2010-05-28 20:45:14 +00:00
|
|
|
/*
|
|
|
|
* CDDL HEADER START
|
|
|
|
*
|
|
|
|
* The contents of this file are subject to the terms of the
|
|
|
|
* Common Development and Distribution License (the "License").
|
|
|
|
* You may not use this file except in compliance with the License.
|
|
|
|
*
|
|
|
|
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
|
|
|
|
* or http://www.opensolaris.org/os/licensing.
|
|
|
|
* See the License for the specific language governing permissions
|
|
|
|
* and limitations under the License.
|
|
|
|
*
|
|
|
|
* When distributing Covered Code, include this CDDL HEADER in each
|
|
|
|
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
|
|
|
|
* If applicable, add the following below this CDDL HEADER, with the
|
|
|
|
* fields enclosed by brackets "[]" replaced with your own identifying
|
|
|
|
* information: Portions Copyright [yyyy] [name of copyright owner]
|
|
|
|
*
|
|
|
|
* CDDL HEADER END
|
|
|
|
*/
|
|
|
|
/*
|
|
|
|
* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
|
2017-02-09 18:19:12 +00:00
|
|
|
* Copyright (c) 2012, 2016 by Delphix. All rights reserved.
|
2015-04-02 03:44:32 +00:00
|
|
|
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
|
2010-05-28 20:45:14 +00:00
|
|
|
*/
|
|
|
|
|
|
|
|
#include <sys/dsl_dataset.h>
|
|
|
|
#include <sys/dmu.h>
|
|
|
|
#include <sys/refcount.h>
|
|
|
|
#include <sys/zap.h>
|
|
|
|
#include <sys/zfs_context.h>
|
|
|
|
#include <sys/dsl_pool.h>
|
|
|
|
|
2011-11-17 18:14:36 +00:00
|
|
|
/*
|
|
|
|
* Deadlist concurrency:
|
|
|
|
*
|
|
|
|
* Deadlists can only be modified from the syncing thread.
|
|
|
|
*
|
|
|
|
* Except for dsl_deadlist_insert(), it can only be modified with the
|
|
|
|
* dp_config_rwlock held with RW_WRITER.
|
|
|
|
*
|
|
|
|
* The accessors (dsl_deadlist_space() and dsl_deadlist_space_range()) can
|
|
|
|
* be called concurrently, from open context, with the dl_config_rwlock held
|
|
|
|
* with RW_READER.
|
|
|
|
*
|
|
|
|
* Therefore, we only need to provide locking between dsl_deadlist_insert() and
|
|
|
|
* the accessors, protecting:
|
|
|
|
* dl_phys->dl_used,comp,uncomp
|
|
|
|
* and protecting the dl_tree from being loaded.
|
|
|
|
* The locking is provided by dl_lock. Note that locking on the bpobj_t
|
|
|
|
* provides its own locking, and dl_oldfmt is immutable.
|
|
|
|
*/
|
|
|
|
|
2010-05-28 20:45:14 +00:00
|
|
|
static int
|
|
|
|
dsl_deadlist_compare(const void *arg1, const void *arg2)
|
|
|
|
{
|
2016-08-27 18:12:53 +00:00
|
|
|
const dsl_deadlist_entry_t *dle1 = (const dsl_deadlist_entry_t *)arg1;
|
|
|
|
const dsl_deadlist_entry_t *dle2 = (const dsl_deadlist_entry_t *)arg2;
|
2010-05-28 20:45:14 +00:00
|
|
|
|
2016-08-27 18:12:53 +00:00
|
|
|
return (AVL_CMP(dle1->dle_mintxg, dle2->dle_mintxg));
|
2010-05-28 20:45:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
dsl_deadlist_load_tree(dsl_deadlist_t *dl)
|
|
|
|
{
|
|
|
|
zap_cursor_t zc;
|
|
|
|
zap_attribute_t za;
|
|
|
|
|
2017-02-09 18:19:12 +00:00
|
|
|
ASSERT(MUTEX_HELD(&dl->dl_lock));
|
|
|
|
|
2010-05-28 20:45:14 +00:00
|
|
|
ASSERT(!dl->dl_oldfmt);
|
|
|
|
if (dl->dl_havetree)
|
|
|
|
return;
|
|
|
|
|
|
|
|
avl_create(&dl->dl_tree, dsl_deadlist_compare,
|
|
|
|
sizeof (dsl_deadlist_entry_t),
|
|
|
|
offsetof(dsl_deadlist_entry_t, dle_node));
|
|
|
|
for (zap_cursor_init(&zc, dl->dl_os, dl->dl_object);
|
|
|
|
zap_cursor_retrieve(&zc, &za) == 0;
|
|
|
|
zap_cursor_advance(&zc)) {
|
2017-06-13 03:16:28 +00:00
|
|
|
dsl_deadlist_entry_t *dle = kmem_alloc(sizeof (*dle), KM_SLEEP);
|
|
|
|
dle->dle_mintxg = zfs_strtonum(za.za_name, NULL);
|
2010-05-28 20:45:14 +00:00
|
|
|
VERIFY3U(0, ==, bpobj_open(&dle->dle_bpobj, dl->dl_os,
|
|
|
|
za.za_first_integer));
|
|
|
|
avl_add(&dl->dl_tree, dle);
|
|
|
|
}
|
|
|
|
zap_cursor_fini(&zc);
|
|
|
|
dl->dl_havetree = B_TRUE;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
dsl_deadlist_open(dsl_deadlist_t *dl, objset_t *os, uint64_t object)
|
|
|
|
{
|
|
|
|
dmu_object_info_t doi;
|
|
|
|
|
OpenZFS 7614, 9064 - zfs device evacuation/removal
OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete
This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk. The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.
