Zabbix Helm chart 8.0.1
Caution
Pay attention to changes of release 8.0.0: https://github.com/zabbix-community/helm-zabbix/releases/tag/zabbix-8.0.0
Fixes
- Zabbix Server HA works again. With chart 8.0.0,
zabbixServer.zabbixServerHA.enabled=truefailed withImagePullBackOff: the default DB job image8.0-20261009150714was never published. The chart now uses DB job images built from Zabbix 8.0.0 (tagSuffix: "20261107172746") and the new HA labels sidecar20261107173134. - First
helm upgradeenabling HA (#201): the DB job now has its own ServiceAccount/Role/RoleBinding (<fullname>-create-upgrade-db), created as hooks before it runs. - Argo CD (#231): the job no longer relies on
.Release.IsUpgrade. WithdbCreateUpgradeJob.hookType: auto(default), it detects an install or upgrade by checking whether the Zabbix Server Deployment exists. - Passive Zabbix Proxy (
ZBX_PROXYMODE: 1) starts again: no port is appended toZBX_SERVER_HOST. In active mode,ZBX_SERVER_PORTis appended to each;-separated HA node without a port. - A failed DB job no longer blocks the next upgrade (
before-hook-creation). - DB job image setting: a
tagSuffixwithout quotes no longer renders an invalid image tag such as8.0-%!s(int64=...)(zabbix-community/helm-zabbix-image-db-init-upgrade-job#2). prepare_db: command not found(#219): the new job images support the modular entrypoint of the Zabbix images.- Trunk images in HA mode (#233): they are still not supported, but the chart now fails with a clear message that explains how to set
dbCreateUpgradeJob.image.tag.
New values
| Value | Default | Description |
|---|---|---|
zabbixServer.zabbixServerHA.dbCreateUpgradeJob.hookType
| auto
| auto, install or upgrade
|
zabbixServer.zabbixServerHA.dbCreateUpgradeJob.waitTimeout
| 900
| Maximum seconds for each wait of the job |
zabbixServer.zabbixServerHA.dbCreateUpgradeJob.activeDeadlineSeconds
| empty | Maximum job duration. Schema upgrades of big databases can take hours |
zabbixProxy.affinity
| {}
| Replaces the global affinity for the proxy only
|
zabbixProxy.minReadySeconds
| 0
| For example, 60 avoids restarting all the replicas of a proxy group at once
|
zabbixProxy.updateStrategy
| {}
| StatefulSet update strategy, for example rollingUpdate.partition
|
Upgrading from 8.0.0
No changes to values.yaml are required. With HA, the <fullname>-ha-helper ServiceAccount is replaced by <fullname>-create-upgrade-db, which gets the annotations from zabbixServerHA.serviceAccount.annotations. For long schema upgrades, increase Helm's --timeout.
Docs
- New HA tutorial for kind with Helm and Argo CD: docs/example/ha/README.md
- HA architecture, job decision flow and troubleshooting: README
PgBouncer
If a PgBouncer in front of PostgreSQL rejects the search_path option sent by the DB job, set POSTGRES_USE_IMPLICIT_SEARCH_PATH=true in zabbixServer.extraEnv. The job receives the same variables.
Related
- #237
- zabbix-community/helm-zabbix-image-db-init-upgrade-job#4
- zabbix-community/helm-zabbix-image-ha-labels-sidecar#1
Thanks
This release includes changes proposed in #223 (@Jonher937), #229 and #230 (@jDmacD) and #236 (@Qetesh).
Thanks to everyone who reported and investigated the HA issues in #201, #219, #231 and #233.
Contributors: @aeciopires @Jonher937 @jDmacD @Qetesh