]> sourceware.org Git - lvm2.git/log
lvm2.git
6 years agolvconvert: avoid superfluous interim raid type
Heinz Mauelshagen [Fri, 31 Aug 2018 17:03:52 +0000 (19:03 +0200)]
lvconvert: avoid superfluous interim raid type

When converting striped/raid0*/raid6_n_6 <-> raid4,
avoid superfluous interim raid5_n layout.

Related: rhbz1447809
(cherry picked from commit 22a13043683a5647e8cc4e3aead911e5269ffd2f)

6 years agoscripts: add After=rbdmap.service to {lvm2-activation-net,blk-availability}.service
Peter Rajnoha [Thu, 30 Aug 2018 10:35:58 +0000 (12:35 +0200)]
scripts: add After=rbdmap.service to {lvm2-activation-net,blk-availability}.service

We need to have Ceph RBD devices mapped first before use in a stack
where LVM is on top so make sure rbdmap.service is called before
generated lvm2-activation-net.service.

On shutdown, we need to stop blk-availability first before we stop the
rbdmap.service.

Resolves: rhbz1623479
(cherry picked from commit cb17ef221bdefea3625a22c19c6d8f5504441771)

Conflicts:
WHATS_NEW

6 years agotests: check activation of many thin-pool
Zdenek Kabelac [Wed, 5 Sep 2018 12:36:16 +0000 (14:36 +0200)]
tests: check activation of many thin-pool

Artifitical testing of monitoring of many thin-pools with low number
of resources in use (need only few pools to actually hit the race).

6 years agodmeventd: lvm2 plugin uses envvar registry
Zdenek Kabelac [Mon, 27 Aug 2018 08:18:26 +0000 (10:18 +0200)]
dmeventd: lvm2 plugin uses envvar registry

Thin plugin started to use configuble setting to allow to configure
usage of external scripts - however to read this value it needed to
execute internal command as dmeventd itself has no access to lvm.conf
and the API for dmeventd plugin has been kept stable.

The call of command itself was not normally 'a big issue' until users
started to use higher number of monitored LVs and execution of command
got stuck because other monitored resource already started to execute
some other lvm2 command and become blocked waiting on VG lock.

This scenario revealed necesity to somehow avoid calling lvm2 command
during resource registration - but this requires bigger changes - so
meanwhile this patch tries to minimize the possibility to hit this race
by obtaining any configurable setting just once - such patch is small
and covers majority of problem - yet better solution needs to be
introduced likely with bigger rework of dmeventd.

TODO: Avoid blocking registration of resource with execution of lvm2
commands since those can get stuck waiting on mutexes.

6 years agoUpdate WHATS_NEW
Marian Csontos [Tue, 28 Aug 2018 13:31:55 +0000 (15:31 +0200)]
Update WHATS_NEW

6 years agoWHATS_NEW: recent fixes
David Teigland [Mon, 27 Aug 2018 19:41:29 +0000 (14:41 -0500)]
WHATS_NEW: recent fixes

6 years agolvmetad: fix pvs for many devices
David Teigland [Mon, 27 Aug 2018 16:42:25 +0000 (11:42 -0500)]
lvmetad: fix pvs for many devices

When using lvmetad, 'pvs' still evaluates full filters
on all devices (lvmetad only provides info about PVs,
but pvs needs to report info about all devices, at
least sometimes.)

Because some filters read the devices, pvs still reads
every device, even with lvmetad (i.e. lvmetad is no help
for the pvs command.)  Because the device reads are not
being managed by the standard label scan layer, but only
happen incidentally through the filters, there is nothing
to control and limit the bcache content and the open file
descriptors for the devices.  When there are a lot of devs
on the system, the number of open fd's excedes the limit
and all opens begin failing.

The proper solution for this would be for pvs to really
use lvmetad and not scan devs, or for pvs to do a proper
label scan even when lvmetad is enabled.  To avoid any
major changes to the way this has worked, just work around
this problem by dropping bcache and closing the fd after
pvs evaluates the filter on each device.

