Add internal profiling to debug tool

doc/user/data/mz-debug/emulator_options.yml

@@ -2,24 +2,16 @@ columns:
   - column: "Option"
   - column: "Description"
 rows:
-  - Option: "`--dump-docker <boolean>`"
-    Description: |
-
-      <a name="dump-docker"></a> If `true`, dump debug information from the Docker container.
-
-      Defaults to `true`.
-
   - Option: "`--docker-container-id <ID>`"
     Description: |
 
       <a name="docker-container-id"></a> The Docker container to dump.
 
-      Required if [`--dump-docker`](#dump-docker) is true.
+      Required.
 
-  - Option: "`--mz-connection-url <URL>`"
+  - Option: "`--dump-docker <boolean>`"
     Description: |
 
-      <a name="mz-connection-url"></a>The URL of the Materialize's SQL
-      connection.
+      <a name="dump-docker"></a> If `true`, dump debug information from the Docker container.
 
-      Defaults to `postgres://127.0.0.1:6875/materialize?sslmode=prefer`.
+      Defaults to `true`.

doc/user/data/mz-debug/mz_debug_option.yml

@@ -2,10 +2,36 @@ columns:
   - column: "Option"
   - column: "Description"
 rows:
+  - Option: "`--dump-heap-profiles <boolean>`"
+    Description: |
+
+      <a name="dump-heap-profiles"></a> If `true`, dump heap profiles (.pprof.gz) from
+      your Materialize instance.
+
+      Defaults to `true`.
+  - Option: "`--dump-prometheus-metrics <boolean>`"
+    Description: |
+
+      <a name="dump-prometheus-metrics"></a> If `true`, dump prometheus metrics from
+      your Materialize instance.
+
+      Defaults to `true`.
+
+
   - Option: "`--dump-system-catalog <boolean>`"
     Description: |
 
       <a name="dump-system-catalog"></a> If `true`, dump the system catalog from
       your Materialize instance.
 
       Defaults to `true`.
+
+
+  - Option: "`--mz-connection-url <URL>`"
+    Description: |
+
+      <a name="mz-connection-url"></a>The Materialize instance's [PostgreSQL
+      connection
+      URL](https://www.postgresql.org/docs/14/libpq-connect.html#LIBPQ-CONNSTRING).
+
+      Defaults to `postgres://127.0.0.1:6875/materialize?sslmode=prefer`.

doc/user/data/mz-debug/self_managed_options.yml

@@ -57,18 +57,6 @@ rows:
 
       Defaults to `6875`.
 
-  - Option: "`--mz-connection-url <URL>`"
-    Description: |
-
-      <a name="mz-connection-url"></a>The Materialize instance's [PostgreSQL
-      connection
-      URL](https://www.postgresql.org/docs/14/libpq-connect.html#LIBPQ-CONNSTRING).
-
-      Defaults to a connection URL constructed from:
-
-      [`--port-forward-local-address`](#port-forward-local-address) and
-      [`--port-forward-local-port`](#port-forward-local-port) values.
-
   - Option: "`-h`, `--help`"
     Description: |
 

src/mz-debug/BUILD.bazel

@@ -34,7 +34,7 @@ rust_binary(
         "@//misc/bazel/platforms:xlang_lto_enabled": ["-Clinker-plugin-lto"],
         "//conditions:default": [],
     }),
-    version = "0.1.0",
+    version = "0.2.0",
     deps = [
         "//src/build-info:mz_build_info",
         "//src/cloud-resources:mz_cloud_resources",

src/mz-debug/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "mz-debug"
 description = "Debug tool for self-managed Materialize."
-version = "0.1.0"
+version = "0.2.0"
 edition.workspace = true
 rust-version.workspace = true
 publish = false
@@ -22,13 +22,15 @@ mz-cloud-resources = { path = "../cloud-resources"}
 mz-ore = { path = "../ore", features = ["cli", "test"] }
 mz-tls-util = { path = "../tls-util" }
 postgres-openssl = { version = "0.5.0" }
+reqwest = { version = "0.11", features = ["stream"] }
 serde = "1.0.219"
 serde_yaml = "0.9.34"
 tokio = "1.44.1"
 tokio-postgres = { version = "0.7.8" }
 tokio-util = { version = "0.7.15", features = ["io"] }
 tracing = "0.1.37"
 tracing-subscriber = { version = "0.3.19", default-features = false, features = ["env-filter", "fmt"] }
+url = { version = "2.3.1", features = ["serde"] }
 walkdir = "2.5"
 workspace-hack = { version = "0.0.0", path = "../workspace-hack", optional = true }
 zip = { version = "2.6", default-features = false, features=["deflate-flate2", "flate2"]}

src/mz-debug/src/docker_dumper.rs

@@ -20,12 +20,13 @@ use std::io::Write;
 use std::path::PathBuf;
 use std::time::Duration;
 
+use anyhow::Context as AnyhowContext;
 use mz_ore::retry::{self, RetryResult};
 use tracing::{info, warn};
 
-use crate::utils::format_base_path;
 use crate::{ContainerDumper, Context};
 
+static DOCKER_DUMP_DIR: &str = "docker";
 static DOCKER_RESOURCE_DUMP_TIMEOUT: Duration = Duration::from_secs(30);
 
 pub struct DockerDumper {
@@ -36,9 +37,7 @@ pub struct DockerDumper {
 impl DockerDumper {
     pub fn new(context: &Context, container_id: String) -> Self {
         Self {
-            directory_path: format_base_path(context.start_time)
-                .join("docker")
-                .join(&container_id),
+            directory_path: context.base_path.join(DOCKER_DUMP_DIR).join(&container_id),
             container_id,
         }
     }
@@ -149,3 +148,34 @@ fn write_output(
     info!("Exported {}", file_path.display());
     Ok(())
 }
+
+/// Gets the IP address of a Docker container using the container ID.
+pub async fn get_container_ip(container_id: &str) -> Result<String, anyhow::Error> {
+    let output = tokio::process::Command::new("docker")
+        .args([
+            "inspect",
+            "-f",
+            "{{range .NetworkSettings.Networks}}{{.IPAddress}}{{end}}",
+            container_id,
+        ])
+        .output()
+        .await
+        .with_context(|| format!("Failed to get container IP address for {}", container_id))?;
+
+    if !output.status.success() {
+        return Err(anyhow::anyhow!(
+            "Docker command failed: {}",
+            String::from_utf8_lossy(&output.stderr)
+        ));
+    }
+
+    let ip = String::from_utf8(output.stdout)
+        .with_context(|| "Failed to convert container IP address to string")?
+        .trim()
+        .to_string();
+    if ip.is_empty() {
+        return Err(anyhow::anyhow!("Container IP address not found"));
+    }
+
+    Ok(ip)
+}