vector() function produces wrong timestamp on remote write

[mprokopov]

Describe the bug
If vector() function is used to produce a default value for extracted metrics from logs, it's being sent to prometheus using remote write with bad timestamp. Related to #6946

The major issue it brings, is that Prometheus fully stops accepting any new metrics after this event. The only way to resume Prometheus operations is to stop and completely clear TSDB.

To Reproduce
Steps to reproduce the behavior:

1. Started Loki 2.7.1 with the following config snippet for ruler

ruler:
  storage:
    type: local
    local:
      directory: /rules
  wal:
    dir: /loki/rules-wal
  rule_path: /tmp/rules
  enable_api: true
  remote_write:
    enabled: true
    client:
      url: http://prometheus.my.internal/api/v1/write

2. Started Promtail 2.7.1
3. Started Prometheus with remote write option enabled.
4. Rule:

   groups: 
   - name: Traefik
     interval: 1m
    rules:
    - record: traefic_count
      expr: |
        sum(count_over_time({namespace="traefik"}[5m]))
        or
        vector(0)
   

Expected behavior
Exported metrics to prometheus endpoint using remote write should use correct timestamp.

Environment:

• Infrastructure: bare-metal with docker
• Deployment tool: ansible

Screenshots, Promtail config, or terminal output

[liguozhong]

I am very sorry that this bug has troubled you, I will fix this bug as soon as possible

[mprokopov]

I really appreciate your efforts, @liguozhong!
Thank you very much for responding to this issue so fast.