6 years agolvmetad: improve scan for pvscan all
David Teigland [Mon, 27 Aug 2018 16:15:35 +0000 (11:15 -0500)]
lvmetad: improve scan for pvscan all

For 'pvscan --cache' avoid using dev_iter in the loop
after the label_scan by passing the necessary devs back
from the label_scan for the continued pvscan.
The dev_iter functions reapply the filters which will
trigger more io when we don't need or want it.  With
many devs, incidental opens from the filters (not controlled
by the label scan) can lead to too many open files.

6 years agospec: Disable python bindings on newer versions
Marian Csontos [Mon, 27 Aug 2018 14:17:11 +0000 (16:17 +0200)]
spec: Disable python bindings on newer versions

6 years agobcache: reduce MAX_IO to 256
David Teigland [Fri, 24 Aug 2018 19:46:51 +0000 (14:46 -0500)]
bcache: reduce MAX_IO to 256

This is the number of concurrent async io requests that
the scan layer will submit to the bcache layer.  There
will be an open fd for each of these, so it is best to
keep this well below the default limit for max open files
(1024), otherwise lvm may get EMFILE from open(2) when
there are around 1024 devices to scan on the system.

6 years agotest: add striped -> raid0 test script
Heinz Mauelshagen [Wed, 22 Aug 2018 15:11:03 +0000 (17:11 +0200)]
test: add striped -> raid0 test script

(cherry picked from commit 3c966e637fe1bec587ceb9ad13aa009db64b4f8e)

6 years agolvconvert: fix conversion attempts to linear
Heinz Mauelshagen [Wed, 22 Aug 2018 14:39:36 +0000 (16:39 +0200)]
lvconvert: fix conversion attempts to linear

"lvconvert --type linear RaidLV" on striped and raid4/5/6/10
have to provide the convenient interim layouts.  Fix involves
a cleanup to the convenience type function.

As a result of testing, add missing sync waits to
lvconvert-raid-reshape-linear_to_raid6-single-type.sh.

Resolves: rhbz1447809
(cherry picked from commit e83c4f07ca4a84808178d5d22cba655e5e370cd8)

Conflicts:
WHATS_NEW

6 years agospec: Add vdo plugin for dmeventd
Marian Csontos [Thu, 23 Aug 2018 09:27:17 +0000 (11:27 +0200)]
spec: Add vdo plugin for dmeventd

6 years agolvconvert: fix regression preventing direct striped conversion
Heinz Mauelshagen [Tue, 21 Aug 2018 15:12:33 +0000 (17:12 +0200)]
lvconvert: fix regression preventing direct striped conversion

Conversion to striped from raid0/raid0_meta is directly possible.

Fix a regression setting superfluous interim raid5_n conversion type
introduced by commit bd7cdd0b09ba123b064937fddde08daacbed7dab.

Add new test script lvconvert-raid0-striped.sh.

Resolves: rhbz1608067
(cherry picked from commit 4578411633a40c8c9068ff439ef3c33cbe78d25a)

Conflicts:
WHATS_NEW

6 years agotests: check policy mq can be used with format2
Zdenek Kabelac [Tue, 7 Aug 2018 13:20:31 +0000 (15:20 +0200)]
tests: check policy mq can be used with format2

6 years agotests: splitmirror for mirror type
Zdenek Kabelac [Tue, 7 Aug 2018 08:34:59 +0000 (10:34 +0200)]
tests: splitmirror for mirror type

6 years agomirror: fix splitmirrors for mirror type
Zdenek Kabelac [Tue, 7 Aug 2018 08:34:17 +0000 (10:34 +0200)]
mirror: fix splitmirrors for mirror type

With improved mirror activation code --splitmirror issue poppedup
since there was missing proper preload code and deactivation
for splitted mirror leg.

6 years agocache: drop metadata_format validation
Zdenek Kabelac [Tue, 7 Aug 2018 08:33:19 +0000 (10:33 +0200)]
cache: drop metadata_format validation

Allow to use any combination of cache metadata format for policy.

6 years agomirrors: fix read_only_volume_list
David Teigland [Thu, 2 Aug 2018 16:26:59 +0000 (11:26 -0500)]
mirrors: fix read_only_volume_list

If a mirror LV is listed in read_only_volume_list, it would
still be activated rw.  The activation would initially be
readonly, but the monitoring function would immediately
change it to rw.  This was a regression from commit

fade45b1d14c mirror: improve table update

The monitoring function needs to copy the read_only setting
into the new set of mirror activation options it uses.

