Prometheus v2.41.0内存持续上涨，咨询内存限制配置及是否正常

环境与配置 #

通过kube-prometheus-stack Helm Chart安装Prometheus v2.41.0，使用的配置如下：

prometheus:
  prometheusSpec:
    retention: 30d
    retentionSize: 49GiB
    storageSpec:
      volumeClaimTemplate:
        spec:
          resources:
            requests:
              storage: 50Gi

运行现象 #

• 内存使用：运行数天后内存持续上涨，无停止趋势，认为无需将所有数据一直驻留内存，希望了解可限制内存占用的配置项。
• PVC存储：占用量约为内存的两倍，符合预期。
• pprof堆数据：

Fetching profile over HTTP from http://localhost:55947/debug/pprof/heap
File: prometheus
Type: inuse_space
Time: Jun 12, 2023 at 2:54pm (CEST)
Entering interactive mode (type "help" for commands, "o" for options)
(pprof) top
Showing nodes accounting for 216MB, 69.80% of 309.47MB total
Dropped 147 nodes (cum <= 1.55MB)
Showing top 10 nodes out of 135
      flat  flat%   sum%        cum   cum%
   58.55MB 18.92% 18.92%    58.55MB 18.92%  github.com/prometheus/prometheus/model/labels.(*Builder).Labels
   39.97MB 12.92% 31.84%    39.97MB 12.92%  github.com/prometheus/prometheus/scrape.newScrapePool.func1
   24.51MB  7.92% 39.76%    24.51MB  7.92%  github.com/prometheus/prometheus/model/textparse.(*PromParser).Metric
      23MB  7.43% 47.19%       31MB 10.02%  github.com/prometheus/prometheus/tsdb.newMemSeries
   17.50MB  5.66% 52.84%    17.50MB  5.66%  github.com/prometheus/prometheus/tsdb/chunkenc.NewXORChunk
   12.50MB  4.04% 56.88%    18.02MB  5.82%  github.com/prometheus/prometheus/tsdb.(*memSeries).mmapCurrentHeadChunk
   11.83MB  3.82% 60.71%    11.83MB  3.82%  github.com/prometheus/prometheus/scrape.(*scrapeCache).trackStaleness
   10.51MB  3.39% 64.10%    10.51MB  3.39%  github.com/prometheus/prometheus/tsdb/chunkenc.(*bstream).writeByte
    8.84MB  2.86% 66.96%     8.84MB  2.86%  github.com/prometheus/prometheus/scrape.(*scrapeCache).addRef
    8.79MB  2.84% 69.80%     8.79MB  2.84%  github.com/golang/snappy.Encode

• TSDB数据：数据表现符合对应监控图表展示的状态。

疑问 #

pprof显示堆内存仅使用309MB，但实际进程内存占用达3-4GB，对此感到困惑。请问这是Prometheus的正常表现吗？是否有办法限制其内存需求？此前查找过类似问题，但未找到合适的解决方案。

内容的提问来源于stack exchange，提问作者vitas-pm