@infernetprotocol/cli 0.1.90 → 0.1.92
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/commands/start.js +16 -0
- package/package.json +9 -9
package/commands/start.js
CHANGED
|
@@ -423,6 +423,10 @@ async function runDaemon(args, ctx) {
|
|
|
423
423
|
}
|
|
424
424
|
return {
|
|
425
425
|
active_jobs: stats.activeJobIds.size,
|
|
426
|
+
// How many jobs this daemon runs at once. Without it the router
|
|
427
|
+
// guesses (1 on CPU-only nodes, 4 with a GPU) and a GPU node
|
|
428
|
+
// running one job at a time gets 3 queued behind it.
|
|
429
|
+
concurrency_cap: maxConcurrentJobs,
|
|
426
430
|
load_avg_1m: Array.isArray(host.load_avg) ? +host.load_avg[0]?.toFixed(2) : null,
|
|
427
431
|
ram: {
|
|
428
432
|
total_gb: +totalRamGb.toFixed(2),
|
|
@@ -629,10 +633,21 @@ async function runDaemon(args, ctx) {
|
|
|
629
633
|
// frees up.
|
|
630
634
|
if (stats.activeJobIds.size >= maxConcurrentJobs) return;
|
|
631
635
|
const polledAt = Date.now();
|
|
636
|
+
let refetch = false;
|
|
632
637
|
try {
|
|
633
638
|
const result = await client.pollJobs({ limit: 5 });
|
|
634
639
|
stats.pollsOk += 1;
|
|
640
|
+
const fetchedAt = Date.now();
|
|
635
641
|
for (const job of result?.jobs ?? []) {
|
|
642
|
+
// Jobs are awaited one by one, so after a long job the rest of
|
|
643
|
+
// this batch is minutes old. The control plane may have moved
|
|
644
|
+
// a queued job to a free node meanwhile (chat rebalancing), and
|
|
645
|
+
// running it here would only be refused at the first event.
|
|
646
|
+
// Fetch a fresh list instead.
|
|
647
|
+
if (Date.now() - fetchedAt > pollMs) {
|
|
648
|
+
refetch = true;
|
|
649
|
+
break;
|
|
650
|
+
}
|
|
636
651
|
// Skip jobs already being processed by this daemon — the
|
|
637
652
|
// server keeps returning 'assigned' jobs until we mark them
|
|
638
653
|
// complete, so without this guard the poll loop queues the
|
|
@@ -652,6 +667,7 @@ async function runDaemon(args, ctx) {
|
|
|
652
667
|
stats.pollsFailed += 1;
|
|
653
668
|
process.stderr.write(`job poll error: ${err?.message ?? err}\n`);
|
|
654
669
|
}
|
|
670
|
+
if (refetch && !stats.draining) await pollJobs();
|
|
655
671
|
}
|
|
656
672
|
|
|
657
673
|
/**
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@infernetprotocol/cli",
|
|
3
|
-
"version": "0.1.
|
|
3
|
+
"version": "0.1.92",
|
|
4
4
|
"description": "GPU-node CLI / daemon for the Infernet Protocol — register a GPU server with a Supabase control plane and start earning.",
|
|
5
5
|
"license": "MIT",
|
|
6
6
|
"type": "module",
|
|
@@ -32,14 +32,14 @@
|
|
|
32
32
|
"ora": "^9.4.1",
|
|
33
33
|
"react": "^17.0.2",
|
|
34
34
|
"react-blessed": "^0.7.2",
|
|
35
|
-
"@infernetprotocol/auth": "0.1.
|
|
36
|
-
"@infernetprotocol/config": "0.1.
|
|
37
|
-
"@infernetprotocol/
|
|
38
|
-
"@infernetprotocol/
|
|
39
|
-
"@infernetprotocol/gpu": "0.1.
|
|
40
|
-
"@infernetprotocol/liquid-agent": "0.1.
|
|
41
|
-
"@infernetprotocol/
|
|
42
|
-
"@infernetprotocol/
|
|
35
|
+
"@infernetprotocol/auth": "0.1.92",
|
|
36
|
+
"@infernetprotocol/config": "0.1.92",
|
|
37
|
+
"@infernetprotocol/engine": "0.1.92",
|
|
38
|
+
"@infernetprotocol/deploy-providers": "0.1.92",
|
|
39
|
+
"@infernetprotocol/gpu": "0.1.92",
|
|
40
|
+
"@infernetprotocol/liquid-agent": "0.1.92",
|
|
41
|
+
"@infernetprotocol/training": "0.1.92",
|
|
42
|
+
"@infernetprotocol/rpc-adapter": "0.1.92"
|
|
43
43
|
},
|
|
44
44
|
"repository": {
|
|
45
45
|
"type": "git",
|