osd: log 'slow op' debug messages for individual slow ops

Otherwise it is very hard to identify which OSD ops are slow when we've
seen a SLOW_OPS health warning in a qa run.

Notably, without this, bugs like http://tracker.ceph.com/issues/23769
are very challenging to track down.

Signed-off-by: Sage Weil <sage@redhat.com>

diff --git a/src/osd/OSD.cc b/src/osd/OSD.cc
index 1285c22dce15b48e09fd323d48129ef595330f37..b801b63c795a1b2a24d6f74a0970f58ce9b5f3e9 100644 (file)
--- a/src/osd/OSD.cc
+++ b/src/osd/OSD.cc
@@ -7011,6 +7011,9 @@ vector<DaemonHealthMetric> OSD::get_health_metrics()
     int slow = 0;
     auto count_slow_ops = [&](TrackedOp& op) {
       if (op.get_initiated() < too_old) {
+       lgeneric_subdout(cct,osd,20) << "slow op " << op.get_desc()
+                                    << " initiated "
+                                    << op.get_initiated() << dendl;
        slow++;
        return true;
       } else {