Skip to content

platformd: remove WaitAfterServerInit config option #326

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
freggy merged 2 commits into from
May 23, 2026
+41 −45
Merged

platformd: remove WaitAfterServerInit config option #326

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
7d15576
remove WaitAfterServerInit config option
freggy May 23, 2026
5ee53fd
forgot
freggy May 23, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion Makefile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -131,7 +131,7 @@ functests-shared: $(TEST_IMG)
.PHONY: ebpf
ebpf:
$(RUN) $(SUDO) go generate ./internal/datapath
$(RUN) $(SUDO) go test ./internal/datapath -run TestVerifier -count 1
$(RUN) $(SUDO) go test -tags bpf ./internal/datapath -run TestVerifier -count 1

$(TEST_IMG):
@docker build -t test-img -f $(IMG_TESTDATA_DIR)/Dockerfile $(IMG_TESTDATA_DIR)
Expand Down
56 changes: 27 additions & 29 deletions cmd/platformd/main.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,34 +25,33 @@ func main() {
mgmtListenSock = fs.String("management-server-listen-sock", "/run/platformd/platformd.sock", "path to the unix domain socket to listen on") //nolint:lll
mgmtSockUID = fs.Uint64("management-server-listen-sock-uid", 9012, "unix domain socket uid")
mgmtSockGID = fs.Uint64("management-server-listen-sock-gid", 9012, "unix domain socket gid")
criListenSock = fs.String("cri-listen-sock", "/var/run/crio/crio.sock", "path to the unix domain socket the CRI is listening on") //nolint:lll
envoyImage = fs.String("envoy-image", "", "container image to use for envoy") //nolint:lll
coreDNSImage = fs.String("coredns-image", "", "container image to use for CoreDNS") //nolint:lll
getsockoptCgroup = fs.String("getsockopt-cgroup", "", "container image to use for coredns") //nolint:lll
dnsServer = fs.String("dns-server", "", "dns server used by the containers") //nolint:lll
hostIface = fs.String("host-iface", "", "internet-facing network interface for ingress and egress traffic") //nolint:lll
maxAttempts = fs.Uint("max-attempts", 5, "maximum number of attempts workload creation attempts") //nolint:lll
syncInterval = fs.Duration("sync-interval", 200*time.Millisecond, "i") //nolint:lll
nodeID = fs.String("node-id", "", "unique node id") //nolint:lll
minPort = fs.Uint("min-port", 30000, "start of the port range") //nolint:lll
maxPort = fs.Uint("max-port", 40000, "end of the port range") //nolint:lll
workloadNamespace = fs.String("workload-namespace", "", "namespace where the workload is deployed") //nolint:lll
registryEndpoint = fs.String("registry-endpoint", "", "registry endpoint where base images will be pulled from and checkpoints pushed to") //nolint:lll
registryUser = fs.String("registry-user", "", "user for the registry") //nolint:lll
registryPass = fs.String("registry-password", "", "password for the registry") //nolint:lll
controlPlaneEndpoint = fs.String("control-plane-endpoint", "", "control plane endpoint") //nolint:lll
checkCPUPeriod = fs.Uint64("checkpoint-cpu-period", 0, "period of checking CPU period") //nolint:lll
checkCPUQuota = fs.Uint64("checkpoint-cpu-quota", 0, "quota of checking CPU quota") //nolint:lll
checkMemoryLimitInBytes = fs.Uint64("checkpoint-memory-limit-bytes", 0, "memory limit of the container that will be checkpointed") //nolint:lll
checkLocationDir = fs.String("checkpoint-file-dir", "/tmp/platformd", "directory where checkpoint files will be stored") //nolint:lll
checkTimeout = fs.Uint64("checkpoint-timeout-seconds", 60, "timeout for checkpoint creation") //nolint:lll
checkListenAddr = fs.String("checkpoint-listen-addr", "", "timeout for checkpoint creation") //nolint:lll
checkStatusRetentionDuration = fs.Duration("checkpoint-status-retention-period", 1*time.Minute, "timeout for checkpoint creation") //nolint:lll
checkContainerReadyTimeout = fs.Duration("checkpoint-container-ready-timeout", 1*time.Minute, "maximum time to wait until the container is ready for checkpointing") //nolint:lll
checkWaitAfterServerInit = fs.Duration("checkpoint-wait-server-init", 10*time.Second, "how long to wait before performing a checkpoint after server has initialized ") //nolint:lll
