From: Casey Bodley <cbodley@redhat.com>
Date: Thu, 13 Feb 2020 19:26:20 +0000 (-0500)
Subject: rgw: only trim logs for uninitialized bucket sync status
X-Git-Tag: v15.1.1~324^2~1
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=82af57afa22dd9af5b5c7415f8c290c227c5dc52;p=ceph.git

rgw: only trim logs for uninitialized bucket sync status

the pubsub zone cannot trim logs when it's in the Stopped state, because
it needs to resume from a specific position - so take that position into
account for trim

a similar issue effects full sync - it also records a marker position
where it should resume incremental sync after full sync finishes, and we
can't trim entries past that either

Signed-off-by: Casey Bodley <cbodley@redhat.com>
---

diff --git a/src/rgw/rgw_trim_bilog.cc b/src/rgw/rgw_trim_bilog.cc
index 8ccc6d547438..9e2ac89d218e 100644
--- a/src/rgw/rgw_trim_bilog.cc
+++ b/src/rgw/rgw_trim_bilog.cc
@@ -364,8 +364,8 @@ int take_min_status(CephContext *cct, Iter first, Iter last,
     auto m = status->begin();
     for (auto& shard : *peer) {
       auto& marker = *m++;
-      // only consider incremental sync markers
-      if (shard.state != rgw_bucket_shard_sync_info::StateIncrementalSync) {
+      // if no sync has started, we can safely trim everything
+      if (shard.state == rgw_bucket_shard_sync_info::StateInit) {
         continue;
       }
       // always take the first marker, or any later marker that's smaller
diff --git a/src/test/rgw/rgw_multi/tests.py b/src/test/rgw/rgw_multi/tests.py
index 4ff642e68fe1..2ddf88c6170e 100644
--- a/src/test/rgw/rgw_multi/tests.py
+++ b/src/test/rgw/rgw_multi/tests.py
@@ -270,16 +270,12 @@ def bucket_sync_status(target_zone, source_zone, bucket_name):
         assert(retcode == 2) # ENOENT
 
     bucket_sync_status_json = bucket_sync_status_json.decode('utf-8')
-    log.debug('current bucket sync markers=%s', bucket_sync_status_json)
     sync_status = json.loads(bucket_sync_status_json)
 
     markers={}
     for entry in sync_status:
         val = entry['val']
-        if val['status'] == 'incremental-sync':
-            pos = val['inc_marker']['position'].split('#')[-1] # get rid of shard id; e.g., 6#00000000002.132.3 -> 00000000002.132.3
-        else:
-            pos = ''
+        pos = val['inc_marker']['position'].split('#')[-1] # get rid of shard id; e.g., 6#00000000002.132.3 -> 00000000002.132.3
         markers[entry['key']] = pos
 
     return markers