mc/mcp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

agent: recover down components on startup when no boot sequence

Browse Source 
A unikernel VM has no runtime restart policy, so if it exits — including
when an agent restart's cgroup kill takes it down — nothing restarts it,
and it sits in drift. Recover() already handles this (and unikernels, via
runtimeFor), but only ran inside RunBootSequence, which is gated on a
[boot] sequence that worker nodes don't define. Now the agent also runs
Recover once in the background on startup when there is no boot sequence,
so desired=running components (VMs especially) come back after an agent
or host restart without delaying registration.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Kyle Isom
2026-06-11 12:58:50 -07:00
parent 4a55972455
commit 98b166fa7b
1 changed files with 15 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
internal/agent/agent.go
View File

@@ -11,6 +11,7 @@ import (
"os/signal" "os/signal"
"path/filepath" "path/filepath"
"syscall" "syscall"
"time"
mcpv1 "git.wntrmute.dev/mc/mcp/gen/mcp/v1" mcpv1 "git.wntrmute.dev/mc/mcp/gen/mcp/v1"
"git.wntrmute.dev/mc/mcp/internal/auth" "git.wntrmute.dev/mc/mcp/internal/auth"
@@ -155,6 +156,20 @@ func Run(cfg *config.AgentConfig, version string) error {
logger.Error("boot sequence failed", "err", err) logger.Error("boot sequence failed", "err", err)
// Continue starting the gRPC server — partial boot is better than no agent. // Continue starting the gRPC server — partial boot is better than no agent.
} }
} else {
// No ordered boot sequence: still reconcile once, in the background, so
// desired=running components are brought back up after an agent or host
// restart without delaying registration. This matters most for
// unikernel VMs: unlike podman containers (which have a restart policy),
// a VM that exits — including when an agent restart's cgroup kill takes
// it down — has nothing to restart it.
go func() {
recCtx, recCancel := context.WithTimeout(context.Background(), 10*time.Minute)
defer recCancel()
if err := a.Recover(recCtx); err != nil {
logger.Error("startup recover failed", "err", err)
}
}()
} }
// Start heartbeat client (registers with master and sends heartbeats). // Start heartbeat client (registers with master and sends heartbeats).