mcServerManagementAPIToken = fs.String("mc-server-management-api-token", "", "token to use for the minecraft server management api") //nolint:lll
serverMonImage = fs.String("servermon-image", "", "image to use for the servermon container") //nolint:lll
_ = fs.String("config", "/etc/platformd/config.json", "path to the config file") //nolint:lll
criListenSock = fs.String("cri-listen-sock", "/var/run/crio/crio.sock", "path to the unix domain socket the CRI is listening on") //nolint:lll
envoyImage = fs.String("envoy-image", "", "container image to use for envoy") //nolint:lll
coreDNSImage = fs.String("coredns-image", "", "container image to use for CoreDNS") //nolint:lll
getsockoptCgroup = fs.String("getsockopt-cgroup", "", "container image to use for coredns") //nolint:lll
dnsServer = fs.String("dns-server", "", "dns server used by the containers") //nolint:lll
hostIface = fs.String("host-iface", "", "internet-facing network interface for ingress and egress traffic") //nolint:lll
maxAttempts = fs.Uint("max-attempts", 5, "maximum number of attempts workload creation attempts") //nolint:lll
syncInterval = fs.Duration("sync-interval", 200*time.Millisecond, "i") //nolint:lll
nodeID = fs.String("node-id", "", "unique node id") //nolint:lll
minPort = fs.Uint("min-port", 30000, "start of the port range") //nolint:lll
maxPort = fs.Uint("max-port", 40000, "end of the port range") //nolint:lll
workloadNamespace = fs.String("workload-namespace", "", "namespace where the workload is deployed") //nolint:lll
registryEndpoint = fs.String("registry-endpoint", "", "registry endpoint where base images will be pulled from and checkpoints pushed to") //nolint:lll
registryUser = fs.String("registry-user", "", "user for the registry") //nolint:lll
registryPass = fs.String("registry-password", "", "password for the registry") //nolint:lll
controlPlaneEndpoint = fs.String("control-plane-endpoint", "", "control plane endpoint") //nolint:lll
checkCPUPeriod = fs.Uint64("checkpoint-cpu-period", 0, "period of checking CPU period") //nolint:lll
checkCPUQuota = fs.Uint64("checkpoint-cpu-quota", 0, "quota of checking CPU quota") //nolint:lll
checkMemoryLimitInBytes = fs.Uint64("checkpoint-memory-limit-bytes", 0, "memory limit of the container that will be checkpointed") //nolint:lll
checkLocationDir = fs.String("checkpoint-file-dir", "/tmp/platformd", "directory where checkpoint files will be stored") //nolint:lll
checkTimeout = fs.Uint64("checkpoint-timeout-seconds", 60, "timeout for checkpoint creation") //nolint:lll
checkListenAddr = fs.String("checkpoint-listen-addr", "", "timeout for checkpoint creation") //nolint:lll
checkStatusRetentionDuration = fs.Duration("checkpoint-status-retention-period", 1*time.Minute, "timeout for checkpoint creation") //nolint:lll
checkContainerReadyTimeout = fs.Duration("checkpoint-container-ready-timeout", 1*time.Minute, "maximum time to wait until the container is ready for checkpointing") //nolint:lll
mcServerManagementAPIToken = fs.String("mc-server-management-api-token", "", "token to use for the minecraft server management api") //nolint:lll
serverMonImage = fs.String("servermon-image", "", "image to use for the servermon container") //nolint:lll
_ = fs.String("config", "/etc/platformd/config.json", "path to the config file") //nolint:lll
)
if err := ff.Parse(fs, os.Args[1:],
ff.WithEnvVarPrefix("PLATFORMD"),
Expand Down Expand Up @@ -97,7 +96,6 @@ func main() {
ListenAddr: *checkListenAddr,
StatusRetentionPeriod: *checkStatusRetentionDuration,
ContainerReadyTimeout: *checkContainerReadyTimeout,
WaitAfterServerInit: *checkWaitAfterServerInit,
},
ManagementSocketUID: *mgmtSockUID,
ManagementSocketGID: *mgmtSockGID,
Expand Down
12 changes: 7 additions & 5 deletions internal/datapath/objects.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -127,8 +127,9 @@ func (o *Objects) BlockIP4Connections(cgroupPath string) error {
Attach: ebpf.AttachCGroupInet4Connect,
Path: cgroupPath,
}); err != nil {
return err
return fmt.Errorf("attach: %w", err)
}