6 years agoMerge branch '2018-06-01-stable' of git://sourceware.org/git/lvm2 into 2018-06-01...
Marian Csontos [Thu, 2 Aug 2018 06:08:51 +0000 (08:08 +0200)]
Merge branch '2018-06-01-stable' of git://sourceware.org/git/lvm2 into 2018-06-01-stable

* '2018-06-01-stable' of git://sourceware.org/git/lvm2:
  vgcreate: close exclusive fd after pvcreate

6 years agopost-release
Marian Csontos [Wed, 1 Aug 2018 14:55:04 +0000 (16:55 +0200)]
post-release

6 years agopre-release
Marian Csontos [Wed, 1 Aug 2018 15:30:40 +0000 (17:30 +0200)]
pre-release

6 years agovgcreate: close exclusive fd after pvcreate
David Teigland [Wed, 1 Aug 2018 15:26:28 +0000 (10:26 -0500)]
vgcreate: close exclusive fd after pvcreate

When vgcreate does an automatic pvcreate, it opens the
dev with O_EXCL to ensure no other subsystem is using
the device.  This exclusive fd remained in bcache and
prevented activation parts of lvm from using the dev.

This appeared with vgcreate of a sanlock VG because of
the unique combination where the dev is not yet a PV,
so pvcreate is needed, and the vgcreate also creates
and activates an internal LV for sanlock.

Fix this by closing the exclusive fd after it's used
by pvcreate so that it won't interfere with other
bits of lvm that may try to use the device.

6 years agopre-release
Marian Csontos [Wed, 1 Aug 2018 14:47:09 +0000 (16:47 +0200)]
pre-release

6 years agobuild: make generate
Marian Csontos [Tue, 31 Jul 2018 15:36:04 +0000 (17:36 +0200)]
build: make generate

6 years agoWHATS_NEW
Marian Csontos [Tue, 31 Jul 2018 15:02:47 +0000 (17:02 +0200)]
WHATS_NEW

6 years agodmeventd: rebase to stable branch
Zdenek Kabelac [Fri, 27 Jul 2018 09:17:26 +0000 (11:17 +0200)]
dmeventd: rebase to stable branch

Some minimal set of changes to make vdo plugin compilable in stable branch:

Use older headers.
Implement simple vdo status parser to only resolve use-percentage.

6 years agodmeventd: base vdo plugin
Zdenek Kabelac [Mon, 9 Jul 2018 09:41:26 +0000 (11:41 +0200)]
dmeventd: base vdo plugin

Introduce VDO plugin for monitoring VDO devices.

This plugin can be used also by other users, as plugin checks
for UUID prefix 'LVM-' and run  lvm actions only on those
devices.

Non LVM- device are only monitored and log warnings
when usage threshold reaches 80%.

6 years agoAdd BSD 2-Clause License
Marian Csontos [Fri, 27 Jul 2018 15:08:19 +0000 (17:08 +0200)]
Add BSD 2-Clause License

This is required by C++ test harness.

6 years agotest: Check flavour is used and exists
Marian Csontos [Thu, 26 Jul 2018 13:01:50 +0000 (15:01 +0200)]
test: Check flavour is used and exists

(cherry picked from commit 9cd05d1f1eef6ceb2dd0b18f792730f711334578)

6 years agolvconvert: reject conversions on raid1 split trackchanges LVs
Heinz Mauelshagen [Wed, 25 Jul 2018 23:47:39 +0000 (01:47 +0200)]
lvconvert: reject conversions on raid1 split trackchanges LVs

Prohibit, because the tracking can't continue and
further conversions may fail with bogus error messages.

Resolves: rhbz1579072
(cherry picked from commit a004bb07f14baeed158ac9ea571cf5a6882c436f)

Conflicts:
WHATS_NEW

6 years agolvconvert: reject conversions on raid1 split trackchanges SubLVs
Heinz Mauelshagen [Wed, 25 Jul 2018 23:00:11 +0000 (01:00 +0200)]
lvconvert: reject conversions on raid1 split trackchanges SubLVs

Prohibit conversions of raid1 split trackchanges SubLVs
because they will fail to get merged back into the RaidLV.

