This PR replaces Thanos with Cortex in MLA stack in order to provide multi-tenancy Alertmanager support.
To do that, it introduces few other components:
Concul: for communications between cortex components.
Minio: for storing Alertmanger configurations.
For Alertmanger, a public gateway service is added (mla-gateway-alert) to see the Alertmanger UI and upload alertmanger configurations.
Also, currently, for seeing the Alertmanger UI, we should create the Alertmanger configuration first. In the future, we can define a fallback Alertmanger configuration (see: https://cortexmetrics.io/docs/configuration/configuration-file/#alertmanager_config for more details), which will be used if no Alertmanger configuration is specified. In current PoC, we don't use the fallback configuration because in that way the same configuration will be used for all tenants, and we can not see if multi-tenancy is working.
For now, Memcached is disabled in Cortex, this is because Loki is also using it, and we don't want to deploy memchched stack twice, and we should find a way of sharing Memcached between Cortex and Loki.
This PR replaces Thanos with Cortex in MLA stack in order to provide multi-tenancy Alertmanager support. To do that, it introduces few other components:
For Alertmanger, a public gateway service is added (mla-gateway-alert) to see the Alertmanger UI and upload alertmanger configurations.
Also, currently, for seeing the Alertmanger UI, we should create the Alertmanger configuration first. In the future, we can define a fallback Alertmanger configuration (see: https://cortexmetrics.io/docs/configuration/configuration-file/#alertmanager_config for more details), which will be used if no Alertmanger configuration is specified. In current PoC, we don't use the fallback configuration because in that way the same configuration will be used for all tenants, and we can not see if multi-tenancy is working.
For now, Memcached is disabled in Cortex, this is because Loki is also using it, and we don't want to deploy memchched stack twice, and we should find a way of sharing Memcached between Cortex and Loki.