diff --git a/services/horizon/CHANGELOG.md b/services/horizon/CHANGELOG.md
index 8ff6e84d49..bd173723cb 100644
--- a/services/horizon/CHANGELOG.md
+++ b/services/horizon/CHANGELOG.md
@@ -12,6 +12,7 @@ This is a minor release with no DB Schema migrations nor explicit state rebuild.
 * Use the correct asset when calculating liquidity pool disbursements ([4018](https://github.com/stellar/go/pull/4018))
 * Make sure Stellar-Core is not started before previous instance termination ([4020](https://github.com/stellar/go/pull/4020))
 * Add a new feature flag `--ingest-enable-extended-log-ledger-stats` (`false` by default) that enables extra ledger stats when logging ledger processing info ([4017](https://github.com/stellar/go/pull/4017))
+* Add a new command `horizon record-metrics` that records `:[ADMIN_PORT]/metrics` into a zip file for debugging purposes ([4023](https://github.com/stellar/go/pull/4023))
 * Expose the `Latest-Ledger` header to browser web pages ([3995](https://github.com/stellar/go/pull/3995))
 * Correct `horizon db reingest range` output command name when invoking `horizon db detect-gaps` ([4007](https://github.com/stellar/go/pull/4007))
 * Add new prometheus metrics:
diff --git a/services/horizon/cmd/record_metrics.go b/services/horizon/cmd/record_metrics.go
new file mode 100644
index 0000000000..a69ae379c8
--- /dev/null
+++ b/services/horizon/cmd/record_metrics.go
@@ -0,0 +1,89 @@
+package cmd
+
+import (
+	"archive/zip"
+	"fmt"
+	"io"
+	"net/http"
+	"os"
+	"time"
+
+	"github.com/spf13/cobra"
+	horizon "github.com/stellar/go/services/horizon/internal"
+	"github.com/stellar/go/support/errors"
+	"github.com/stellar/go/support/log"
+)
+
+var recordMetricsCmd = &cobra.Command{
+	Use:   "record-metrics",
+	Short: "records `/metrics` on admin port for debuging purposes",
+	Long:  "",
+	RunE: func(cmd *cobra.Command, args []string) error {
+		if err := horizon.ApplyFlags(config, flags, horizon.ApplyOptions{}); err != nil {
+			return err
+		}
+
+		const (
+			timeFormat            = "2006-01-02-15-04-05"
+			scrapeIntervalSeconds = 15
+			scrapesCount          = (60 / scrapeIntervalSeconds) * 10 // remember about rounding if change is required
+		)
+
+		client := &http.Client{
+			Timeout: 2 * time.Second,
+		}
+
+		outputFileName := fmt.Sprintf("./metrics-%s.zip", time.Now().Format(timeFormat))
+		outputFile, err := os.Create(outputFileName)
+		if err != nil {
+			return err
+		}
+
+		w := zip.NewWriter(outputFile)
+		defer w.Close()
+
+		for i := 1; i <= scrapesCount; i++ {
+			log.Infof(
+				"Getting metrics %d/%d... ETA: %s",
+				i,
+				scrapesCount,
+				time.Duration(time.Duration(scrapeIntervalSeconds*(scrapesCount-i))*time.Second),
+			)
+
+			metricsResponse, err := client.Get(fmt.Sprintf("http://127.0.0.1:%d/metrics", config.AdminPort))
+			if err != nil {
+				return errors.Wrap(err, "Error fetching metrics. Is admin server running?")
+			}
+
+			if metricsResponse.StatusCode != http.StatusOK {
+				return errors.Errorf("Invalid status code: %d. Is admin server running?", metricsResponse.StatusCode)
+			}
+
+			metricsFile, err := w.Create(time.Now().Format(timeFormat))
+			if err != nil {
+				return err
+			}
+
+			if _, err = io.Copy(metricsFile, metricsResponse.Body); err != nil {
+				return errors.Wrap(err, "Error reading response body. Is admin server running?")
+			}
+
+			// Flush to keep memory usage log and save at least some records in case of errors later.
+			err = w.Flush()
+			if err != nil {
+				return err
+			}
+
+			if i < scrapesCount {
+				time.Sleep(scrapeIntervalSeconds * time.Second)
+			}
+		}
+
+		log.Infof("Metrics recorded to %s!", outputFileName)
+		return nil
+	},
+}
+
+func init() {
+	RootCmd.AddCommand(recordMetricsCmd)
+}