The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool. An entry becomes obsolete when all the blocks that use
it are freed. An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones). Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible. This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.
Note that when a device is removed, we do not verify the checksum of
the data that is copied. This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.
At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.
Porting Notes:
* Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
* Remove comment regarding "mpt" driver where zfs_remove_max_segment
is initialized to SPA_MAXBLOCKSIZE.
Change zfs_condense_indirect_commit_entry_delay_ticks to
zfs_condense_indirect_commit_entry_delay_ms for consistency with
most other tunables in which delays are specified in ms.
* ZTS changes:
Use set_tunable rather than mdb
Use zpool sync as appropriate
Use sync_pool instead of sync
Kill jobs during test_removal_with_operation to allow unmount/export
Don't add non-disk names such as "mirror" or "raidz" to $DISKS
Use $TEST_BASE_DIR instead of /tmp
Increase HZ from 100 to 1000 which is more common on Linux
removal_multiple_indirection.ksh
Reduce iterations in order to not time out on the code
coverage builders.
removal_resume_export:
Functionally, the test case is correct but there exists a race
where the kernel thread hasn't been fully started yet and is
not visible. Wait for up to 1 second for the removal thread
to be started before giving up on it. Also, increase the
amount of data copied in order that the removal not finish
before the export has a chance to fail.
* MMP compatibility, the concept of concrete versus non-concrete devices
has slightly changed the semantics of vdev_writeable(). Update
mmp_random_leaf_impl() accordingly.
* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
feature which is not supported by OpenZFS.
* Added support for new vdev removal tracepoints.
* Test cases removal_with_zdb and removal_condense_export have been
intentionally disabled. When run manually they pass as intended,
but when running in the automated test environment they produce
unreliable results on the latest Fedora release.
They may work better once the upstream pool import refectoring is
merged into ZoL at which point they will be re-enabled.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alex Reece <alex@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Richard Laager <rlaager@wiktel.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Garrett D'Amore <garrett@damore.org>
Ported-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900
2016-09-22 16:30:13 +00:00
|
|
|
ASSERT(!dsl_deadlist_is_open(dl));
|
|
|
|
|
2010-05-28 20:45:14 +00:00
|
|
|
mutex_init(&dl->dl_lock, NULL, MUTEX_DEFAULT, NULL);
|
|
|
|
dl->dl_os = os;
|
|
|
|
dl->dl_object = object;
|
|
|
|
VERIFY3U(0, ==, dmu_bonus_hold(os, object, dl, &dl->dl_dbuf));
|
|
|
|
dmu_object_info_from_db(dl->dl_dbuf, &doi);
|
|
|
|
if (doi.doi_type == DMU_OT_BPOBJ) {
|
|
|
|
dmu_buf_rele(dl->dl_dbuf, dl);
|
|
|
|
dl->dl_dbuf = NULL;
|
|
|
|
dl->dl_oldfmt = B_TRUE;
|
|
|
|
VERIFY3U(0, ==, bpobj_open(&dl->dl_bpobj, os, object));
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
dl->dl_oldfmt = B_FALSE;
|
|
|
|
dl->dl_phys = dl->dl_dbuf->db_data;
|
|
|
|
dl->dl_havetree = B_FALSE;
|
|
|
|
}
|
|
|
|
|
OpenZFS 7614, 9064 - zfs device evacuation/removal
OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete
This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk. The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.
The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool. An entry becomes obsolete when all the blocks that use
it are freed. An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones). Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible. This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.
Note that when a device is removed, we do not verify the checksum of
the data that is copied. This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.
At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.
Porting Notes:
* Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
* Remove comment regarding "mpt" driver where zfs_remove_max_segment
is initialized to SPA_MAXBLOCKSIZE.
Change zfs_condense_indirect_commit_entry_delay_ticks to
zfs_condense_indirect_commit_entry_delay_ms for consistency with
most other tunables in which delays are specified in ms.
* ZTS changes:
Use set_tunable rather than mdb
Use zpool sync as appropriate
Use sync_pool instead of sync
Kill jobs during test_removal_with_operation to allow unmount/export
Don't add non-disk names such as "mirror" or "raidz" to $DISKS
Use $TEST_BASE_DIR instead of /tmp
Increase HZ from 100 to 1000 which is more common on Linux
removal_multiple_indirection.ksh
Reduce iterations in order to not time out on the code
coverage builders.
removal_resume_export:
Functionally, the test case is correct but there exists a race
where the kernel thread hasn't been fully started yet and is
not visible. Wait for up to 1 second for the removal thread
to be started before giving up on it. Also, increase the
amount of data copied in order that the removal not finish
before the export has a chance to fail.
* MMP compatibility, the concept of concrete versus non-concrete devices
has slightly changed the semantics of vdev_writeable(). Update
mmp_random_leaf_impl() accordingly.
* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
feature which is not supported by OpenZFS.
* Added support for new vdev removal tracepoints.
* Test cases removal_with_zdb and removal_condense_export have been
intentionally disabled. When run manually they pass as intended,
but when running in the automated test environment they produce
unreliable results on the latest Fedora release.
They may work better once the upstream pool import refectoring is
merged into ZoL at which point they will be re-enabled.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alex Reece <alex@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Richard Laager <rlaager@wiktel.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Garrett D'Amore <garrett@damore.org>
Ported-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900
2016-09-22 16:30:13 +00:00
|
|
|
boolean_t
|
|
|
|
dsl_deadlist_is_open(dsl_deadlist_t *dl)
|
|
|
|
{
|
|
|
|
return (dl->dl_os != NULL);
|
|
|
|
}
|
|
|
|
|
2010-05-28 20:45:14 +00:00
|
|
|
void
|
|
|
|
dsl_deadlist_close(dsl_deadlist_t *dl)
|
|
|
|
{
|
|
|
|
void *cookie = NULL;
|
|
|
|
dsl_deadlist_entry_t *dle;
|
|
|
|
|
OpenZFS 7614, 9064 - zfs device evacuation/removal
OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete
This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk. The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.
The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool. An entry becomes obsolete when all the blocks that use
it are freed. An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones). Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible. This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.
Note that when a device is removed, we do not verify the checksum of
the data that is copied. This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.
At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.
Porting Notes:
* Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
* Remove comment regarding "mpt" driver where zfs_remove_max_segment
is initialized to SPA_MAXBLOCKSIZE.
Change zfs_condense_indirect_commit_entry_delay_ticks to
zfs_condense_indirect_commit_entry_delay_ms for consistency with
most other tunables in which delays are specified in ms.
* ZTS changes:
Use set_tunable rather than mdb
Use zpool sync as appropriate
Use sync_pool instead of sync
Kill jobs during test_removal_with_operation to allow unmount/export
Don't add non-disk names such as "mirror" or "raidz" to $DISKS
Use $TEST_BASE_DIR instead of /tmp
Increase HZ from 100 to 1000 which is more common on Linux
removal_multiple_indirection.ksh
Reduce iterations in order to not time out on the code
coverage builders.
removal_resume_export:
Functionally, the test case is correct but there exists a race
where the kernel thread hasn't been fully started yet and is
not visible. Wait for up to 1 second for the removal thread
to be started before giving up on it. Also, increase the
amount of data copied in order that the removal not finish
before the export has a chance to fail.
* MMP compatibility, the concept of concrete versus non-concrete devices
has slightly changed the semantics of vdev_writeable(). Update
mmp_random_leaf_impl() accordingly.
* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
feature which is not supported by OpenZFS.
* Added support for new vdev removal tracepoints.
* Test cases removal_with_zdb and removal_condense_export have been
intentionally disabled. When run manually they pass as intended,
but when running in the automated test environment they produce
unreliable results on the latest Fedora release.
They may work better once the upstream pool import refectoring is
merged into ZoL at which point they will be re-enabled.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alex Reece <alex@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Richard Laager <rlaager@wiktel.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Garrett D'Amore <garrett@damore.org>
Ported-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900
2016-09-22 16:30:13 +00:00
|
|
|
ASSERT(dsl_deadlist_is_open(dl));
|
2016-11-26 20:30:44 +00:00
|
|
|
mutex_destroy(&dl->dl_lock);
|
2015-04-02 03:44:32 +00:00
|
|
|
|
2010-05-28 20:45:14 +00:00
|
|
|
if (dl->dl_oldfmt) {
|
|
|
|
dl->dl_oldfmt = B_FALSE;
|
|
|
|
bpobj_close(&dl->dl_bpobj);
|
OpenZFS 7614, 9064 - zfs device evacuation/removal
OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete
This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk. The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.
The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool. An entry becomes obsolete when all the blocks that use
it are freed. An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones). Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible. This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.
Note that when a device is removed, we do not verify the checksum of
the data that is copied. This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.
At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.
Porting Notes:
* Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
* Remove comment regarding "mpt" driver where zfs_remove_max_segment
is initialized to SPA_MAXBLOCKSIZE.
Change zfs_condense_indirect_commit_entry_delay_ticks to
zfs_condense_indirect_commit_entry_delay_ms for consistency with
most other tunables in which delays are specified in ms.
* ZTS changes:
Use set_tunable rather than mdb
Use zpool sync as appropriate
Use sync_pool instead of sync
Kill jobs during test_removal_with_operation to allow unmount/export
Don't add non-disk names such as "mirror" or "raidz" to $DISKS
Use $TEST_BASE_DIR instead of /tmp
Increase HZ from 100 to 1000 which is more common on Linux
removal_multiple_indirection.ksh
Reduce iterations in order to not time out on the code
coverage builders.
removal_resume_export:
Functionally, the test case is correct but there exists a race
where the kernel thread hasn't been fully started yet and is
not visible. Wait for up to 1 second for the removal thread
to be started before giving up on it. Also, increase the
amount of data copied in order that the removal not finish
before the export has a chance to fail.
* MMP compatibility, the concept of concrete versus non-concrete devices
has slightly changed the semantics of vdev_writeable(). Update
mmp_random_leaf_impl() accordingly.
* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
feature which is not supported by OpenZFS.
* Added support for new vdev removal tracepoints.
* Test cases removal_with_zdb and removal_condense_export have been
intentionally disabled. When run manually they pass as intended,
but when running in the automated test environment they produce
unreliable results on the latest Fedora release.
They may work better once the upstream pool import refectoring is
merged into ZoL at which point they will be re-enabled.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alex Reece <alex@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Richard Laager <rlaager@wiktel.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Garrett D'Amore <garrett@damore.org>
Ported-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900
2016-09-22 16:30:13 +00:00
|
|
|
dl->dl_os = NULL;
|
|
|
|
dl->dl_object = 0;
|
2010-05-28 20:45:14 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (dl->dl_havetree) {
|
|
|
|
while ((dle = avl_destroy_nodes(&dl->dl_tree, &cookie))
|
|
|
|
!= NULL) {
|
|
|
|
bpobj_close(&dle->dle_bpobj);
|
|
|
|
kmem_free(dle, sizeof (*dle));
|
|
|
|
}
|
|
|
|
avl_destroy(&dl->dl_tree);
|
|
|
|
}
|
|
|
|
dmu_buf_rele(dl->dl_dbuf, dl);
|
|
|
|
dl->dl_dbuf = NULL;
|
|
|
|
dl->dl_phys = NULL;
|
OpenZFS 7614, 9064 - zfs device evacuation/removal
OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete
This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk. The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.
