agora inbox for pgsql-hackers@postgresql.org  

help / color / mirror / Atom feed

From: Kyotaro Horiguchi <horiguchi.kyotaro@lab.ntt.co.jp>
Subject: [PATCH 3/3] Fix failure of standby startup caused by tablespace removal
Date: Mon, 22 Apr 2019 20:59:15 +0900

When standby restarts after a crash after drop of a tablespace,
there's a possibility that recovery fails trying an object-creation in
already removed tablespace directory. Allow recovery to continue by
ignoring the error if not reaching consistency point.
---
 src/backend/access/transam/xlogutils.c | 34 ++++++++++++++++++++++++++++++++++
 src/backend/storage/file/copydir.c     | 12 +++++++-----
 src/include/access/xlogutils.h         |  1 +
 3 files changed, 42 insertions(+), 5 deletions(-)

diff --git a/src/backend/access/transam/xlogutils.c b/src/backend/access/transam/xlogutils.c
index 10a663bae6..75cdb882cd 100644
--- a/src/backend/access/transam/xlogutils.c
+++ b/src/backend/access/transam/xlogutils.c
@@ -522,6 +522,40 @@ XLogReadBufferExtended(RelFileNode rnode, ForkNumber forknum,
 	return buffer;
 }
 
+/*
+ * XLogMakePGDirectory
+ *
+ * There is a possibility that WAL replay causes an error by creation of a
+ * directory under a directory removed before the previous crash. Issuing
+ * ERROR prevents the caller from continuing recovery.
+ *
+ * To prevent that case, this function issues WARNING instead of ERROR on
+ * error if consistency is not reached yet.
+ */
+int
+XLogMakePGDirectory(const char *directoryName)
+{
+	int ret;
+
+	ret = MakePGDirectory(directoryName);
+
+	if (ret != 0)
+	{
+		int elevel = ERROR;
+
+		/* Don't issue ERROR for this failure before reaching consistency. */
+		if (InRecovery && !reachedConsistency)
+			elevel = WARNING;
+
+		ereport(elevel,
+				(errcode_for_file_access(),
+				 errmsg("could not create directory \"%s\": %m", directoryName)));
+		return ret;
+	}
+
+	return 0;
+}
+
 /*
  * Struct actually returned by XLogFakeRelcacheEntry, though the declared
  * return type is Relation.
diff --git a/src/backend/storage/file/copydir.c b/src/backend/storage/file/copydir.c
index 30f6200a86..0216270dd3 100644
--- a/src/backend/storage/file/copydir.c
+++ b/src/backend/storage/file/copydir.c
@@ -22,11 +22,11 @@
 #include <unistd.h>
 #include <sys/stat.h>
 
+#include "access/xlogutils.h"
 #include "storage/copydir.h"
 #include "storage/fd.h"
 #include "miscadmin.h"
 #include "pgstat.h"
-
 /*
  * copydir: copy a directory
  *
@@ -41,10 +41,12 @@ copydir(char *fromdir, char *todir, bool recurse)
 	char		fromfile[MAXPGPATH * 2];
 	char		tofile[MAXPGPATH * 2];
 
-	if (MakePGDirectory(todir) != 0)
-		ereport(ERROR,
-				(errcode_for_file_access(),
-				 errmsg("could not create directory \"%s\": %m", todir)));
+	/*
+	 * We might have to skip copydir to continue recovery. See the function
+	 * for details.
+	 */
+	if (XLogMakePGDirectory(todir) != 0)
+		return;
 
 	xldir = AllocateDir(fromdir);
 
diff --git a/src/include/access/xlogutils.h b/src/include/access/xlogutils.h
index 0ab5ba62f5..46a7596315 100644
--- a/src/include/access/xlogutils.h
+++ b/src/include/access/xlogutils.h
@@ -43,6 +43,7 @@ extern XLogRedoAction XLogReadBufferForRedoExtended(XLogReaderState *record,
 
 extern Buffer XLogReadBufferExtended(RelFileNode rnode, ForkNumber forknum,
 					   BlockNumber blkno, ReadBufferMode mode);
+extern int XLogMakePGDirectory(const char *directoryName);
 
 extern Relation CreateFakeRelcacheEntry(RelFileNode rnode);
 extern void FreeFakeRelcacheEntry(Relation fakerel);
-- 
2.16.3


----Next_Part(Mon_Apr_22_21_19_33_2019_510)----

view thread (6+ messages)  latest in thread

Message-ID: <no-message-id-1883275@localhost>
Permalink:  ../../no-message-id-1883275@localhost/
Also on:    postgresql.org/message-id/no-message-id-1883275@localhost

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-hackers@postgresql.org
  Cc: horiguchi.kyotaro@lab.ntt.co.jp
  Subject: Re: [PATCH 3/3] Fix failure of standby startup caused by tablespace removal
  In-Reply-To: <no-message-id-1883275@localhost>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox