Let a node describe itself instead of being described by hand
CI / test (push) Successful in 1m47s

Every agent file on the cluster named its own node, its own address, the
path to llama-server and two weight directories — different on every node,
kept true by hand, and not true: node05 carried "ram: 32g" beside a measured
31 GB, looked for weights in an empty directory while fetching them into
another, and the other four never looked in theirs at all, so four nodes
reported no disk. Growing a node's RAM would have gone unnoticed, because
memory was read once, at registration.

The agent now finds these out. Its id is the hostname, which is also the
certificate's common name. Its address is the one on the route to the
control plane, which the kernel picks for a UDP dial without sending
anything. llama-server is looked for on PATH and then where release archives
are unpacked. A missing nvidia-smi is a node without NVIDIA cards rather
than a failed discovery — which also cut discovery short before InfiniBand
on every CPU node not told to skip GPUs. Fetched weights land in
/var/lib/llmgw-weights by default, and wherever they land is always looked
in.

Memory is read on every heartbeat, and installed engines are probed again
each minute and sent when they change, so a VM given more RAM or a node
given vLLM becomes a candidate without anyone restarting its agent. The
control plane logs both changes, since "the node now qualifies" is the
answer to why a model placed an hour after it was asked for.

What is written still wins. The agent file shrinks to the control plane's
addresses and the TLS files, the same on every node; onboarding writes only
the overrides it was given. What cannot be found stays configuration: which
fabric an adapter is on, because OpenSM's default prefix is the same on all
of them.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Ilker ManapandClaude Opus 5.5 committed 2026-09-28 10:13:42 +02:00
1 parent 2cae92d113
commit 20deb22dfa
14 files changed
+552 -134

No files matched your search

+44 -1
View File
@@ -1,6 +1,7 @@
package agentsvc
import (
"log/slog"
"testing"
llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
@@ -11,7 +12,8 @@ import (
func nodeWith(s *Service, n inventory.Node) { s.inv.Upsert(n) }
func newTestService() *Service {
return &Service{inv: inventory.NewStore(), conns: map[string]*conn{}}
return &Service{inv: inventory.NewStore(), conns: map[string]*conn{},
log: slog.New(slog.DiscardHandler)}
}
func TestDiskIsRefreshedWhileTheAgentRuns(t *testing.T) {
@@ -112,3 +114,44 @@ func TestAHeartbeatAboutNothingChangesNothing(t *testing.T) {
t.Errorf("disk changed: %d", free)
}
}
func TestAHeartbeatCarriesMemoryAddedToARunningNode(t *testing.T) {
// Read once at registration, a node that was given 32 GB went on looking
// like the 7 GB machine it had been, and every model that needed the
// difference stayed unplaced until somebody restarted its agent.
s := newTestService()
nodeWith(s, inventory.Node{ID: "n1", State: inventory.Ready, MemoryTotalMB: 7 * 1024})
s.applyHeartbeat("n1", &llmgwv1.Heartbeat{NodeId: "n1", MemoryTotalMb: 32 * 1024})
if n, _ := s.inv.Get("n1"); n.MemoryTotalMB != 32*1024 {
t.Errorf("memory is %d MB, expected the 32 GB the node now has", n.MemoryTotalMB)
}
// Zero is an agent that could not read it, not a node that lost its RAM.
s.applyHeartbeat("n1", &llmgwv1.Heartbeat{NodeId: "n1"})
if n, _ := s.inv.Get("n1"); n.MemoryTotalMB != 32*1024 {
t.Errorf("a heartbeat without memory erased it: %d", n.MemoryTotalMB)
}
}
func TestAHeartbeatCarriesAnEngineInstalledAfterRegistration(t *testing.T) {
s := newTestService()
nodeWith(s, inventory.Node{ID: "n1", State: inventory.Ready,
EngineVersions: map[string]string{"llamacpp": "unknown"}})
// Unchanged engines are not resent, and silence must not clear them.
s.applyHeartbeat("n1", &llmgwv1.Heartbeat{NodeId: "n1", MemoryTotalMb: 1024})
if n, _ := s.inv.Get("n1"); len(n.EngineVersions) != 1 {
t.Fatalf("engines changed without being reported: %v", n.EngineVersions)
}
s.applyHeartbeat("n1", &llmgwv1.Heartbeat{
NodeId: "n1", EnginesChanged: true,
EngineVersions: map[string]string{"llamacpp": "unknown", "vllm": "unknown"},
})
n, _ := s.inv.Get("n1")
if _, ok := n.EngineVersions["vllm"]; !ok {
t.Errorf("the newly installed engine is not known: %v", n.EngineVersions)
}
}
+24 -1
View File
@@ -6,6 +6,7 @@ import (
"fmt"
"io"
"log/slog"
"maps"
"net/http"
"slices"
"sync"
@@ -269,7 +270,9 @@ func (s *Service) applyHeartbeat(nodeID string, hb *llmgwv1.Heartbeat) {
used := hb.GetGpuMemoryUsedMb()
problems := hb.GetNewProblems()
storage := hb.GetStorage()
if len(used) == 0 && len(problems) == 0 && len(storage) == 0 && !hb.GetWeightsChanged() {
memMB := int(hb.GetMemoryTotalMb())
if len(used) == 0 && len(problems) == 0 && len(storage) == 0 &&
!hb.GetWeightsChanged() && memMB == 0 && !hb.GetEnginesChanged() {
return
}
@@ -316,6 +319,26 @@ func (s *Service) applyHeartbeat(nodeID string, hb *llmgwv1.Heartbeat) {
})
}
}
// RAM and engines change under a running agent too: memory added to a VM,
// an engine installed this afternoon. Said in the log when they do, because
// "the node now qualifies" is the answer to why a model placed an hour after
// it was asked for, and "the node no longer does" is the answer to why one
// stopped placing.
if memMB > 0 && memMB != n.MemoryTotalMB {
s.log.Info("node memory changed", "node", nodeID,
"from_mb", n.MemoryTotalMB, "to_mb", memMB)
n.MemoryTotalMB = memMB
}
if hb.GetEnginesChanged() {
vers := maps.Clone(hb.GetEngineVersions())
if !maps.Equal(vers, n.EngineVersions) {
s.log.Info("node engines changed", "node", nodeID,
"from", slices.Sorted(maps.Keys(n.EngineVersions)),
"to", slices.Sorted(maps.Keys(vers)))
}
n.EngineVersions = vers
}
s.inv.Upsert(n)
}