import logging
import os
-from textwrap import dedent
+import time
import traceback
+
+from io import BytesIO
from collections import namedtuple, defaultdict
+from textwrap import dedent
from teuthology.orchestra.run import CommandFailedError
from tasks.cephfs.cephfs_test_case import CephFSTestCase, for_teuthology
# Exactly stripe_count objects will exist
self.os * self.sc,
# Fewer than stripe_count objects will exist
- self.os * self.sc / 2,
- self.os * (self.sc - 1) + self.os / 2,
- self.os * (self.sc - 1) + self.os / 2 - 1,
- self.os * (self.sc + 1) + self.os / 2,
- self.os * (self.sc + 1) + self.os / 2 + 1,
+ self.os * self.sc // 2,
+ self.os * (self.sc - 1) + self.os // 2,
+ self.os * (self.sc - 1) + self.os // 2 - 1,
+ self.os * (self.sc + 1) + self.os // 2,
+ self.os * (self.sc + 1) + self.os // 2 + 1,
# More than stripe_count objects will exist
- self.os * self.sc + self.os * self.sc / 2
+ self.os * self.sc + self.os * self.sc // 2
]
def write(self):
log.info(str(self.mds_cluster.status()))
# Mount a client
- self.mount_a.mount()
- self.mount_a.wait_until_mounted()
+ self.mount_a.mount_wait()
# See that the files are present and correct
errors = workload.validate()
That when injecting a dentry into a fragmented directory, we put it in the right fragment.
"""
- self.fs.set_allow_dirfrags(True)
-
file_count = 100
file_names = ["%s" % n for n in range(0, file_count)]
+ # Make sure and disable dirfrag auto merging and splitting
+ self.fs.set_ceph_conf('mds', 'mds bal merge size', 0)
+ self.fs.set_ceph_conf('mds', 'mds bal split size', 100 * file_count)
+
# Create a directory of `file_count` files, each named after its
# decimal number and containing the string of its decimal number
self.mount_a.run_python(dedent("""
# Start filesystem back up, observe that the file appears to be gone in an `ls`
self.fs.mds_restart()
self.fs.wait_for_daemons()
- self.mount_a.mount()
- self.mount_a.wait_until_mounted()
+ self.mount_a.mount_wait()
files = self.mount_a.run_shell(["ls", "subdir/"]).stdout.getvalue().strip().split("\n")
self.assertListEqual(sorted(files), sorted(list(set(file_names) - set([victim_dentry]))))
# by checking the omap now has the dentry's key again
self.fs.data_scan(["scan_extents", self.fs.get_data_pool_name()])
self.fs.data_scan(["scan_inodes", self.fs.get_data_pool_name()])
+ self.fs.data_scan(["scan_links"])
self.assertIn(victim_key, self._dirfrag_keys(frag_obj_id))
# Start the filesystem and check that the dentry we deleted is now once again visible
# and points to the correct file data.
self.fs.mds_restart()
self.fs.wait_for_daemons()
- self.mount_a.mount()
- self.mount_a.wait_until_mounted()
+ self.mount_a.mount_wait()
out = self.mount_a.run_shell(["cat", "subdir/{0}".format(victim_dentry)]).stdout.getvalue().strip()
self.assertEqual(out, victim_dentry)
keys = self._dirfrag_keys(frag_obj_id)
self.assertListEqual(sorted(keys), sorted(["%s_head" % f for f in file_names]))
+ # run scrub to update and make sure rstat.rbytes info in subdir inode and dirfrag
+ # are matched
+ out_json = self.fs.rank_tell(["scrub", "start", "/subdir", "repair", "recursive"])
+ self.assertNotEqual(out_json, None)
+
+ # Remove the whole 'sudbdir' directory
+ self.mount_a.run_shell(["rm", "-rf", "subdir/"])
+
@for_teuthology
def test_parallel_execution(self):
self._rebuild_metadata(ManyFilesWorkload(self.fs, self.mount_a, 25), workers=7)
pgs_to_files[pgid].append(file_path)
log.info("{0}: {1}".format(file_path, pgid))
- pg_count = self.fs.get_pgs_per_fs_pool()
+ pg_count = self.fs.pgs_per_fs_pool
for pg_n in range(0, pg_count):
pg_str = "{0}.{1}".format(self.fs.get_data_pool_id(), pg_n)
out = self.fs.data_scan(["pg_files", "mydir", pg_str])
log.info("{0}: {1}".format(pg_str, lines))
self.assertSetEqual(set(lines), set(pgs_to_files[pg_str]))
- def test_scan_links(self):
+ def test_rebuild_linkage(self):
"""
The scan_links command fixes linkage errors
"""
# introduce duplicated primary link
file1_key = "file1_head"
self.assertIn(file1_key, dirfrag1_keys)
- file1_omap_data = self.fs.rados(["getomapval", dirfrag1_oid, file1_key, '-'])
+ file1_omap_data = self.fs.rados(["getomapval", dirfrag1_oid, file1_key, '-'],
+ stdout_data=BytesIO())
self.fs.rados(["setomapval", dirfrag2_oid, file1_key], stdin_data=file1_omap_data)
self.assertIn(file1_key, self._dirfrag_keys(dirfrag2_oid))
self.fs.mds_restart()
self.fs.wait_for_daemons()
- self.mount_a.mount()
- self.mount_a.wait_until_mounted()
+ self.mount_a.mount_wait()