The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool. An entry becomes obsolete when all the blocks that use
it are freed. An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones). Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible. This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.
Note that when a device is removed, we do not verify the checksum of
the data that is copied. This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.
At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.
Porting Notes:
* Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
* Remove comment regarding "mpt" driver where zfs_remove_max_segment
is initialized to SPA_MAXBLOCKSIZE.
Change zfs_condense_indirect_commit_entry_delay_ticks to
zfs_condense_indirect_commit_entry_delay_ms for consistency with
most other tunables in which delays are specified in ms.
* ZTS changes:
Use set_tunable rather than mdb
Use zpool sync as appropriate
Use sync_pool instead of sync
Kill jobs during test_removal_with_operation to allow unmount/export
Don't add non-disk names such as "mirror" or "raidz" to $DISKS
Use $TEST_BASE_DIR instead of /tmp
Increase HZ from 100 to 1000 which is more common on Linux
removal_multiple_indirection.ksh
Reduce iterations in order to not time out on the code
coverage builders.
removal_resume_export:
Functionally, the test case is correct but there exists a race
where the kernel thread hasn't been fully started yet and is
not visible. Wait for up to 1 second for the removal thread
to be started before giving up on it. Also, increase the
amount of data copied in order that the removal not finish
before the export has a chance to fail.
* MMP compatibility, the concept of concrete versus non-concrete devices
has slightly changed the semantics of vdev_writeable(). Update
mmp_random_leaf_impl() accordingly.
* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
feature which is not supported by OpenZFS.
* Added support for new vdev removal tracepoints.
* Test cases removal_with_zdb and removal_condense_export have been
intentionally disabled. When run manually they pass as intended,
but when running in the automated test environment they produce
unreliable results on the latest Fedora release.
They may work better once the upstream pool import refectoring is
merged into ZoL at which point they will be re-enabled.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alex Reece <alex@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Richard Laager <rlaager@wiktel.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Garrett D'Amore <garrett@damore.org>
Ported-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900
2016-09-22 16:30:13 +00:00
|
|
|
dl->dl_os = NULL;
|
|
|
|
dl->dl_object = 0;
|
2010-05-28 20:45:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t
|
|
|
|
dsl_deadlist_alloc(objset_t *os, dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
|
2014-11-03 20:15:08 +00:00
|
|
|
return (bpobj_alloc(os, SPA_OLD_MAXBLOCKSIZE, tx));
|
2010-05-28 20:45:14 +00:00
|
|
|
return (zap_create(os, DMU_OT_DEADLIST, DMU_OT_DEADLIST_HDR,
|
|
|
|
sizeof (dsl_deadlist_phys_t), tx));
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
dsl_deadlist_free(objset_t *os, uint64_t dlobj, dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
dmu_object_info_t doi;
|
|
|
|
zap_cursor_t zc;
|
|
|
|
zap_attribute_t za;
|
|
|
|
|
|
|
|
VERIFY3U(0, ==, dmu_object_info(os, dlobj, &doi));
|
|
|
|
if (doi.doi_type == DMU_OT_BPOBJ) {
|
|
|
|
bpobj_free(os, dlobj, tx);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (zap_cursor_init(&zc, os, dlobj);
|
|
|
|
zap_cursor_retrieve(&zc, &za) == 0;
|
2012-12-23 23:57:14 +00:00
|
|
|
zap_cursor_advance(&zc)) {
|
|
|
|
uint64_t obj = za.za_first_integer;
|
|
|
|
if (obj == dmu_objset_pool(os)->dp_empty_bpobj)
|
|
|
|
bpobj_decr_empty(os, tx);
|
|
|
|
else
|
|
|
|
bpobj_free(os, obj, tx);
|
|
|
|
}
|
2010-05-28 20:45:14 +00:00
|
|
|
zap_cursor_fini(&zc);
|
|
|
|
VERIFY3U(0, ==, dmu_object_free(os, dlobj, tx));
|
|
|
|
}
|
|
|
|
|
2012-12-23 23:57:14 +00:00
|
|
|
static void
|
|
|
|
dle_enqueue(dsl_deadlist_t *dl, dsl_deadlist_entry_t *dle,
|
|
|
|
const blkptr_t *bp, dmu_tx_t *tx)
|
|
|
|
{
|
2017-02-09 18:19:12 +00:00
|
|
|
ASSERT(MUTEX_HELD(&dl->dl_lock));
|
2012-12-23 23:57:14 +00:00
|
|
|
if (dle->dle_bpobj.bpo_object ==
|
|
|
|
dmu_objset_pool(dl->dl_os)->dp_empty_bpobj) {
|
2014-11-03 20:15:08 +00:00
|
|
|
uint64_t obj = bpobj_alloc(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
|
2012-12-23 23:57:14 +00:00
|
|
|
bpobj_close(&dle->dle_bpobj);
|
|
|
|
bpobj_decr_empty(dl->dl_os, tx);
|
|
|
|
VERIFY3U(0, ==, bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
|
|
|
|
VERIFY3U(0, ==, zap_update_int_key(dl->dl_os, dl->dl_object,
|
|
|
|
dle->dle_mintxg, obj, tx));
|
|
|
|
}
|
|
|
|
