Auto-detect available memory when AvailableMemory is unset and apply MaxMemoryPercentage

config/file_config.go

@@ -11,6 +11,8 @@ import (
 
 	"golang.org/x/exp/slices"
 	"gopkg.in/yaml.v3"
+
+	"github.com/honeycombio/refinery/internal/system"
 )
 
 // In order to be able to unmarshal "15s" etc. into time.Duration, we need to
@@ -544,6 +546,13 @@ func newFileConfig(opts *CmdEnv, cData, rulesData []configData, currentVersion .
 		return nil, err
 	}
 
+	// If AvailableMemory is not set, AND MaxAlloc is not set, try to detect it from the system
+	if mainconf.Collection.AvailableMemory == 0 && mainconf.Collection.MaxAlloc == 0 {
+		if mem, err := system.GetTotalMemory(); err == nil && mem > 0 {
+			allowed := uint64(float64(mem) * float64(mainconf.Collection.MaxMemoryPercentage) / 100.0)
+			mainconf.Collection.MaxAlloc = MemorySize(allowed)
+		}
+	}
 	var rulesconf *V2SamplerConfig
 	ruleshash, err := applyConfigInto(&rulesconf, rulesData, nil)
 	if err != nil {

config/metadata/configMeta.yaml

@@ -1320,8 +1320,9 @@ groups:
         description: >
           This value will typically be set through an environment variable
           controlled by the container or deploy script. If this value is zero or
-          not set, then `MaxMemoryPercentage` cannot be used to calculate the
-          maximum allocation and `MaxAlloc` will be used instead. If set, then
+          not set, Refinery will attempt to detect the total available memory from
+          the system (cgroups or /proc/meminfo). If detection fails, `MaxAlloc`
+          will be used. If set, then
           this must be a memory size. Sizes with standard unit suffixes (such as
           `MB` and `GiB`) and Kubernetes units (such as `M` and `Gi`) are
           supported. Fractional values with a suffix are supported. If
@@ -1338,8 +1339,7 @@ groups:
         validations:
           - type: minimum
             arg: 10
-          - type: requiredWith
-            arg: AvailableMemory
+
         summary: is the maximum percentage of memory that should be allocated by the span collector.
         description: >
           If nonzero, then it must be an integer value between 1 and 100,

internal/system/memory.go

@@ -0,0 +1,82 @@
+package system
+
+import (
+	"bufio"
+	"bytes"
+	"errors"
+	"math"
+	"os"
+	"strconv"
+	"strings"
+)
+
+// GetTotalMemory returns the total available memory in bytes.
+// It tries to read from cgroup limits first, then falls back to system memory.
+func GetTotalMemory() (uint64, error) {
+	// Try cgroup v2
+	if mem, err := getCgroupV2Memory(); err == nil {
+		return mem, nil
+	}
+
+	// Try cgroup v1
+	if mem, err := getCgroupV1Memory(); err == nil {
+		return mem, nil
+	}
+
+	// Fall back to system memory
+	return getSystemMemory()
+}
+
+func getCgroupV2Memory() (uint64, error) {
+	data, err := os.ReadFile("/sys/fs/cgroup/memory.max")
+	if err != nil {
+		return 0, err
+	}
+	s := bytes.TrimSpace(data)
+	if bytes.Equal(s, []byte("max")) {
+		return 0, errors.New("no limit set in cgroup v2")
+	}
+	return strconv.ParseUint(string(s), 10, 64)
+}
+
+func getCgroupV1Memory() (uint64, error) {
+	data, err := os.ReadFile("/sys/fs/cgroup/memory/memory.limit_in_bytes")
+	if err != nil {
+		return 0, err
+	}
+	s := bytes.TrimSpace(data)
+	mem, err := strconv.ParseUint(string(s), 10, 64)
+	if err != nil {
+		return 0, err
+	}
+	// A very large number indicates no limit in cgroup v1
+	if mem > math.MaxInt64 {
+		return 0, errors.New("no limit set in cgroup v1")
+	}
+	return mem, nil
+}
+
+func getSystemMemory() (uint64, error) {
+	file, err := os.Open("/proc/meminfo")
+	if err != nil {
+		return 0, err
+	}
+	defer file.Close()
+
+	scanner := bufio.NewScanner(file)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if strings.HasPrefix(line, "MemTotal:") {
+			parts := strings.Fields(line)
+			if len(parts) < 2 {
+				return 0, errors.New("invalid MemTotal format")
+			}
+			kb, err := strconv.ParseUint(parts[1], 10, 64)
+			if err != nil {
+				return 0, err
+			}
+			return kb * 1024, nil
+		}
+	}
+	return 0, errors.New("MemTotal not found in /proc/meminfo")
+}

refinery_config.md

@@ -884,7 +884,8 @@ Its minimum value should be at least three times the `CacheCapacity`.
 `AvailableMemory` is the amount of system memory available to the Refinery process.
 
 This value will typically be set through an environment variable controlled by the container or deploy script.
-If this value is zero or not set, then `MaxMemoryPercentage` cannot be used to calculate the maximum allocation and `MaxAlloc` will be used instead.
+If this value is zero or not set, Refinery will attempt to detect the total available memory from the system (cgroups or /proc/meminfo).
+If detection fails, `MaxAlloc` will be used.
 If set, then this must be a memory size.
 Sizes with standard unit suffixes (such as `MB` and `GiB`) and Kubernetes units (such as `M` and `Gi`) are supported.
 Fractional values with a suffix are supported.