Add 'zpool status -e' flag to see unhealthy vdevs

When very large pools are present, it can be laborious to find reasons for why a pool is degraded and/or where an unhealthy vdev is. This option filters out vdevs that are ONLINE and with no errors to make it easier to see where the issues are. Root and parents of unhealthy vdevs will always be printed. Testing: ZFS errors and drive failures for multiple vdevs were simulated with zinject. Sample vdev listings with '-e' option - All vdevs healthy NAME STATE READ WRITE CKSUM iron5 ONLINE 0 0 0 - ZFS errors NAME STATE READ WRITE CKSUM iron5 ONLINE 0 0 0 raidz2-5 ONLINE 1 0 0 L23 ONLINE 1 0 0 L24 ONLINE 1 0 0 L37 ONLINE 1 0 0 - Vdev faulted NAME STATE READ WRITE CKSUM iron5 DEGRADED 0 0 0 raidz2-6 DEGRADED 0 0 0 L67 FAULTED 0 0 0 too many errors - Vdev faults and data errors NAME STATE READ WRITE CKSUM iron5 DEGRADED 0 0 0 raidz2-1 DEGRADED 0 0 0 L2 FAULTED 0 0 0 too many errors raidz2-5 ONLINE 1 0 0 L23 ONLINE 1 0 0 L24 ONLINE 1 0 0 L37 ONLINE 1 0 0 raidz2-6 DEGRADED 0 0 0 L67 FAULTED 0 0 0 too many errors - Vdev missing NAME STATE READ WRITE CKSUM iron5 DEGRADED 0 0 0 raidz2-6 DEGRADED 0 0 0 L67 UNAVAIL 3 1 0 - Slow devices when -s provided with -e NAME STATE READ WRITE CKSUM SLOW iron5 DEGRADED 0 0 0 - raidz2-5 DEGRADED 0 0 0 - L10 FAULTED 0 0 0 0 external device fault L51 ONLINE 0 0 0 14 Reviewed-by: Tony Hutter <hutter2@llnl.gov> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Cameron Harr <harr1@llnl.gov> Closes #15769
2026-05-22 02:27:36 +03:00 · 2024-02-07 09:12:12 -08:00
parent 9bb8d26bd5
commit 40e20d808c
7 changed files with 169 additions and 7 deletions
@@ -2161,6 +2161,7 @@ typedef struct status_cbdata {
 	boolean_t	cb_explain;
 	boolean_t	cb_first;
 	boolean_t	cb_dedup_stats;
+	boolean_t	cb_print_unhealthy;
 	boolean_t	cb_print_status;
 	boolean_t	cb_print_slow_ios;
 	boolean_t	cb_print_vdev_init;
@@ -2357,6 +2358,35 @@ health_str_to_color(const char *health)
 	return (NULL);
 }

+/*
+ * Called for each leaf vdev.  Returns 0 if the vdev is healthy.
+ * A vdev is unhealthy if any of the following are true:
+ * 1) there are read, write, or checksum errors,
+ * 2) its state is not ONLINE, or
+ * 3) slow IO reporting was requested (-s) and there are slow IOs.
+ */
+static int
+vdev_health_check_cb(void *hdl_data, nvlist_t *nv, void *data)
+{
+	status_cbdata_t *cb = data;
+	vdev_stat_t *vs;
+	uint_t vsc;
+	(void) hdl_data;
+
+	if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
+	    (uint64_t **)&vs, &vsc) != 0)
+		return (1);
+
+	if (vs->vs_checksum_errors || vs->vs_read_errors ||
+	    vs->vs_write_errors || vs->vs_state != VDEV_STATE_HEALTHY)
+		return (1);
+
+	if (cb->cb_print_slow_ios && vs->vs_slow_ios)
+		return (1);
+
+	return (0);
+}
+
 /*
 * Print out configuration state as requested by status_callback.
 */
@@ -2375,7 +2405,8 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 	const char *state;
 	const char *type;
 	const char *path = NULL;
-	const char *rcolor = NULL, *wcolor = NULL, *ccolor = NULL;
+	const char *rcolor = NULL, *wcolor = NULL, *ccolor = NULL,
+	    *scolor = NULL;

 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
 	    &child, &children) != 0)
@@ -2402,6 +2433,15 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 			state = gettext("AVAIL");
 	}

+	/*
+	 * If '-e' is specified then top-level vdevs and their children
+	 * can be pruned if all of their leaves are healthy.
+	 */
+	if (cb->cb_print_unhealthy && depth > 0 &&
+	    for_each_vdev_in_nvlist(nv, vdev_health_check_cb, cb) == 0) {
+		return;
+	}
+
 	printf_color(health_str_to_color(state),
 	    "\t%*s%-*s  %-8s", depth, "", cb->cb_namewidth - depth,
 	    name, state);
@@ -2416,6 +2456,9 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 		if (vs->vs_checksum_errors)
 			ccolor = ANSI_RED;

+		if (vs->vs_slow_ios)
+			scolor = ANSI_BLUE;
+
 		if (cb->cb_literal) {
 			fputc(' ', stdout);
 			printf_color(rcolor, "%5llu",
@@ -2448,9 +2491,10 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 			}

 			if (cb->cb_literal)
-				printf(" %5llu", (u_longlong_t)vs->vs_slow_ios);
+				printf_color(scolor, " %5llu",
+				    (u_longlong_t)vs->vs_slow_ios);
 			else
-				printf(" %5s", rbuf);
+				printf_color(scolor, " %5s", rbuf);
 		}
 		if (cb->cb_print_power) {
 			if (children == 0)  {
@@ -8999,9 +9043,11 @@ status_callback(zpool_handle_t *zhp, void *data)
 				(void) printf(gettext(
 				    "errors: No known data errors\n"));
 			} else if (!cbp->cb_verbose) {
+				color_start(ANSI_RED);
 				(void) printf(gettext("errors: %llu data "
 				    "errors, use '-v' for a list\n"),
 				    (u_longlong_t)nerr);
+				color_end();
 			} else {
 				print_error_log(zhp);
 			}
@@ -9022,6 +9068,7 @@ status_callback(zpool_handle_t *zhp, void *data)
 *              [pool] [interval [count]]
 *
 *	-c CMD	For each vdev, run command CMD
+ *	-e	Display only unhealthy vdevs
 *	-i	Display vdev initialization status.
 *	-g	Display guid for individual vdev name.
 *	-L	Follow links when resolving vdev path name.
@@ -9053,7 +9100,7 @@ zpool_do_status(int argc, char **argv)
 	};

 	/* check options */
-	while ((c = getopt_long(argc, argv, "c:igLpPsvxDtT:", long_options,
+	while ((c = getopt_long(argc, argv, "c:eigLpPsvxDtT:", long_options,
 	    NULL)) != -1) {
 		switch (c) {
 		case 'c':
@@ -9080,6 +9127,9 @@ zpool_do_status(int argc, char **argv)
 			}
 			cmd = optarg;
 			break;
+		case 'e':
+			cb.cb_print_unhealthy = B_TRUE;
+			break;
 		case 'i':
 			cb.cb_print_vdev_init = B_TRUE;
 			break;