Refactor backend status handling and improve idle management

- Updated the idle-killer logic to treat SwarmUI `empty` and `disabled` states as busy, preventing unnecessary idle time during provisioning.
- Enhanced the `wait_backend_idle` function to recognize suspended backends as ready, improving resource utilization and user feedback.
- Refined the `install_swarm_comfy` script to skip installation when backends are already present, streamlining the setup process.
- Improved the `resolve_llm_runtime` function to prioritize live configuration over stale state notes, ensuring accurate runtime detection.
- Added tests to validate the new backend status handling and idle management logic, ensuring robustness and reliability.
This commit is contained in:
Leonid Pershin
2026-08-21 10:07:16 +03:00
parent 26f3be6e96
commit 1785ab369c
13 changed files with 302 additions and 90 deletions
+100 -9
View File
@@ -234,18 +234,109 @@ def seed_autocomplete(cfg: Config, host: str, log: Log) -> bool:
applied = bool(json.loads(raw).get("settings_applied"))
except json.JSONDecodeError:
applied = False
if not remote_exists(cfg, host, settings) or not applied:
fds = (
"DefaultUser:\n"
" AutoComplete:\n"
f" Source: {cfg.autocomplete_filename}\n"
" EscapeParens: true\n"
)
put_text(cfg, host, settings, fds)
log(f"Settings.fds AutoComplete.Source = {cfg.autocomplete_filename}")
if not applied:
_merge_autocomplete_into_settings(cfg, host, settings, cfg.autocomplete_filename, log)
if remote_exists(cfg, host, meta_path):
try:
meta_obj = json.loads(
run_ssh(cfg, host, f"cat {meta_path}", check=False)
)
except json.JSONDecodeError:
meta_obj = {}
if isinstance(meta_obj, dict):
meta_obj["settings_applied"] = True
put_text(cfg, host, meta_path, json.dumps(meta_obj, indent=2) + "\n")
return changed
_AUTOCOMPLETE_MERGE_PY = r'''
#!/usr/bin/env python3
"""Merge AutoComplete.Source into Settings.fds without wiping other keys."""
from __future__ import annotations
import os
import re
import sys
from pathlib import Path
p = Path(os.environ.get("GPU_RENT_SETTINGS_FDS") or "/mnt/swarm_data/Data/Settings.fds")
fname = (os.environ.get("GPU_RENT_AUTOCOMPLETE_FILE") or "").strip()
if not fname:
print("no GPU_RENT_AUTOCOMPLETE_FILE", file=sys.stderr)
raise SystemExit(1)
block = (
"DefaultUser:\n"
" AutoComplete:\n"
f" Source: {fname}\n"
" EscapeParens: true\n"
)
if not p.is_file():
p.parent.mkdir(parents=True, exist_ok=True)
p.write_text(block, encoding="utf-8")
print(f"created Settings.fds AutoComplete.Source={fname}")
raise SystemExit(0)
text = p.read_text(encoding="utf-8", errors="replace")
if re.search(rf"^\s*Source:\s*{re.escape(fname)}\s*$", text, re.M):
print(f"AutoComplete.Source already {fname}")
raise SystemExit(0)
# Replace Source line if AutoComplete section exists
new, n = re.subn(
r"(^[ \t]*Source:\s*).*$",
rf"\1{fname}",
text,
count=1,
flags=re.M,
)
if n and "AutoComplete" in text:
p.write_text(new, encoding="utf-8")
print(f"patched AutoComplete.Source={fname}")
raise SystemExit(0)
if re.search(r"^DefaultUser:\s*$", text, re.M):
new = re.sub(
r"^(DefaultUser:\s*\n)",
(
r"\1 AutoComplete:\n"
f" Source: {fname}\n"
" EscapeParens: true\n"
),
text,
count=1,
flags=re.M,
)
p.write_text(new, encoding="utf-8")
print(f"inserted AutoComplete under DefaultUser Source={fname}")
else:
p.write_text(text.rstrip() + "\n\n" + block, encoding="utf-8")
print(f"appended DefaultUser.AutoComplete Source={fname}")
'''
def _merge_autocomplete_into_settings(
cfg: Config, host: str, settings_path: str, filename: str, log: Log
) -> None:
"""Patch AutoComplete.Source in Settings.fds without wiping the rest."""
out = run_python(
cfg,
host,
_AUTOCOMPLETE_MERGE_PY,
remote_path="/tmp/gpu-rent-patch_autocomplete.py",
timeout=60,
log=None,
env={
"GPU_RENT_SETTINGS_FDS": settings_path,
"GPU_RENT_AUTOCOMPLETE_FILE": filename,
},
)
for line in (out or "").splitlines():
if line.strip():
log(line.strip())
def _download_url(host: str, version_id: int, file_info: dict) -> str:
raw = str(file_info.get("downloadUrl") or "")
if "civitai." in raw and "/api/download/" in raw: