mds: allow boot on read-only

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit c7ce967b778a0b86b335f6801301e484aaf6ebc3)

Conflicts:
	src/mds/MDSRank.cc
- no ceph_assert in mimic

diff --git a/qa/tasks/cephfs/fuse_mount.py b/qa/tasks/cephfs/fuse_mount.py
index a7c394718abd0c4ed7cef6079dc8815d9503faf8..66b97d34406044c7e4b99cd65ad3790e9e5d331d 100644 (file)
--- a/qa/tasks/cephfs/fuse_mount.py
+++ b/qa/tasks/cephfs/fuse_mount.py
@@ -226,8 +226,15 @@ class FuseMount(CephFSMount):
 
         # Now that we're mounted, set permissions so that the rest of the test will have
         # unrestricted access to the filesystem mount.
-        self.client_remote.run(
-            args=['sudo', 'chmod', '1777', self.mountpoint], timeout=(15*60))
+        try:
+            stderr = StringIO()
+            self.client_remote.run(args=['sudo', 'chmod', '1777', self.mountpoint], timeout=(15*60), stderr=stderr)
+        except run.CommandFailedError:
+            stderr = stderr.getvalue()
+            if "Read-only file system".lower() in stderr.lower():
+                pass
+            else:
+                raise
 
     def _mountpoint_exists(self):
         return self.client_remote.run(args=["ls", "-d", self.mountpoint], check_status=False, timeout=(15*60)).exitstatus == 0

diff --git a/qa/tasks/cephfs/test_damage.py b/qa/tasks/cephfs/test_damage.py
index 3d14244ad9f718671283fc70fd1d8c461650ae33..f534a12759d96f78de1ff86aa9048c414c4b58d6 100644 (file)
--- a/qa/tasks/cephfs/test_damage.py
+++ b/qa/tasks/cephfs/test_damage.py
@@ -12,6 +12,7 @@ DAMAGED_ON_START = "damaged_on_start"
 DAMAGED_ON_LS = "damaged_on_ls"
 CRASHED = "server crashed"
 NO_DAMAGE = "no damage"
+READONLY = "readonly"
 FAILED_CLIENT = "client failed"
 FAILED_SERVER = "server failed"
 
@@ -161,14 +162,22 @@ class TestDamage(CephFSTestCase):
             ))
 
         # Blatant corruptions
-        mutations.extend([
-            MetadataMutation(
-                o,
-                "Corrupt {0}".format(o),
-                lambda o=o: self.fs.rados(["put", o, "-"], stdin_data=junk),
-                DAMAGED_ON_START
-            ) for o in data_objects
-        ])
+        for obj_id in data_objects:
+            if obj_id == "500.00000000":
+                # purge queue corruption results in read-only FS
+                mutations.append(MetadataMutation(
+                    obj_id,
+                    "Corrupt {0}".format(obj_id),
+                    lambda o=obj_id: self.fs.rados(["put", o, "-"], stdin_data=junk),
+                    READONLY
+                ))
+            else:
+                mutations.append(MetadataMutation(
+                    obj_id,
+                    "Corrupt {0}".format(obj_id),
+                    lambda o=obj_id: self.fs.rados(["put", o, "-"], stdin_data=junk),
+                    DAMAGED_ON_START
+                ))
 
         # Truncations
         for o in data_objects:
@@ -316,7 +325,17 @@ class TestDamage(CephFSTestCase):
                     else:
                         log.error("Result: Failed to go damaged on mutation '{0}'".format(mutation.desc))
                         results[mutation] = FAILED_SERVER
-
+            elif mutation.expectation == READONLY:
+                proc = self.mount_a.run_shell(["mkdir", "foo"], wait=False)
+                try:
+                    proc.wait()
+                except CommandFailedError:
+                    stderr = proc.stderr.getvalue()
+                    log.info(stderr)
+                    if "Read-only file system".lower() in stderr.lower():
+                        pass
+                    else:
+                        raise
             else:
                 try:
                     wait([proc], 20)

diff --git a/src/mds/MDSRank.cc b/src/mds/MDSRank.cc
index 325dc6cab9bafbfa30dd45c7a89c000a13f3389f..90dfe64a4774892de761bc42f640387e9305e7ac 100644 (file)
--- a/src/mds/MDSRank.cc
+++ b/src/mds/MDSRank.cc
@@ -1501,6 +1501,8 @@ void MDSRank::boot_start(BootStep step, int r)
         << cpp_strerror(r);
       damaged();
       assert(r == 0);  // Unreachable, damaged() calls respawn()
+    } else if (r == -EROFS) {
+      dout(0) << "boot error forcing transition to read-only; MDS will try to continue" << dendl;
     } else {
       // Completely unexpected error, give up and die
       dout(0) << "boot_start encountered an error, failing" << dendl;