pillar: configure GOGC based on pillar memory limit or global config

Patch introduces two settings for Golang runtime which
impacts garbage collector behavior:

1. `gogc.memory.limit.bytes` provides the runtime with a soft memory
    limit.  The runtime undertakes several processes to try to respect
    this memory limit, including adjustments to the frequency of
    garbage collections and returning memory to the underlying system
    more aggressively. The Go API call is described here:
    https://pkg.go.dev/runtime/debug#SetMemoryLimit

    By default, EVE setting is disabled (set to 0), meaning the Golang
    runtime memory limit will be set according to the following
    equation based on the `memory.limit_in_bytes` hard memory limit
    provided by the pillar `cgroups`:

    `limit = memory.limit_in_bytes * 0.6`

    The constant 0.6 was chosen empirically and is explained by simple
    logic: `memory.limit_in_bytes` is a hard limit for the whole
    pillar cgroup, meaning when reached, likely one of the processes
    will be killed by OOM. In turn Golang runtime memory limit is a
    soft limit, so the difference must be significant to ensure that
    after the soft limit is reached, there will be enough memory for
    the Go garbage collector to do its job and, fortunately, not to
    hit the hard limit.

2. `gogc.percent` sets the garbage collection target percentage: a
    collection is triggered when the ratio of freshly allocated data
    to live data remaining after the previous collection reaches this
    percentage. The Go API call is described here:
    https://pkg.go.dev/runtime/debug#SetGCPercent

The patch is motivated by a frequently observed bloated `zedbox`
application (up to 500MB) that causes an OOM kill call to the /eve or
/pillar cgroups. It is assumed that the bloated `zedbox` application
is not caused by memory leaks, but by a delayed GC sweep cycle and a
unconditionally growing runtime heap size. An explicit memory limit
set for the Golang runtime (~400MB in the current version of EVE)
should make the GC more aggressive when the soft memory limit is hit,
which should result in a significant reduction in allocated but unused
memory.

Signed-off-by: Roman Penyaev <r.peniaev@gmail.com>

pkg/pillar/cmd/zedmanager/zedmanager.go

@@ -1422,6 +1422,17 @@ func quantifyChanges(config types.AppInstanceConfig, oldConfig types.AppInstance
 	return needPurge, needRestart, purgeReason, restartReason
 }
 
+func configureGOGC(gcp *types.ConfigItemValueMap) {
+	lim := gcp.GlobalValueInt(types.GOGCMemoryLimitInBytes)
+	per := gcp.GlobalValueInt(types.GOGCPercent)
+	plim, pper, err := types.ConfigureGOGC(int64(lim), int(per))
+	if err != nil {
+		log.Warningf("configureGOGC: failed '%v'", err)
+	} else {
+		log.Functionf("configureGOGC: memory limit set to '%v' (previous '%v'), GC percent set to '%v' (previous '%v')", lim, plim, per, pper)
+	}
+}
+
 func handleGlobalConfigCreate(ctxArg interface{}, key string,
 	statusArg interface{}) {
 	handleGlobalConfigImpl(ctxArg, key, statusArg)
@@ -1447,6 +1458,7 @@ func handleGlobalConfigImpl(ctxArg interface{}, key string,
 		ctx.globalConfig = gcp
 		ctx.GCInitialized = true
 	}
+	configureGOGC(gcp)
 	log.Functionf("handleGlobalConfigImpl done for %s", key)
 }
 

pkg/pillar/types/global.go

@@ -231,6 +231,10 @@ const (
 	EveMemoryLimitInBytes GlobalSettingKey = "memory.eve.limit.bytes"
 	// How much memory overhead is allowed for VMM needs
 	VmmMemoryLimitInMiB GlobalSettingKey = "memory.vmm.limit.MiB"
+	// GOGCMemoryLimitInBytes global setting key
+	GOGCMemoryLimitInBytes GlobalSettingKey = "gogc.memory.limit.bytes"
+	// GOGCPercent global setting key
+	GOGCPercent GlobalSettingKey = "gogc.percent"
 	// IgnoreMemoryCheckForApps global setting key
 	IgnoreMemoryCheckForApps GlobalSettingKey = "memory.apps.ignore.check"
 	// IgnoreDiskCheckForApps global setting key
@@ -892,7 +896,10 @@ func NewConfigItemSpecMap() ConfigItemSpecMap {
 		100*1024*1024, 0xFFFFFFFF)
 	configItemSpecMap.AddIntItem(StorageZfsReserved, 20, 1, 99)
 	configItemSpecMap.AddIntItem(ForceFallbackCounter, 0, 0, 0xFFFFFFFF)
-
+	// Default GOGC memory limit is 0
+	configItemSpecMap.AddIntItem(GOGCMemoryLimitInBytes, 0, 0, 0xFFFFFFFF)
+	// Default GOGC target percentage is 100, 0 means disable GC
+	configItemSpecMap.AddIntItem(GOGCPercent, 100, 0, 500)
 	configItemSpecMap.AddIntItem(EveMemoryLimitInBytes, uint32(eveMemoryLimitInBytes),
 		uint32(eveMemoryLimitInBytes), 0xFFFFFFFF)
 	// Limit manual vmm overhead override to 1 PiB

pkg/pillar/types/global_test.go

@@ -183,6 +183,8 @@ func TestNewConfigItemSpecMap(t *testing.T) {
 		ConsoleAccess,
 		VncShimVMAccess,
 		AllowAppVnc,
+		GOGCMemoryLimitInBytes,
+		GOGCPercent,
 		EveMemoryLimitInBytes,
 		VmmMemoryLimitInMiB,
 		IgnoreMemoryCheckForApps,

pkg/pillar/types/locationconsts.go

@@ -98,6 +98,8 @@ const (
 	NewlogUploadAppDir = NewlogDir + "/appUpload"
 	// NewlogKeepSentQueueDir - a circular queue of gzip files already been sent
 	NewlogKeepSentQueueDir = NewlogDir + "/keepSentQueue"
+	// PillarHardMemoryLimitFile - hard memory reserved for pillar
+	PillarHardMemoryLimitFile = "/hostfs/sys/fs/cgroup/memory/eve/services/pillar/memory.limit_in_bytes"
 	// EveMemoryLimitFile - stores memory reserved for eve
 	EveMemoryLimitFile = "/hostfs/sys/fs/cgroup/memory/eve/memory.soft_limit_in_bytes"
 	// EveMemoryUsageFile - current usage

pkg/pillar/types/memory.go

@@ -4,11 +4,19 @@
 package types
 
 import (
+	"fmt"
 	"os"
+	"runtime/debug"
 	"strconv"
 	"strings"
 )
 
+// GetPillarHardMemoryLimitInBytes returns hard memory limit
+// reserved for pillar in bytes
+func GetPillarHardMemoryLimitInBytes() (uint64, error) {
+	return readUint64File(PillarHardMemoryLimitFile)
+}
+
 // GetEveMemoryLimitInBytes returns memory limit
 // reserved for eve in bytes
 func GetEveMemoryLimitInBytes() (uint64, error) {
@@ -42,3 +50,36 @@ func readUint64File(filename string) (uint64, error) {
 	dataUint64, err := strconv.ParseUint(dataString, 10, 64)
 	return dataUint64, err
 }
+
+// ConfigureGOGC sets two main configuration parameters for the
+// garbage collector (GOGC): memory limit and percentage (see
+// explanation here: https://tip.golang.org/doc/gc-guide).
+// If limit is 0, create GOGC limit from the pillar cgroups hard
+// memory limit.
+func ConfigureGOGC(limit int64, percent int) (int64, int, error) {
+	if limit == 0 {
+		// Fallback to value from cgroups if no limit in the configuration
+		ulimit, err := GetPillarHardMemoryLimitInBytes()
+		if err != nil {
+			err := fmt.Errorf("can't receive pillar memory hard limit: '%w'", err)
+			return -1, -1, err
+		}
+		// Reduce actual memory limit to 0.6 of cgroup limit. The logic behind
+		// the constant is simple: cgroup limit is a hard limit for the whole
+		// pillar cgroup, meaning when reached, we are killed by OOM. In turn
+		// GOGC memory limit is a soft limit, so the difference must be
+		// significant to ensure that after the soft limit is reached, there
+		// will be enough memory for the GOGC to do its job and, fortunately,
+		// not to hit the hard limit.
+		limit = int64(ulimit) * 600 / 1000
+	}
+	if percent == 0 {
+		// Disable GC
+		percent = -1
+	}
+	// Set new and retrieve previous values
+	limit = debug.SetMemoryLimit(limit)
+	percent = debug.SetGCPercent(percent)
+
+	return limit, percent, nil
+}