Merge pull request ceph#58858 from ronen-fr/wip-rf-entry

osd/scrub: a scrub queue of level-specific entries

Reviewed-by: Samuel Just <[email protected]>
Reviewed-by: Nitzan Mordechai <[email protected]>

Author: ronen-fr

qa/standalone/scrub/osd-recovery-scrub.sh

@@ -99,11 +99,11 @@ function TEST_recovery_scrub_1() {
     kill_daemons $dir #|| return 1
 
     declare -a err_strings
-    err_strings[0]="recovery in progress. Only high priority scrubs allowed."
+    err_strings[0]="recovery in progress.*scrubs"
 
     for osd in $(seq 0 $(expr $OSDS - 1))
     do
-        grep "recovery in progress. Only high priority scrubs allowed." $dir/osd.${osd}.log
+        grep "recovery in progress.*scrubs" $dir/osd.${osd}.log
     done
     for err_string in "${err_strings[@]}"
     do
@@ -242,7 +242,7 @@ function TEST_recovery_scrub_2() {
     setup $dir || return 1
     run_mon $dir a --osd_pool_default_size=1 --mon_allow_pool_size_one=true || return 1
     run_mgr $dir x || return 1
-    local ceph_osd_args="--osd-scrub-interval-randomize-ratio=0 "
+    local ceph_osd_args="--osd-scrub-interval-randomize-ratio=0.1 "
     ceph_osd_args+="--osd_scrub_backoff_ratio=0 "
     ceph_osd_args+="--osd_stats_update_period_not_scrubbing=3 "
     ceph_osd_args+="--osd_stats_update_period_scrubbing=2"
@@ -269,7 +269,7 @@ function TEST_recovery_scrub_2() {
     ceph pg dump pgs
 
     # note that the following will be needed if the mclock scheduler is specified
-    #ceph tell osd.* config get osd_mclock_override_recovery_settings
+    ceph tell osd.* config get osd_mclock_override_recovery_settings
 
     # the '_max_active' is expected to be 0
     ceph tell osd.1 config get osd_recovery_max_active
@@ -327,11 +327,11 @@ function TEST_recovery_scrub_2() {
     kill_daemons $dir #|| return 1
 
     declare -a err_strings
-    err_strings[0]="not scheduling scrubs due to active recovery"
-
+    ## we do not expect a refusal to scrub
+    err_strings[0]="recovery in progress.*scrubs"
     for osd in $(seq 0 $(expr $OSDS - 1))
     do
-        grep "not scheduling scrubs" $dir/osd.${osd}.log
+        grep "recovery in progress.*scrubs" $dir/osd.${osd}.log
     done
     for err_string in "${err_strings[@]}"
     do

qa/standalone/scrub/osd-scrub-repair.sh

@@ -5754,11 +5754,13 @@ function TEST_corrupt_scrub_erasure_overwrites() {
 
 #
 # Test to make sure that a periodic scrub won't cause deep-scrub info to be lost
+# Update 2024: this functionality was removed from the code. The test will be skipped.
 #
 function TEST_periodic_scrub_replicated() {
     local dir=$1
     local poolname=psr_pool
     local objname=POBJ
+    return 0
 
     run_mon $dir a --osd_pool_default_size=2 || return 1
     run_mgr $dir x || return 1
@@ -5795,12 +5797,13 @@ function TEST_periodic_scrub_replicated() {
 
     flush_pg_stats
     local last_scrub=$(get_last_scrub_stamp $pg)
-    # Fake a schedule scrub
+    # Fake a scheduled deep scrub
     ceph tell $pg schedule-scrub || return 1
     # Wait for schedule regular scrub
     wait_for_scrub $pg "$last_scrub"
 
     # It needed to be upgraded
+    # update 2024: the "upgrade" functionality has been removed
     grep -q "Deep scrub errors, upgrading scrub to deep-scrub" $dir/osd.${primary}.log || return 1
 
     # Bad object still known

qa/standalone/scrub/osd-scrub-test.sh

@@ -205,7 +205,9 @@ function TEST_interval_changes() {
     perf_counters $dir $OSDS
 }
 
-function TEST_scrub_extended_sleep() {
+# RRR 6aug24: this test cannot work as expected, following the changes in the
+#   scrub type to overrides matrix. Disabled for now.
+function NO_scrub_extended_sleep() {
     local dir=$1
     local poolname=test
     local OSDS=3