Resolves: rhbz1579438
(cherry picked from commit 8b0729af0fd357e66e95834e92913ea2a42d4e93)

Conflicts:
WHATS_NEW

6 years agodmsetup: fix error propagation in _display_info_cols()
Bryn M. Reeves [Thu, 28 Jun 2018 13:25:30 +0000 (14:25 +0100)]
dmsetup: fix error propagation in _display_info_cols()

Commit 3f35146 added a check on the value returned by the
_display_info_cols() function:

  1024         if (!_switches[COLS_ARG])
  1025                 _display_info_long(dmt, &info);
  1026         else
  1027                 r = _display_info_cols(dmt, &info);
  1028
  1029         return r;

This exposes a bug in the dmstats code in _display_info_cols:
the fact that a device has no regions is explicitly not an error
(and is documented as such in the code), but since the return
code is not changed before leaving the function it is now treated
as an error leading to:

  # dmstats list
  Command failed.

When no regions exist.

Set the return code to the correct value before returning.

(cherry picked from commit 29b9ccd261be025aaf75e58e5d2547e818ef22c3)

6 years agolvconvert: reject conversions of LVs under snapshot
Heinz Mauelshagen [Mon, 23 Jul 2018 17:20:18 +0000 (19:20 +0200)]
lvconvert: reject conversions of LVs under snapshot

Conversions of LVs under snapshot to thinpool or cachepool
correctly fail but leave them inactive and provide cryptic
error messages like 'Internal error: #LVs (10) != #visible
LVs (2) + #snapshots (1) + #internal LVs (5) in VG VG'.

Reject and provide better error message.

Resolves: rhbz1514146
(cherry picked from commit 2214dc12c34890c78b05456f58d0aa5d6dd08f4c)

6 years agolvconvert: improve text about splitmirrors
David Teigland [Mon, 23 Jul 2018 17:28:48 +0000 (12:28 -0500)]
lvconvert: improve text about splitmirrors

in messages and man page.

6 years agolvconvert: restrict command matching for no option variant
David Teigland [Mon, 23 Jul 2018 16:08:12 +0000 (11:08 -0500)]
lvconvert: restrict command matching for no option variant

The 'lvconvert LV' command def has caused multiple problems
for command matching because it matches the required options
of any lvconvert command.  Any lvconvert with incorrect options
ends up matching 'lvconvert LV', which then produces an error
about incorrect options being used for 'lvconvert LV'.  This
prevents suggestions from nearest-command partial command matches.

Add a special case for 'lvconvert LV' so that it won't be used
as a partial match for a command that has options specified.

6 years agopost-release
Marian Csontos [Thu, 19 Jul 2018 16:43:10 +0000 (18:43 +0200)]
post-release

6 years agopre-release v2_02_180
Marian Csontos [Thu, 19 Jul 2018 16:35:42 +0000 (18:35 +0200)]
pre-release

6 years agoWHATS_NEW
Marian Csontos [Thu, 19 Jul 2018 16:33:59 +0000 (18:33 +0200)]
WHATS_NEW

6 years agoallocation: add check for passing log allocation
Zdenek Kabelac [Sun, 8 Jul 2018 22:54:16 +0000 (00:54 +0200)]
allocation: add check for passing log allocation

Updates previous commit.

6 years agoWHATS_NEW: update
Zdenek Kabelac [Sun, 8 Jul 2018 22:36:11 +0000 (00:36 +0200)]
WHATS_NEW: update

6 years agodev_io: no discard in testmode
Zdenek Kabelac [Sun, 8 Jul 2018 19:11:01 +0000 (21:11 +0200)]
dev_io: no discard in testmode

When lvm2 command is executed in test mode, discard ioctl is skipped.
This may cause even data-loose in case, issuing discard for released
areas was enabled and user 'tested'  lvreduce.

6 years agoallocator: fix thin-pool allocation
Zdenek Kabelac [Sun, 8 Jul 2018 16:34:38 +0000 (18:34 +0200)]
allocator: fix thin-pool allocation

When allocating thin-pool with more then 1 device - try to
allocate 'metadataLV' with reuse of log-type allocation for mirror LV.
It should be naturally place on other device then 'dataLV'.

