osd: EC optimizations fix bug when recovering only partial write objects

PGLog::reset_complete_to is not handling the scenario where all the
missing objects have a partial write that excludes updating the shard being
recovered as their most recent update. In this scenario the oldest need
is newer than newest log entry. Setting last_compelte to the head of the
log confuses code and makes it think that recovery has completed.

The fix is to hold last_complete one entry behind the head of the log
until all missing objects have been recovered.

PGLog::recover_got already does this when an object is recovered and the
remaining objects to recover match this scenario, so this fix just makes
reset_complete_to behave the same way as recover_got.

Signed-off-by: Bill Scales <bill_scales@uk.ibm.com>

diff --git a/src/osd/PGLog.h b/src/osd/PGLog.h
index c829f0534898312b04debb4e8274cbef40c581bd..53ca8e92eb2ab2f4e75d0f54467c7d796c6335f2 100644 (file)
--- a/src/osd/PGLog.h
+++ b/src/osd/PGLog.h
@@ -1010,6 +1010,9 @@ public:
        // partial writes allow a shard which did not participate in a write to
        // have a missing version that is newer that the most recent log entry
        if (ec_optimizations_enabled && (log.complete_to == log.log.end())) {
+         // keep complete_to one entry behind the end of the log to stop
+         // code incorrectly using it to deduce that recovery has completed
+         --log.complete_to;
          break;
        }
         ceph_assert(log.complete_to != log.log.end());