{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiccihfygwpl5zzcnttn6uqlfb5477seb4zkjm4unr52esie24bfke",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mjgu26frzny2"
  },
  "path": "/t/continous-increase-in-memory-usage/127891#post_17",
  "publishedAt": "2026-04-14T06:20:02.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "@John6666"
  ],
  "textContent": "@John6666 Previously my RSS was not increasing or it was increasing and comes back down to original point after request completion and the free -h used was growing continuously. Now Currently I’ve tried your approach but in this case both RSS and free -h used are growing with requests.\n\nI have tested your approach by giving 3600 requests for 1 hr (1 req/s)\n\nStats before request start\n\nStats after requests completion\n\nEvery 1.0s: free -b openclaw-serv-02: Tue Apr 14 11:17:09 2026\n\n\n                 total        used        free      shared   buff/cache   available\n\n    Mem:     33514856448  5407645696 22296809472    19968000  6310006784 28107210752\n    Swap:     8589930496           0  8589930496\n\n\nEvery 10.0s: ps -C uvicorn -o rss,vsz,command openclaw-serv-02: Tue Apr 14 11:17:18 2026\n\n\n      RSS      VSZ   COMMAND\n    1503860 14842804 python uvicorn app:app --host 0.0.0.0 --port 9000\n\n\nI want both RSS and free -h used come back to its original state after every request completion. What could be the approach for it as according to the above approach my GPU server will collapse after some time due to RAM limit reach. Need your opinion on this.",
  "title": "Continous increase in Memory usage"
}