bpobj_enqueue(&dle->dle_bpobj, bp, tx);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
dle_enqueue_subobj(dsl_deadlist_t *dl, dsl_deadlist_entry_t *dle,
|
|
|
|
uint64_t obj, dmu_tx_t *tx)
|
|
|
|
{
|
2017-02-09 18:19:12 +00:00
|
|
|
ASSERT(MUTEX_HELD(&dl->dl_lock));
|
2012-12-23 23:57:14 +00:00
|
|
|
if (dle->dle_bpobj.bpo_object !=
|
|
|
|
dmu_objset_pool(dl->dl_os)->dp_empty_bpobj) {
|
|
|
|
bpobj_enqueue_subobj(&dle->dle_bpobj, obj, tx);
|
|
|
|
} else {
|
|
|
|
bpobj_close(&dle->dle_bpobj);
|
|
|
|
bpobj_decr_empty(dl->dl_os, tx);
|
|
|
|
VERIFY3U(0, ==, bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
|
|
|
|
VERIFY3U(0, ==, zap_update_int_key(dl->dl_os, dl->dl_object,
|
|
|
|
dle->dle_mintxg, obj, tx));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-05-28 20:45:14 +00:00
|
|
|
void
|
|
|
|
dsl_deadlist_insert(dsl_deadlist_t *dl, const blkptr_t *bp, dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
dsl_deadlist_entry_t dle_tofind;
|
|
|
|
dsl_deadlist_entry_t *dle;
|
|
|
|
avl_index_t where;
|
|
|
|
|
|
|
|
if (dl->dl_oldfmt) {
|
|
|
|
bpobj_enqueue(&dl->dl_bpobj, bp, tx);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_enter(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
dsl_deadlist_load_tree(dl);
|
|
|
|
|
|
|
|
dmu_buf_will_dirty(dl->dl_dbuf, tx);
|
|
|
|
dl->dl_phys->dl_used +=
|
|
|
|
bp_get_dsize_sync(dmu_objset_spa(dl->dl_os), bp);
|
|
|
|
dl->dl_phys->dl_comp += BP_GET_PSIZE(bp);
|
|
|
|
dl->dl_phys->dl_uncomp += BP_GET_UCSIZE(bp);
|
|
|
|
|
|
|
|
dle_tofind.dle_mintxg = bp->blk_birth;
|
|
|
|
dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
|
|
|
|
if (dle == NULL)
|
|
|
|
dle = avl_nearest(&dl->dl_tree, where, AVL_BEFORE);
|
|
|
|
else
|
|
|
|
dle = AVL_PREV(&dl->dl_tree, dle);
|
2015-12-11 19:09:41 +00:00
|
|
|
|
|
|
|
if (dle == NULL) {
|
|
|
|
zfs_panic_recover("blkptr at %p has invalid BLK_BIRTH %llu",
|
|
|
|
bp, (longlong_t)bp->blk_birth);
|
|
|
|
dle = avl_first(&dl->dl_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
ASSERT3P(dle, !=, NULL);
|
2012-12-23 23:57:14 +00:00
|
|
|
dle_enqueue(dl, dle, bp, tx);
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_exit(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Insert new key in deadlist, which must be > all current entries.
|
|
|
|
* mintxg is not inclusive.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
dsl_deadlist_add_key(dsl_deadlist_t *dl, uint64_t mintxg, dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
uint64_t obj;
|
|
|
|
dsl_deadlist_entry_t *dle;
|
|
|
|
|
|
|
|
if (dl->dl_oldfmt)
|
|
|
|
return;
|
|
|
|
|
2014-11-21 00:09:39 +00:00
|
|
|
dle = kmem_alloc(sizeof (*dle), KM_SLEEP);
|
2010-05-28 20:45:14 +00:00
|
|
|
dle->dle_mintxg = mintxg;
|
2017-02-09 18:19:12 +00:00
|
|
|
|
|
|
|
mutex_enter(&dl->dl_lock);
|
|
|
|
dsl_deadlist_load_tree(dl);
|
|
|
|
|
2014-11-03 20:15:08 +00:00
|
|
|
obj = bpobj_alloc_empty(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
|
2010-05-28 20:45:14 +00:00
|
|
|
VERIFY3U(0, ==, bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
|
|
|
|
avl_add(&dl->dl_tree, dle);
|
|
|
|
|
|
|
|
VERIFY3U(0, ==, zap_add_int_key(dl->dl_os, dl->dl_object,
|
|
|
|
mintxg, obj, tx));
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_exit(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Remove this key, merging its entries into the previous key.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
dsl_deadlist_remove_key(dsl_deadlist_t *dl, uint64_t mintxg, dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
dsl_deadlist_entry_t dle_tofind;
|
|
|
|
dsl_deadlist_entry_t *dle, *dle_prev;
|
|
|
|
|
|
|
|
if (dl->dl_oldfmt)
|
|
|
|
return;
|
|
|
|
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_enter(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
dsl_deadlist_load_tree(dl);
|
|
|
|
|
|
|
|
dle_tofind.dle_mintxg = mintxg;
|
|
|
|
dle = avl_find(&dl->dl_tree, &dle_tofind, NULL);
|
|
|
|
dle_prev = AVL_PREV(&dl->dl_tree, dle);
|
|
|
|
|
2012-12-23 23:57:14 +00:00
|
|
|
dle_enqueue_subobj(dl, dle_prev, dle->dle_bpobj.bpo_object, tx);
|
2010-05-28 20:45:14 +00:00
|
|
|
|
|
|
|
avl_remove(&dl->dl_tree, dle);
|
|
|
|
bpobj_close(&dle->dle_bpobj);
|
|
|
|
kmem_free(dle, sizeof (*dle));
|
|
|
|
|
|
|
|
VERIFY3U(0, ==, zap_remove_int(dl->dl_os, dl->dl_object, mintxg, tx));
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_exit(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Walk ds's snapshots to regenerate generate ZAP & AVL.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
dsl_deadlist_regenerate(objset_t *os, uint64_t dlobj,
|
|
|
|
uint64_t mrs_obj, dmu_tx_t *tx)
|
|
|
|
{
|
OpenZFS 7614, 9064 - zfs device evacuation/removal
OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete
This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk. The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.