However due to somewhat hard to follow allocation logic code,
it's been rejected allocation in cases where there was not
enough space for data or metadata on single PV, thus to successed,
usage of segments was mandatory.

While user may use:

allocation/thin_pool_metadata_require_separate_pvs=1

to enforce separe meta and data LV - on default settings, this is not
enable thus segment allocation is meant to work.

NOTE:

As already said - the original intention of this whole  'if()' is unclear,
so try to split this test into multiple more simple tests that are more readable.

TODO: more validation.

6 years agomemlock: extend exception list
Zdenek Kabelac [Sat, 30 Jun 2018 09:05:14 +0000 (11:05 +0200)]
memlock: extend exception list

Amound of linked libraries grows.
Most of them we don't need to lock in, since we are not using
them in locked section, so skip locking them in memory.

6 years agotests: update with --yes
Zdenek Kabelac [Sun, 24 Jun 2018 18:49:59 +0000 (20:49 +0200)]
tests: update with --yes

vgcfgrestore needs to confirm restore while LVs from VG are present.

6 years agovcfgrestore: add prompt with active volumes
Zdenek Kabelac [Fri, 22 Jun 2018 16:45:48 +0000 (18:45 +0200)]
vcfgrestore: add prompt with active volumes

Add check for active device with names matching restored VG.
When such devices are present in dm table, prompt user, if he
wish to continue.

6 years agoRevert "man: fix lvreduce example"
David Teigland [Wed, 27 Jun 2018 14:19:01 +0000 (09:19 -0500)]
Revert "man: fix lvreduce example"

 -l -3 is correct, meaning reduce by 3.

This reverts commit d5bcc56eefde8349ffee93d0411c6f9fd71ceb84.

6 years agoman: fix lvreduce example
David Teigland [Wed, 27 Jun 2018 13:58:22 +0000 (08:58 -0500)]
man: fix lvreduce example

6 years agodevs: use bcache fd for read ahead ioctl
David Teigland [Tue, 26 Jun 2018 17:05:39 +0000 (12:05 -0500)]
devs: use bcache fd for read ahead ioctl

to avoid an unnecessary open of the device in
most cases.

6 years agoscan: reopen RDWR during rescan
David Teigland [Tue, 26 Jun 2018 16:58:11 +0000 (11:58 -0500)]
scan: reopen RDWR during rescan

Commit a30e6222799:
  "scan: work around udev problems by avoiding open RDWR"

had us reopen a device RDWR in the write function.  Since
we know earlier that the command intends to write to devices
in the VG, we can reopen the VG's devices RDWR during the
rescan instead of waiting until the writes to happen.

6 years agobcache.c add missing {
David Teigland [Tue, 26 Jun 2018 17:15:07 +0000 (12:15 -0500)]
bcache.c add missing {

6 years agobcache: Fix null pointer dereferencing
Marian Csontos [Tue, 26 Jun 2018 15:04:18 +0000 (17:04 +0200)]
bcache: Fix null pointer dereferencing

(cherry picked from commit a14f21bf1ddbb93b13a917cad8720fed6c94d6de)

Conflicts:
lib/device/bcache.c

6 years agoWHATS_NEW
Heinz Mauelshagen [Tue, 5 Jun 2018 14:24:19 +0000 (16:24 +0200)]
WHATS_NEW

(cherry picked from commit 11384637fb57c0fdefe5940cf27285ebdcc1dbc6)

Conflicts:
WHATS_NEW

6 years agotest: add convcenience conversion tests linear <-> striped
Heinz Mauelshagen [Wed, 30 May 2018 10:50:34 +0000 (12:50 +0200)]
test: add convcenience conversion tests linear <-> striped

Add tests for linear <-> striped|raid* conversions.

Add region_size config to reshape tests to avoid test
failures in case of it being defined unexpectedly in lvm.conf.

Related: rhbz1439925
Related: rhbz1447809
(cherry picked from commit 3810fd8d0d9cfcc519355f995d9cc70dca2811fd)

6 years agolvconvert: support linear <-> striped convenience conversions
Heinz Mauelshagen [Wed, 30 May 2018 10:49:15 +0000 (12:49 +0200)]
lvconvert: support linear <-> striped convenience conversions

"lvconvert --type {linear|striped|raid*} ..." on a striped/linear
LV provides convenience interim type to convert to the requested
final layout similar to the given raid* <-> raid* conveninece types.

Whilst on it, add missing raid5_n convenince type from raid5* to raid10.

Resolves: rhbz1439925
Resolves: rhbz1447809
Resolves: rhbz1573255
(cherry picked from commit bd7cdd0b09ba123b064937fddde08daacbed7dab)

6 years agosegtype: add linear
Heinz Mauelshagen [Wed, 30 May 2018 10:44:24 +0000 (12:44 +0200)]
segtype: add linear

Add linear segtype addressing FIXME in preparation
for linear <-> striped convenience conversion support

(cherry picked from commit de66704253653305e4513637735a078af00912b7)

6 years agofilter: make pointers distinguishable
Marian Csontos [Fri, 22 Jun 2018 13:07:07 +0000 (15:07 +0200)]
filter: make pointers distinguishable

This ammends commit: 4afb5971b93a with suggestions to improve debugging
from Nir Soffer.

6 years agosnapshot: improve checking of merging snapshot
Zdenek Kabelac [Mon, 11 Jun 2018 20:19:20 +0000 (22:19 +0200)]
snapshot: improve checking of merging snapshot

Add runtime detection for 'lvs -o+seg_monitor' and 'vgchange --monitor'.
This fix should avoid unnecessary timeout on systemd shutdown.

6 years agofsadm: missing -l description
Zdenek Kabelac [Fri, 22 Jun 2018 13:00:52 +0000 (15:00 +0200)]
fsadm: missing -l description

6 years agocache: cleaner policy also uses fmt2
Zdenek Kabelac [Fri, 8 Jun 2018 10:37:07 +0000 (12:37 +0200)]
cache:  cleaner policy also uses fmt2

Format 2 is also with cleaner policy.

6 years agopvmove: improve lvs
Zdenek Kabelac [Mon, 4 Jun 2018 09:05:28 +0000 (11:05 +0200)]
pvmove: improve lvs

When pvmoving LV - the target for LV is a mirror so the validation
that checked the type is matching was incorrect.

While we need a more generic enhancment of LVS output for pvmoved LVs,
for now at least stop showing internal errors and  'X' symbols in attrs.

6 years agopvresize: add missing return
Zdenek Kabelac [Tue, 5 Jun 2018 19:21:26 +0000 (21:21 +0200)]
pvresize: add missing return

Log error path missed return 0.
Also fix some unneded bactraces (since log_error already shows
position).

6 years agopvresize: update message
Zdenek Kabelac [Tue, 5 Jun 2018 21:09:55 +0000 (23:09 +0200)]
pvresize: update message

There is always at least PV header update even if the size
of PV remains same (so it's not really resized).
Try to make it a slightly less confusing.

6 years agosystemd: add conficting sockets
Zdenek Kabelac [Tue, 12 Jun 2018 14:27:42 +0000 (16:27 +0200)]
systemd: add conficting sockets

Since we are using "DefaultDependencies=no" we do not get automatic STOP
job on socket connection - so automatically refuse connection on
shutdown by adding this Conflict definition to socket Unit.

6 years agopvscan: code reshape
Zdenek Kabelac [Thu, 14 Jun 2018 19:07:15 +0000 (21:07 +0200)]
pvscan: code reshape

6 years agovgchange: fix error code in error path
Zdenek Kabelac [Wed, 13 Jun 2018 13:57:51 +0000 (15:57 +0200)]
vgchange: fix error code in error path

This rather hard to hit error path used wrong return value to signal
real error.

6 years agofilter: use pointers to real addresses
David Teigland [Thu, 21 Jun 2018 15:52:35 +0000 (10:52 -0500)]
filter: use pointers to real addresses

instead of casting values 1 and 2 to pointers
which gcc optimization can have problems with.