# link count was adjusted?
file1_nlink = self.mount_a.path_to_nlink("testdir1/file1")
self.assertEqual(file1_nlink, 2)
+
+ def test_rebuild_inotable(self):
+ """
+ The scan_links command repair inotables
+ """
+ self.fs.set_max_mds(2)
+ self.fs.wait_for_daemons()
+
+ active_mds_names = self.fs.get_active_names()
+ mds0_id = active_mds_names[0]
+ mds1_id = active_mds_names[1]
+
+ self.mount_a.run_shell(["mkdir", "dir1"])
+ dir_ino = self.mount_a.path_to_ino("dir1")
+ self.mount_a.setfattr("dir1", "ceph.dir.pin", "1")
+ # wait for subtree migration
+
+ file_ino = 0;
+ while True:
+ time.sleep(1)
+ # allocate an inode from mds.1
+ self.mount_a.run_shell(["touch", "dir1/file1"])
+ file_ino = self.mount_a.path_to_ino("dir1/file1")
+ if file_ino >= (2 << 40):
+ break
+ self.mount_a.run_shell(["rm", "-f", "dir1/file1"])
+
+ self.mount_a.umount_wait()
+
+ self.fs.mds_asok(["flush", "journal"], mds0_id)
+ self.fs.mds_asok(["flush", "journal"], mds1_id)
+ self.mds_cluster.mds_stop()
+
+ self.fs.rados(["rm", "mds0_inotable"])
+ self.fs.rados(["rm", "mds1_inotable"])
+
+ self.fs.data_scan(["scan_links", "--filesystem", self.fs.name])
+
+ mds0_inotable = json.loads(self.fs.table_tool([self.fs.name + ":0", "show", "inode"]))
+ self.assertGreaterEqual(
+ mds0_inotable['0']['data']['inotable']['free'][0]['start'], dir_ino)
+
+ mds1_inotable = json.loads(self.fs.table_tool([self.fs.name + ":1", "show", "inode"]))
+ self.assertGreaterEqual(
+ mds1_inotable['1']['data']['inotable']['free'][0]['start'], file_ino)
+
+ def test_rebuild_snaptable(self):
+ """
+ The scan_links command repair snaptable
+ """
+ self.fs.set_allow_new_snaps(True)
+
+ self.mount_a.run_shell(["mkdir", "dir1"])
+ self.mount_a.run_shell(["mkdir", "dir1/.snap/s1"])
+ self.mount_a.run_shell(["mkdir", "dir1/.snap/s2"])
+ self.mount_a.run_shell(["rmdir", "dir1/.snap/s2"])
+
+ self.mount_a.umount_wait()
+
+ mds0_id = self.fs.get_active_names()[0]
+ self.fs.mds_asok(["flush", "journal"], mds0_id)
+
+ # wait for mds to update removed snaps
+ time.sleep(10)
+
+ old_snaptable = json.loads(self.fs.table_tool([self.fs.name + ":0", "show", "snap"]))
+ # stamps may have minor difference
+ for item in old_snaptable['snapserver']['snaps']:
+ del item['stamp']
+
+ self.fs.rados(["rm", "mds_snaptable"])
+ self.fs.data_scan(["scan_links", "--filesystem", self.fs.name])
+
+ new_snaptable = json.loads(self.fs.table_tool([self.fs.name + ":0", "show", "snap"]))
+ for item in new_snaptable['snapserver']['snaps']:
+ del item['stamp']
+ self.assertGreaterEqual(
+ new_snaptable['snapserver']['last_snap'], old_snaptable['snapserver']['last_snap'])
+ self.assertEqual(
+ new_snaptable['snapserver']['snaps'], old_snaptable['snapserver']['snaps'])