The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool. An entry becomes obsolete when all the blocks that use
it are freed. An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones). Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible. This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.
Note that when a device is removed, we do not verify the checksum of
the data that is copied. This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.
At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.
Porting Notes:
* Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
* Remove comment regarding "mpt" driver where zfs_remove_max_segment
is initialized to SPA_MAXBLOCKSIZE.
Change zfs_condense_indirect_commit_entry_delay_ticks to
zfs_condense_indirect_commit_entry_delay_ms for consistency with
most other tunables in which delays are specified in ms.
* ZTS changes:
Use set_tunable rather than mdb
Use zpool sync as appropriate
Use sync_pool instead of sync
Kill jobs during test_removal_with_operation to allow unmount/export
Don't add non-disk names such as "mirror" or "raidz" to $DISKS
Use $TEST_BASE_DIR instead of /tmp
Increase HZ from 100 to 1000 which is more common on Linux
removal_multiple_indirection.ksh
Reduce iterations in order to not time out on the code
coverage builders.
removal_resume_export:
Functionally, the test case is correct but there exists a race
where the kernel thread hasn't been fully started yet and is
not visible. Wait for up to 1 second for the removal thread
to be started before giving up on it. Also, increase the
amount of data copied in order that the removal not finish
before the export has a chance to fail.
* MMP compatibility, the concept of concrete versus non-concrete devices
has slightly changed the semantics of vdev_writeable(). Update
mmp_random_leaf_impl() accordingly.
* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
feature which is not supported by OpenZFS.
* Added support for new vdev removal tracepoints.
* Test cases removal_with_zdb and removal_condense_export have been
intentionally disabled. When run manually they pass as intended,
but when running in the automated test environment they produce
unreliable results on the latest Fedora release.
They may work better once the upstream pool import refectoring is
merged into ZoL at which point they will be re-enabled.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alex Reece <alex@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Richard Laager <rlaager@wiktel.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Garrett D'Amore <garrett@damore.org>
Ported-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900
2016-09-22 16:30:13 +00:00
|
|
|
dsl_deadlist_t dl = { 0 };
|
2010-05-28 20:45:14 +00:00
|
|
|
dsl_pool_t *dp = dmu_objset_pool(os);
|
|
|
|
|
|
|
|
dsl_deadlist_open(&dl, os, dlobj);
|
|
|
|
if (dl.dl_oldfmt) {
|
|
|
|
dsl_deadlist_close(&dl);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
while (mrs_obj != 0) {
|
|
|
|
dsl_dataset_t *ds;
|
|
|
|
VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, mrs_obj, FTAG, &ds));
|
2015-04-01 15:14:34 +00:00
|
|
|
dsl_deadlist_add_key(&dl,
|
|
|
|
dsl_dataset_phys(ds)->ds_prev_snap_txg, tx);
|
|
|
|
mrs_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
|
2010-05-28 20:45:14 +00:00
|
|
|
dsl_dataset_rele(ds, FTAG);
|
|
|
|
}
|
|
|
|
dsl_deadlist_close(&dl);
|
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t
|
|
|
|
dsl_deadlist_clone(dsl_deadlist_t *dl, uint64_t maxtxg,
|
|
|
|
uint64_t mrs_obj, dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
dsl_deadlist_entry_t *dle;
|
|
|
|
uint64_t newobj;
|
|
|
|
|
|
|
|
newobj = dsl_deadlist_alloc(dl->dl_os, tx);
|
|
|
|
|
|
|
|
if (dl->dl_oldfmt) {
|
|
|
|
dsl_deadlist_regenerate(dl->dl_os, newobj, mrs_obj, tx);
|
|
|
|
return (newobj);
|
|
|
|
}
|
|
|
|
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_enter(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
dsl_deadlist_load_tree(dl);
|
|
|
|
|
|
|
|
for (dle = avl_first(&dl->dl_tree); dle;
|
|
|
|
dle = AVL_NEXT(&dl->dl_tree, dle)) {
|
|
|
|
uint64_t obj;
|
|
|
|
|
|
|
|
if (dle->dle_mintxg >= maxtxg)
|
|
|
|
break;
|
|
|
|
|
2014-11-03 20:15:08 +00:00
|
|
|
obj = bpobj_alloc_empty(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
|
2010-05-28 20:45:14 +00:00
|
|
|
VERIFY3U(0, ==, zap_add_int_key(dl->dl_os, newobj,
|
|
|
|
dle->dle_mintxg, obj, tx));
|
|
|
|
}
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_exit(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
return (newobj);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
dsl_deadlist_space(dsl_deadlist_t *dl,
|
|
|
|
uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
|
|
|
|
{
|
OpenZFS 7614, 9064 - zfs device evacuation/removal
OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete
This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk. The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.
The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool. An entry becomes obsolete when all the blocks that use
it are freed. An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones). Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible. This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.
Note that when a device is removed, we do not verify the checksum of
the data that is copied. This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.
At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.
Porting Notes:
* Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
* Remove comment regarding "mpt" driver where zfs_remove_max_segment
is initialized to SPA_MAXBLOCKSIZE.
Change zfs_condense_indirect_commit_entry_delay_ticks to
zfs_condense_indirect_commit_entry_delay_ms for consistency with
most other tunables in which delays are specified in ms.