6 years agocoverity warnings about null info in lvmcache.c
David Teigland [Thu, 21 Jun 2018 14:21:25 +0000 (09:21 -0500)]
coverity warnings about null info in lvmcache.c

6 years agolvmlockd: fix another missing lock_type null check
David Teigland [Thu, 21 Jun 2018 14:00:23 +0000 (09:00 -0500)]
lvmlockd: fix another missing lock_type null check

Same as 347c807f8.

6 years agobcache: fix memory leaks
Marian Csontos [Thu, 21 Jun 2018 08:22:35 +0000 (10:22 +0200)]
bcache: fix memory leaks

6 years agolibdm: fix buffer overflow
Marian Csontos [Thu, 21 Jun 2018 08:20:09 +0000 (10:20 +0200)]
libdm: fix buffer overflow

6 years agoWHATS_NEW: recent changes
David Teigland [Wed, 20 Jun 2018 19:32:29 +0000 (14:32 -0500)]
WHATS_NEW: recent changes

6 years agoscan: work around udev problems by avoiding open RDWR
David Teigland [Wed, 20 Jun 2018 16:32:45 +0000 (11:32 -0500)]
scan: work around udev problems by avoiding open RDWR

udev creates a train wreck of events if we open devices
with RDWR.  Until we can fix/disable/scrap udev, work around
this by opening RDONLY and then closing/reopening RDWR when
a write is needed.  This invalidates the bcache blocks for
the device before writing so it can trigger unnecessary
rereading.

6 years agoclvmd: fix leak of saved_vg struct
David Teigland [Mon, 18 Jun 2018 19:10:48 +0000 (14:10 -0500)]
clvmd: fix leak of saved_vg struct

Commit c016b573ee32f "clvmd: separate saved_vg from vginfo"
created a separate hash table for the saved_vg structs.
The vg's referenced by the saved_vg struct were all being
freed properly, but the svg wrapper struct itself was not
being freed.

6 years agobcache: remove extraneous error message
David Teigland [Mon, 18 Jun 2018 16:59:57 +0000 (11:59 -0500)]
bcache: remove extraneous error message

an error from io_submit is already recognized by
the caller like errors during completion.

6 years agopost-release
Marian Csontos [Mon, 18 Jun 2018 05:30:09 +0000 (07:30 +0200)]
post-release

6 years agopre-release v2_02_179
Marian Csontos [Mon, 18 Jun 2018 05:21:51 +0000 (07:21 +0200)]
pre-release

6 years agoman: update lvmsystemid wording
David Teigland [Thu, 14 Jun 2018 17:34:00 +0000 (12:34 -0500)]
man: update lvmsystemid wording

to refer to "shared VG" instead of "lockd VG".

6 years agoman: updates to lvmlockd
David Teigland [Thu, 14 Jun 2018 17:30:45 +0000 (12:30 -0500)]
man: updates to lvmlockd

The terminology has migrated toward using "shared VG"
rather than "lockd VG".

Also improve the wording in a number of places.

6 years agolvmlockd: update method for changing clustered VG
David Teigland [Wed, 13 Jun 2018 20:54:39 +0000 (15:54 -0500)]
lvmlockd: update method for changing clustered VG

The previous method for forcibly changing a clustered VG to
a local VG involved using -cn and --config locking_type=0.
Add an alternative that is consistent with other forced
lock type changes:
vgchange --locktype none --lockopt force.

6 years agoWHATS_NEW for recent changes
David Teigland [Wed, 13 Jun 2018 20:42:15 +0000 (15:42 -0500)]
WHATS_NEW for recent changes

6 years agoMerge branch 'dev-mcsontos-release' into stable
Marian Csontos [Wed, 13 Jun 2018 17:12:35 +0000 (19:12 +0200)]
Merge branch 'dev-mcsontos-release' into stable

6 years agopost-release
Marian Csontos [Thu, 24 May 2018 13:23:08 +0000 (15:23 +0200)]
post-release

6 years agopre-release v2_02_178
Marian Csontos [Wed, 13 Jun 2018 12:13:35 +0000 (14:13 +0200)]
pre-release

6 years agolvmlockd: skip repair lock upgrade for non shared vgs
David Teigland [Tue, 12 Jun 2018 14:25:51 +0000 (09:25 -0500)]
lvmlockd: skip repair lock upgrade for non shared vgs

Only attempt lvmlockd lock upgrade for shared VGs.

