From: Ronen Friedman <rfriedma@redhat.com>
Date: Wed, 6 Aug 2025 05:38:07 +0000 (-0500)
Subject: osd/scrub: do not limit operator-initiated repairs
X-Git-Url: http://git.apps.os.sepia.ceph.com/?a=commitdiff_plain;h=5afc446fdafe3f2e168f9846f324d1d6a71f0f77;p=ceph.git

osd/scrub: do not limit operator-initiated repairs

'auto-repair' scrubs are limited to a maximum of
'scrub_auto_repair_num_errors' damaged objects.
However, operator-initiated repairs should not be limited
by that number. Alas, a bug in a previous commit
(97de817ad1c253ee1c7c9c9302981ad2435301b9) modified the
code in such a way that it applied the
'scrub_auto_repair_num_errors' limit to all repairs,
including operator-initiated ones. This commit fixes that.

Fixes: https://tracker.ceph.com/issues/72420
Signed-off-by: Ronen Friedman <rfriedma@redhat.com>
---

diff --git a/src/osd/scrubber/pg_scrubber.cc b/src/osd/scrubber/pg_scrubber.cc
index 515b53fed18be..eb9c0c697e373 100644
--- a/src/osd/scrubber/pg_scrubber.cc
+++ b/src/osd/scrubber/pg_scrubber.cc
@@ -1828,10 +1828,17 @@ void PgScrubber::scrub_finish()
   // if the repair request comes from auto-repair and there is a large
   // number of objects known to be damaged, we cancel the auto-repair
   if (m_is_repair && m_flags.auto_repair &&
+      ScrubJob::is_repairs_count_limited(m_active_target->urgency()) &&
       m_be->authoritative_peers_count() >
-	static_cast<int>(m_pg->cct->_conf->osd_scrub_auto_repair_num_errors)) {
+	  static_cast<int>(
+	      m_pg->cct->_conf->osd_scrub_auto_repair_num_errors)) {
 
-    dout(10) << __func__ << " undoing the repair" << dendl;
+    dout(5) << fmt::format(
+		   "{}: undoing the repair. Damaged objects count ({}) is "
+		   "above configured limit ({})",
+		   __func__, m_be->authoritative_peers_count(),
+		   m_pg->cct->_conf->osd_scrub_auto_repair_num_errors)
+	    << dendl;
     state_clear(PG_STATE_REPAIR);  // not expected to be set, anyway
     m_is_repair = false;
     update_op_mode_text();
diff --git a/src/osd/scrubber/scrub_job.cc b/src/osd/scrubber/scrub_job.cc
index da9cbb01820f6..81f45ffca804d 100644
--- a/src/osd/scrubber/scrub_job.cc
+++ b/src/osd/scrubber/scrub_job.cc
@@ -426,3 +426,8 @@ bool ScrubJob::is_autorepair_allowed(urgency_t urgency)
 	 urgency == urgency_t::operator_requested ||
 	 urgency == urgency_t::repairing || urgency == urgency_t::must_repair;
 }
+
+bool ScrubJob::is_repairs_count_limited(urgency_t urgency)
+{
+  return urgency < urgency_t::operator_requested;
+}
diff --git a/src/osd/scrubber/scrub_job.h b/src/osd/scrubber/scrub_job.h
index 5673c20bb903a..3aef5a874a48a 100644
--- a/src/osd/scrubber/scrub_job.h
+++ b/src/osd/scrubber/scrub_job.h
@@ -375,6 +375,13 @@ class ScrubJob {
   static bool is_repair_implied(urgency_t urgency);
 
   static bool is_autorepair_allowed(urgency_t urgency);
+
+  /**
+   * should we cancel the repair if the number of damaged objects
+   * exceeds the configured limit ('osd_scrub_auto_repair_num_errors')?
+   * This does not apply to any repair that was operator-initiated.
+   */
+  static bool is_repairs_count_limited(urgency_t urgency);
 };
 }  // namespace Scrub