Long label name queries returns duplicate value

[cyriltovena]

There must be a bug somewhere in the new schema.

[aocenas]

This is the response from Loki seen by Grafana:

{"values":["__name__","app","ceph_daemon_id","ceph_osd_id","cluster","color","component","container_name","controller_revision_hash","controller_uid","diff_name","diff_namespace","diff_type","diff_verb","etcd_cluster","etcd_node","filename","hosted_metrics","instance","job","job_name","k8s_app","level","mgr","mon","mon_cluster","name","namespace","nodename","org","partitions","partitions_group","plan","pod_template_generation","pod_template_hash","product","redis_role","replica","restart","revision","rgw","role","rook_cluster","rook_object_store","statefulset_kubernetes_io_pod_name","stream","syslog_identifier","systemd_unit","test","tier","__name__","app","ceph_daemon_id","ceph_osd_id","cluster","color","component","container_name","controller_revision_hash","controller_uid","diff_name","diff_namespace","diff_type","diff_verb","etcd_cluster","etcd_node","filename","hosted_metrics","instance","jaeger_infra","job","job_name","k8s_app","level","mgr","mon","mon_cluster","name","namespace","nodename","org","partitions","partitions_group","plan","pod_template_generation","pod_template_hash","product","redis_role","replica","restart","revision","rgw","role","rook_cluster","rook_object_store","statefulset_kubernetes_io_pod_name","stream","syslog_identifier","systemd_unit","test","tier","version"]}

So seems the duplicated labels come from Loki and we do not do any deduplication later on.

[cyriltovena]

This is from cortex with the composite store. I'll send a PR there.