6 years agoman lvmlockd: remove unnecessary reference to lvmetad
David Teigland [Thu, 7 Jun 2018 18:42:11 +0000 (13:42 -0500)]
man lvmlockd: remove unnecessary reference to lvmetad

it's optional to use it with lvmlockd

6 years agolvmlockd: fix missing lock_type null check
David Teigland [Wed, 6 Jun 2018 18:56:02 +0000 (13:56 -0500)]
lvmlockd: fix missing lock_type null check

Missed checking if vg->lock_type is NULL in commit db8d3bdfa:
  lvmlockd: enable mirror split and merge with dlm lock_type

6 years agodevices: clean up io error messages
David Teigland [Wed, 6 Jun 2018 15:05:08 +0000 (10:05 -0500)]
devices: clean up io error messages

Remove the io error message from bcache.c since it is not
very useful without the device path.

Make the io error messages from dev_read_bytes/dev_write_bytes
more user friendly.

6 years agoman lvmlockd: update list of limitations
David Teigland [Thu, 31 May 2018 21:38:39 +0000 (16:38 -0500)]
man lvmlockd: update list of limitations

6 years agotests: enable non-working tests with lvmlockd
David Teigland [Thu, 31 May 2018 21:18:53 +0000 (16:18 -0500)]
tests: enable non-working tests with lvmlockd

Those that are failing for reasons other than lvmlockd
restrictions.

6 years agotests: enable more working tests with lvmlockd
David Teigland [Thu, 31 May 2018 20:55:18 +0000 (15:55 -0500)]
tests: enable more working tests with lvmlockd

6 years agolvmlockd: improve error message for existing lockspace
David Teigland [Thu, 31 May 2018 20:52:23 +0000 (15:52 -0500)]
lvmlockd: improve error message for existing lockspace

When a VG/lockspace already exists with the same name
don't just print the error number.

6 years agotests: enable lvcreate-pvtags with lvmlockd
David Teigland [Thu, 31 May 2018 20:37:25 +0000 (15:37 -0500)]
tests: enable lvcreate-pvtags with lvmlockd

6 years agolvmlockd: fix reverting new lv in error path
David Teigland [Thu, 31 May 2018 20:35:48 +0000 (15:35 -0500)]
lvmlockd: fix reverting new lv in error path

The wrong name was being used to free the LV lock
in lvmlockd in the error exit path.

6 years agotests: enable metadata-balance with lvmlockd
David Teigland [Thu, 31 May 2018 20:12:34 +0000 (15:12 -0500)]
tests: enable metadata-balance with lvmlockd

6 years agotests: enable lvmlockd with tests using lvcreate -H -L LV
David Teigland [Thu, 31 May 2018 19:49:16 +0000 (14:49 -0500)]
tests: enable lvmlockd with tests using lvcreate -H -L LV

6 years agolvmlockd: enable lvcreate -H -L LV
David Teigland [Thu, 31 May 2018 19:20:11 +0000 (14:20 -0500)]
lvmlockd: enable lvcreate -H -L LV

Allow this command in a shared VG which had previously been
disallowed.

6 years agovgs: add report field for shared
David Teigland [Thu, 31 May 2018 15:23:03 +0000 (10:23 -0500)]
vgs: add report field for shared

equivalent to a non-empty -o locktype.

6 years agotests: change lvcreate syntax to allow lvmlockd
David Teigland [Wed, 30 May 2018 21:40:03 +0000 (16:40 -0500)]
tests: change lvcreate syntax to allow lvmlockd

Using the less ambiguous lvcreate syntax for creating a
cache LV allows more tests to run with lvmlockd.

6 years agotests: enable lvcreate cache tests with lvmlockd
David Teigland [Wed, 30 May 2018 20:56:08 +0000 (15:56 -0500)]
tests: enable lvcreate cache tests with lvmlockd

Tests that want to use lvcreate to create a new
origin LV and then combine it with an existing
cache pool to create a cache LV.

This page took 0.065302 seconds and 5 git commands to generate.