* ZTS changes:
Use set_tunable rather than mdb
Use zpool sync as appropriate
Use sync_pool instead of sync
Kill jobs during test_removal_with_operation to allow unmount/export
Don't add non-disk names such as "mirror" or "raidz" to $DISKS
Use $TEST_BASE_DIR instead of /tmp
Increase HZ from 100 to 1000 which is more common on Linux
removal_multiple_indirection.ksh
Reduce iterations in order to not time out on the code
coverage builders.
removal_resume_export:
Functionally, the test case is correct but there exists a race
where the kernel thread hasn't been fully started yet and is
not visible. Wait for up to 1 second for the removal thread
to be started before giving up on it. Also, increase the
amount of data copied in order that the removal not finish
before the export has a chance to fail.
* MMP compatibility, the concept of concrete versus non-concrete devices
has slightly changed the semantics of vdev_writeable(). Update
mmp_random_leaf_impl() accordingly.
* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
feature which is not supported by OpenZFS.
* Added support for new vdev removal tracepoints.
* Test cases removal_with_zdb and removal_condense_export have been
intentionally disabled. When run manually they pass as intended,
but when running in the automated test environment they produce
unreliable results on the latest Fedora release.
They may work better once the upstream pool import refectoring is
merged into ZoL at which point they will be re-enabled.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alex Reece <alex@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Richard Laager <rlaager@wiktel.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Garrett D'Amore <garrett@damore.org>
Ported-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900
2016-09-22 16:30:13 +00:00
|
|
|
ASSERT(dsl_deadlist_is_open(dl));
|
2010-05-28 20:45:14 +00:00
|
|
|
if (dl->dl_oldfmt) {
|
|
|
|
VERIFY3U(0, ==, bpobj_space(&dl->dl_bpobj,
|
|
|
|
usedp, compp, uncompp));
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
mutex_enter(&dl->dl_lock);
|
|
|
|
*usedp = dl->dl_phys->dl_used;
|
|
|
|
*compp = dl->dl_phys->dl_comp;
|
|
|
|
*uncompp = dl->dl_phys->dl_uncomp;
|
|
|
|
mutex_exit(&dl->dl_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return space used in the range (mintxg, maxtxg].
|
|
|
|
* Includes maxtxg, does not include mintxg.
|
|
|
|
* mintxg and maxtxg must both be keys in the deadlist (unless maxtxg is
|
2011-11-17 18:14:36 +00:00
|
|
|
* larger than any bp in the deadlist (eg. UINT64_MAX)).
|
2010-05-28 20:45:14 +00:00
|
|
|
*/
|
|
|
|
void
|
|
|
|
dsl_deadlist_space_range(dsl_deadlist_t *dl, uint64_t mintxg, uint64_t maxtxg,
|
|
|
|
uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
|
|
|
|
{
|
|
|
|
dsl_deadlist_entry_t *dle;
|
2011-11-17 18:14:36 +00:00
|
|
|
dsl_deadlist_entry_t dle_tofind;
|
2010-05-28 20:45:14 +00:00
|
|
|
avl_index_t where;
|
|
|
|
|
|
|
|
if (dl->dl_oldfmt) {
|
|
|
|
VERIFY3U(0, ==, bpobj_space_range(&dl->dl_bpobj,
|
|
|
|
mintxg, maxtxg, usedp, compp, uncompp));
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
*usedp = *compp = *uncompp = 0;
|
|
|
|
|
2011-11-17 18:14:36 +00:00
|
|
|
mutex_enter(&dl->dl_lock);
|
|
|
|
dsl_deadlist_load_tree(dl);
|
2010-05-28 20:45:14 +00:00
|
|
|
dle_tofind.dle_mintxg = mintxg;
|
|
|
|
dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
|
|
|
|
/*
|
|
|
|
* If we don't find this mintxg, there shouldn't be anything
|
|
|
|
* after it either.
|
|
|
|
*/
|
|
|
|
ASSERT(dle != NULL ||
|
|
|
|
avl_nearest(&dl->dl_tree, where, AVL_AFTER) == NULL);
|
2011-11-17 18:14:36 +00:00
|
|
|
|
2010-05-28 20:45:14 +00:00
|
|
|
for (; dle && dle->dle_mintxg < maxtxg;
|
|
|
|
dle = AVL_NEXT(&dl->dl_tree, dle)) {
|
|
|
|
uint64_t used, comp, uncomp;
|
|
|
|
|
|
|
|
VERIFY3U(0, ==, bpobj_space(&dle->dle_bpobj,
|
|
|
|
&used, &comp, &uncomp));
|
|
|
|
|
|
|
|
*usedp += used;
|
|
|
|
*compp += comp;
|
|
|
|
*uncompp += uncomp;
|
|
|
|
}
|
2011-11-17 18:14:36 +00:00
|
|
|
mutex_exit(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
dsl_deadlist_insert_bpobj(dsl_deadlist_t *dl, uint64_t obj, uint64_t birth,
|
|
|
|
dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
dsl_deadlist_entry_t dle_tofind;
|
|
|
|
dsl_deadlist_entry_t *dle;
|
|
|
|
avl_index_t where;
|
|
|
|
