Extend import_progress kstat with a notes field

Detail the import progress of log spacemaps as they can take a very
long time.  Also grab the spa_note() messages to, as they provide
insight into what is happening

Sponsored-By: OpenDrives Inc.
Sponsored-By: Klara Inc.
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Don Brady <don.brady@klarasystems.com>
Co-authored-by: Allan Jude <allan@klarasystems.com>
Closes #15539
This commit is contained in:
Don Brady
2023-12-05 15:27:56 -07:00
committed by GitHub
parent 727497ccdf
commit 687e4d7f9c
6 changed files with 256 additions and 10 deletions
+39 -2
View File
@@ -3109,6 +3109,7 @@ spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type)
spa->spa_load_state = state;
(void) spa_import_progress_set_state(spa_guid(spa),
spa_load_state(spa));
spa_import_progress_set_notes(spa, "spa_load()");
gethrestime(&spa->spa_loaded_ts);
error = spa_load_impl(spa, type, &ereport);
@@ -3337,7 +3338,7 @@ spa_activity_check(spa_t *spa, uberblock_t *ub, nvlist_t *config)
uint64_t mmp_config = ub->ub_mmp_config;
uint16_t mmp_seq = MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0;
uint64_t import_delay;
hrtime_t import_expire;
hrtime_t import_expire, now;
nvlist_t *mmp_label = NULL;
vdev_t *rvd = spa->spa_root_vdev;
kcondvar_t cv;
@@ -3375,7 +3376,17 @@ spa_activity_check(spa_t *spa, uberblock_t *ub, nvlist_t *config)
import_expire = gethrtime() + import_delay;
while (gethrtime() < import_expire) {
spa_import_progress_set_notes(spa, "Checking MMP activity, waiting "
"%llu ms", (u_longlong_t)NSEC2MSEC(import_delay));
int interations = 0;
while ((now = gethrtime()) < import_expire) {
if (interations++ % 30 == 0) {
spa_import_progress_set_notes(spa, "Checking MMP "
"activity, %llu ms remaining",
(u_longlong_t)NSEC2MSEC(import_expire - now));
}
(void) spa_import_progress_set_mmp_check(spa_guid(spa),
NSEC2SEC(import_expire - gethrtime()));
@@ -4995,6 +5006,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
/*
* Retrieve the checkpoint txg if the pool has a checkpoint.
*/
spa_import_progress_set_notes(spa, "Loading checkpoint txg");
error = spa_ld_read_checkpoint_txg(spa);
if (error != 0)
return (error);
@@ -5007,6 +5019,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* initiated. Otherwise we could be reading from indirect vdevs before
* we have loaded their mappings.
*/
spa_import_progress_set_notes(spa, "Loading indirect vdev metadata");
error = spa_ld_open_indirect_vdev_metadata(spa);
if (error != 0)
return (error);
@@ -5015,6 +5028,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* Retrieve the full list of active features from the MOS and check if
* they are all supported.
*/
spa_import_progress_set_notes(spa, "Checking feature flags");
error = spa_ld_check_features(spa, &missing_feat_write);
if (error != 0)
return (error);
@@ -5023,6 +5037,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* Load several special directories from the MOS needed by the dsl_pool
* layer.
*/
spa_import_progress_set_notes(spa, "Loading special MOS directories");
error = spa_ld_load_special_directories(spa);
if (error != 0)
return (error);
@@ -5030,6 +5045,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
/*
* Retrieve pool properties from the MOS.
*/
spa_import_progress_set_notes(spa, "Loading properties");
error = spa_ld_get_props(spa);
if (error != 0)
return (error);
@@ -5038,6 +5054,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* Retrieve the list of auxiliary devices - cache devices and spares -
* and open them.
*/
spa_import_progress_set_notes(spa, "Loading AUX vdevs");
error = spa_ld_open_aux_vdevs(spa, type);
if (error != 0)
return (error);
@@ -5046,14 +5063,17 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* Load the metadata for all vdevs. Also check if unopenable devices
* should be autoreplaced.
*/
spa_import_progress_set_notes(spa, "Loading vdev metadata");
error = spa_ld_load_vdev_metadata(spa);
if (error != 0)
return (error);
spa_import_progress_set_notes(spa, "Loading dedup tables");
error = spa_ld_load_dedup_tables(spa);
if (error != 0)
return (error);
spa_import_progress_set_notes(spa, "Loading BRT");
error = spa_ld_load_brt(spa);
if (error != 0)
return (error);
@@ -5062,6 +5082,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* Verify the logs now to make sure we don't have any unexpected errors
* when we claim log blocks later.
*/
spa_import_progress_set_notes(spa, "Verifying Log Devices");
error = spa_ld_verify_logs(spa, type, ereport);
if (error != 0)
return (error);
@@ -5083,6 +5104,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* state. When performing an extreme rewind, we verify the whole pool,
* which can take a very long time.
*/
spa_import_progress_set_notes(spa, "Verifying pool data");
error = spa_ld_verify_pool_data(spa);
if (error != 0)
return (error);
@@ -5092,6 +5114,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* we write anything to the pool because we'd need to update the space
* accounting using the deflated sizes.
*/
spa_import_progress_set_notes(spa, "Calculating deflated space");
spa_update_dspace(spa);
/*
@@ -5099,6 +5122,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* pool. If we are importing the pool in read-write mode, a few
* additional steps must be performed to finish the import.
*/
spa_import_progress_set_notes(spa, "Starting import");
if (spa_writeable(spa) && (spa->spa_load_state == SPA_LOAD_RECOVER ||
spa->spa_load_max_txg == UINT64_MAX)) {
uint64_t config_cache_txg = spa->spa_config_txg;
@@ -5122,6 +5146,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
(u_longlong_t)spa->spa_uberblock.ub_checkpoint_txg);
}
spa_import_progress_set_notes(spa, "Claiming ZIL blocks");
/*
* Traverse the ZIL and claim all blocks.
*/
@@ -5141,6 +5166,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* will have been set for us by ZIL traversal operations
* performed above.
*/
spa_import_progress_set_notes(spa, "Syncing ZIL claims");
txg_wait_synced(spa->spa_dsl_pool, spa->spa_claim_max_txg);
/*
@@ -5148,6 +5174,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* next sync, we would update the config stored in vdev labels
* and the cachefile (by default /etc/zfs/zpool.cache).
*/
spa_import_progress_set_notes(spa, "Updating configs");
spa_ld_check_for_config_update(spa, config_cache_txg,
update_config_cache);
@@ -5156,6 +5183,7 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* Then check all DTLs to see if anything needs resilvering.
* The resilver will be deferred if a rebuild was started.
*/
spa_import_progress_set_notes(spa, "Starting resilvers");
if (vdev_rebuild_active(spa->spa_root_vdev)) {
vdev_rebuild_restart(spa);
} else if (!dsl_scan_resilvering(spa->spa_dsl_pool) &&
@@ -5169,6 +5197,8 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
*/
spa_history_log_version(spa, "open", NULL);
spa_import_progress_set_notes(spa,
"Restarting device removals");
spa_restart_removal(spa);
spa_spawn_aux_threads(spa);
@@ -5181,19 +5211,26 @@ spa_load_impl(spa_t *spa, spa_import_type_t type, const char **ereport)
* auxiliary threads above (from which the livelist
* deletion zthr is part of).
*/
spa_import_progress_set_notes(spa,
"Cleaning up inconsistent objsets");
(void) dmu_objset_find(spa_name(spa),
dsl_destroy_inconsistent, NULL, DS_FIND_CHILDREN);
/*
* Clean up any stale temporary dataset userrefs.
*/
spa_import_progress_set_notes(spa,
"Cleaning up temporary userrefs");
dsl_pool_clean_tmp_userrefs(spa->spa_dsl_pool);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
spa_import_progress_set_notes(spa, "Restarting initialize");
vdev_initialize_restart(spa->spa_root_vdev);
spa_import_progress_set_notes(spa, "Restarting TRIM");
vdev_trim_restart(spa->spa_root_vdev);
vdev_autotrim_restart(spa);
spa_config_exit(spa, SCL_CONFIG, FTAG);
spa_import_progress_set_notes(spa, "Finished importing");
}
spa_import_progress_remove(spa_guid(spa));