return nil
}

Expand All @@ -138,8 +139,9 @@ func (o *Objects) BlockIP6Connections(cgroupPath string) error {
Attach: ebpf.AttachCGroupInet6Connect,
Path: cgroupPath,
}); err != nil {
return err
return fmt.Errorf("attach: %w", err)
}

return nil
}

Expand All @@ -155,7 +157,7 @@ func (o *Objects) AttachAndPinSNAT(iface *net.Interface) error {

// pin because cni is short-lived
if err := l.Pin(fmt.Sprintf("%s/snat_%s", ProgPinPath, iface.Name)); err != nil {
return fmt.Errorf("pin link: %w", err)
return fmt.Errorf("pin: %w", err)
}

return nil
Expand All @@ -177,7 +179,7 @@ func (o *Objects) AttachAndPinDNAT(iface *net.Interface) error {
// TODO: update prog
return nil
}
return fmt.Errorf("pin link: %w", err)
return fmt.Errorf("pin: %w", err)
}

return nil
Expand All @@ -195,7 +197,7 @@ func (o *Objects) AttachAndPinARP(iface *net.Interface) error {

// pin because cni is short-lived
if err := l.Pin(fmt.Sprintf("%s/arp_%s", ProgPinPath, iface.Name)); err != nil {
return fmt.Errorf("pin link: %w", err)
return fmt.Errorf("pin: %w", err)
}

return nil
Expand Down
Binary file modified internal/datapath/sock_bpfeb.o
View file
Open in desktop
Binary file not shown.
Binary file modified internal/datapath/sock_bpfel.o
View file
Open in desktop
Binary file not shown.
1 change: 0 additions & 1 deletion platformd/checkpoint/config.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,5 +33,4 @@ type Config struct {
ListenAddr string
StatusRetentionPeriod time.Duration
ContainerReadyTimeout time.Duration
WaitAfterServerInit time.Duration
}
13 changes: 6 additions & 7 deletions platformd/checkpoint/service.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -274,17 +274,16 @@ func (s *ServiceImpl) checkpoint(ctx context.Context, id string, baseRef name.Re
return fmt.Errorf("find netns: %w", err)
}

// sockets in state TCP_TIME_WAIT and TCP_CLOSE_WAIT, will be a left-over
// because our ebpf programs will not be able to close them.
//
// in order to make restoring the checkpoint work, we have to specify
// the --tcp-close option in /etc/criu/crun.conf to make it work.
// see https://criu.org/CLI/opt/--tcp-close for reference.
if err := s.sockHandler.DestroySocks(netnsPath); err != nil {
return fmt.Errorf("kill sockets: %w", err)
}

// wait a little after closing all sockets, as it seems that if we directly checkpoint we still get:
//
// Error (criu/sk-inet.c:191): inet: Connected TCP socket, consider using --tcp-established option.
//
// linux might take some time to close all sockets eventually.
time.Sleep(s.cfg.WaitAfterServerInit)

logger.InfoContext(ctx, "checkpointing container", "container_id", ctrID)

if _, err := s.criService.CheckpointContainer(ctx, &runtimev1.CheckpointContainerRequest{
Expand Down
1 change: 0 additions & 1 deletion platformd/config.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,6 @@ type Config struct {
ListenAddr string
StatusRetentionPeriod time.Duration
ContainerReadyTimeout time.Duration
WaitAfterServerInit time.Duration
}
ManagementSocketUID uint64
ManagementSocketGID uint64
Expand Down
1 change: 0 additions & 1 deletion platformd/server.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,7 +145,6 @@ func (s *Server) Run(ctx context.Context, cfg Config) error {
ListenAddr: cfg.CheckpointConfig.ListenAddr,
StatusRetentionPeriod: cfg.CheckpointConfig.StatusRetentionPeriod,
ContainerReadyTimeout: cfg.CheckpointConfig.ContainerReadyTimeout,
WaitAfterServerInit: cfg.CheckpointConfig.WaitAfterServerInit,
},
criSvc,
image.NewService(checkSvcLogger, cfg.RegistryUser, cfg.RegistryPass, "/tmp"),
Expand Down