uint64_t used, comp, uncomp;
|
|
|
|
bpobj_t bpo;
|
|
|
|
|
2017-02-09 18:19:12 +00:00
|
|
|
ASSERT(MUTEX_HELD(&dl->dl_lock));
|
|
|
|
|
2010-05-28 20:45:14 +00:00
|
|
|
VERIFY3U(0, ==, bpobj_open(&bpo, dl->dl_os, obj));
|
|
|
|
VERIFY3U(0, ==, bpobj_space(&bpo, &used, &comp, &uncomp));
|
|
|
|
bpobj_close(&bpo);
|
|
|
|
|
|
|
|
dsl_deadlist_load_tree(dl);
|
|
|
|
|
|
|
|
dmu_buf_will_dirty(dl->dl_dbuf, tx);
|
|
|
|
dl->dl_phys->dl_used += used;
|
|
|
|
dl->dl_phys->dl_comp += comp;
|
|
|
|
dl->dl_phys->dl_uncomp += uncomp;
|
|
|
|
|
|
|
|
dle_tofind.dle_mintxg = birth;
|
|
|
|
dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
|
|
|
|
if (dle == NULL)
|
|
|
|
dle = avl_nearest(&dl->dl_tree, where, AVL_BEFORE);
|
2012-12-23 23:57:14 +00:00
|
|
|
dle_enqueue_subobj(dl, dle, obj, tx);
|
2010-05-28 20:45:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
dsl_deadlist_insert_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
dsl_deadlist_t *dl = arg;
|
|
|
|
dsl_deadlist_insert(dl, bp, tx);
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Merge the deadlist pointed to by 'obj' into dl. obj will be left as
|
|
|
|
* an empty deadlist.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
dsl_deadlist_merge(dsl_deadlist_t *dl, uint64_t obj, dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
zap_cursor_t zc;
|
|
|
|
zap_attribute_t za;
|
|
|
|
dmu_buf_t *bonus;
|
|
|
|
dsl_deadlist_phys_t *dlp;
|
|
|
|
dmu_object_info_t doi;
|
|
|
|
|
|
|
|
VERIFY3U(0, ==, dmu_object_info(dl->dl_os, obj, &doi));
|
|
|
|
if (doi.doi_type == DMU_OT_BPOBJ) {
|
|
|
|
bpobj_t bpo;
|
|
|
|
VERIFY3U(0, ==, bpobj_open(&bpo, dl->dl_os, obj));
|
|
|
|
VERIFY3U(0, ==, bpobj_iterate(&bpo,
|
|
|
|
dsl_deadlist_insert_cb, dl, tx));
|
|
|
|
bpobj_close(&bpo);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_enter(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
for (zap_cursor_init(&zc, dl->dl_os, obj);
|
|
|
|
zap_cursor_retrieve(&zc, &za) == 0;
|
|
|
|
zap_cursor_advance(&zc)) {
|
2017-06-13 03:16:28 +00:00
|
|
|
uint64_t mintxg = zfs_strtonum(za.za_name, NULL);
|
2010-05-28 20:45:14 +00:00
|
|
|
dsl_deadlist_insert_bpobj(dl, za.za_first_integer, mintxg, tx);
|
|
|
|
VERIFY3U(0, ==, zap_remove_int(dl->dl_os, obj, mintxg, tx));
|
|
|
|
}
|
|
|
|
zap_cursor_fini(&zc);
|
|
|
|
|
|
|
|
VERIFY3U(0, ==, dmu_bonus_hold(dl->dl_os, obj, FTAG, &bonus));
|
|
|
|
dlp = bonus->db_data;
|
|
|
|
dmu_buf_will_dirty(bonus, tx);
|
|
|
|
bzero(dlp, sizeof (*dlp));
|
|
|
|
dmu_buf_rele(bonus, FTAG);
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_exit(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Remove entries on dl that are >= mintxg, and put them on the bpobj.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
dsl_deadlist_move_bpobj(dsl_deadlist_t *dl, bpobj_t *bpo, uint64_t mintxg,
|
|
|
|
dmu_tx_t *tx)
|
|
|
|
{
|
|
|
|
dsl_deadlist_entry_t dle_tofind;
|
|
|
|
dsl_deadlist_entry_t *dle;
|
|
|
|
avl_index_t where;
|
|
|
|
|
|
|
|
ASSERT(!dl->dl_oldfmt);
|
2017-02-09 18:19:12 +00:00
|
|
|
|
|
|
|
mutex_enter(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
dmu_buf_will_dirty(dl->dl_dbuf, tx);
|
|
|
|
dsl_deadlist_load_tree(dl);
|
|
|
|
|
|
|
|
dle_tofind.dle_mintxg = mintxg;
|
|
|
|
dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
|
|
|
|
if (dle == NULL)
|
|
|
|
dle = avl_nearest(&dl->dl_tree, where, AVL_AFTER);
|
|
|
|
while (dle) {
|
|
|
|
uint64_t used, comp, uncomp;
|
|
|
|
dsl_deadlist_entry_t *dle_next;
|
|
|
|
|
|
|
|
bpobj_enqueue_subobj(bpo, dle->dle_bpobj.bpo_object, tx);
|
|
|
|
|
|
|
|
VERIFY3U(0, ==, bpobj_space(&dle->dle_bpobj,
|
|
|
|
&used, &comp, &uncomp));
|
|
|
|
ASSERT3U(dl->dl_phys->dl_used, >=, used);
|
|
|
|
ASSERT3U(dl->dl_phys->dl_comp, >=, comp);
|
|
|
|
ASSERT3U(dl->dl_phys->dl_uncomp, >=, uncomp);
|
|
|
|
dl->dl_phys->dl_used -= used;
|
|
|
|
dl->dl_phys->dl_comp -= comp;
|
|
|
|
dl->dl_phys->dl_uncomp -= uncomp;
|
|
|
|
|
|
|
|
VERIFY3U(0, ==, zap_remove_int(dl->dl_os, dl->dl_object,
|
|
|
|
dle->dle_mintxg, tx));
|
|
|
|
|
|
|
|
dle_next = AVL_NEXT(&dl->dl_tree, dle);
|
|
|
|
avl_remove(&dl->dl_tree, dle);
|
|
|
|
bpobj_close(&dle->dle_bpobj);
|
|
|
|
kmem_free(dle, sizeof (*dle));
|
|
|
|
dle = dle_next;
|
|
|
|
}
|
2017-02-09 18:19:12 +00:00
|
|
|
mutex_exit(&dl->dl_lock);
|
2010-05-28 20:45:14 +00:00
|
|
|
}
|