Scrape duration using prometheus is increasing continuously until restart

### problem

We are using Prometheus to scrape metrics from the management server.
We see a issue that the scrape_duration_seconds metric for the cloudstack job is increasing in slow pace and after ~7-10 days the scrape duration is longer then the configured interval and we get no more metrics.
First time we saw this just doing a time curl towards the mangement server it took 40-50s for metrics to come back.

### versions

CloudStack 4.22.0.0

### The steps to reproduce the bug

1. Start scraping
2. See the scrape_duration_seconds{job="cloudstack"} increasing daily.

### What to do about it?

1. Restart the management server solves the issue, temporarily

从 CloudStack 管理服务器的 Prometheus metrics endpoint 开始，通过连续数天对其进行抓取来重现该问题，同时观察 scrape_duration_seconds{job="cloudstack"}。将 endpoint 响应时间与配置的抓取间隔进行比较；当抓取持续时间不再持续增加，并且 metrics 在无需重启管理服务器的情况下仍然可用时，即表示完成。

由索引模型根据 Issue 内容生成。