector. None of them were wired to anything (6571951)
geo-ip: the MaxMind database update runs as a sidecar on the Graylog pod instead of a CronJob. graylog.config.geolocation.maxmindGeoIp.cronSchedule moves to graylog.config.geolocation.sidecar.schedule, and postInstallRun is gone (#129) (e9c0354)
ingress:ingress.forwarder splits into two channels, ingress.forwarder.messageChannel and ingress.forwarder.configChannel, each with its own className, annotations, labels, hosts and tls. Existing forwarder ingress values need rewriting, not renaming (#160) (29c1430)
datanode: the native-libs volume claim template changes name from nativeLibs to native-libs. volumeClaimTemplates is immutable, so Kubernetes rejects the upgrade for anyone running with datanode.persistence.nativeLibs.enabled=true. Delete the StatefulSet with --cascade=orphan first (#75) (5fc7e22)
graylog:terminationGracePeriodSeconds is now 300, and a preStop hook drains the journal before shutdown. Pod deletion, node drains and rolling upgrades take up to five minutes per pod (#147) (c643cd0)
PodDisruptionBudgets are on by default for both graylog and datanode. A single-replica deployment blocks node drains until you relax the budget
helpers: an explicit graylog.config.network.externalUri now wins over the LoadBalancer Service lookup, and a bare hostname gets the scheme and app port appended. Setting both a LoadBalancer service and externalUri changes the advertised http_external_uri (e3c965d)
the chart renames the init script ConfigMap from the fixed init-script-cm to <release>-graylog-init-cm, so two releases can share a namespace. External references to the old name break (#79) (25653e2)
the chart writes the generated root password to the backup Secret instead of printing it in NOTES.txt. Anything scraping the install output for the password must read the Secret (33499f2)
probes: both StatefulSets now get a startupProbe, on by default. It allows 330s before the kubelet restarts the container, so an install that takes longer restarts in a loop until you raise startupProbe.failureThreshold (#169) (b411145)
probes: removed graylog.livenessProbe.successThreshold and datanode.livenessProbe.successThreshold. The schema does not reject unknown keys, so either one still validates and does nothing (#169) (b411145)