IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
With the ability to stack many operations in one udev transaction -
in same cases we are adding and removing same device at the same time
(i.e. deactivate followed by activate).
This leads to a problem of checking stacked operations:
i.e. remove /dev/node1 followed by create /dev/node1
If the node creation is handled with udev - there is a problem as
stacked operation gives warning about existing node1 and will try to
remove it - while next operation needs to recreate it.
Current code removes all previous stacked operation if the fs op is
FS_DEL - patch adds similar behavior for FS_ADD - it will try to
remove any 'delete' operation if udev is in use.
For FS_RENAME operation it seems to be more complex. But as we
are always stacking FS_READ_AHEAD after FS_ADD operation -
should be safe to remove all previous operation on the node
when udev is running.
Code does same checking for stacking libdm and liblvm operations.
As a very simple optimization counters were added for each stacked ops
type to avoid unneeded list scans if some operation does not exists in
the list.
Enable skipping of fs_unlock() (udev sync) if only DEL operations are staked.
as we do not use lv_info for already deleted nodes.
Instead of implicitly syncing udev operation in clustered and
file locking code - call synchronization directly in lock_vol() when
the operation unlocks VG
The problem is missing implicit fs_unlock() in the no_locking code.
This is used with --sysinit on read-only filesystem locking dir.
In this case vgchange -ay could exit before all udev nodes are properly
synchronised and may cause problems with accessing such node right after
vgchange --sysinint command is finished.
Add test case for vgchange --sysinit.
return lockspace reference (even if lockspace already exists)
and thus increases DLM lockspace count. It means that after
clvmd restart the lockspace is still in use.
(The only way to clean environment to enable clean cluster
shutdown is call "dlm_tool leave clvmd" several times.)
Because only one clvmd can run in time, we can use simpler logic,
try to open lockspace with dlm_open_lockspace() and only if it fails
try to create new one. This way the lockspace reference doesn not
increase.
Very easily reproducible with "clvmd -S" command.
Patch also fixes return code when clvmd_restart fails and fixes
double free if debug option was specified during restart.
Fixes https://bugzilla.redhat.com/show_bug.cgi?id=612862
Improve condition within lock_vol so we are not calling extra unlock
if the volume just has been deactivated.
Patch uses lck_type and replaces negative 'and' condition to more
readable 'or' condition.
Few missing strace traces added.
As sync_local_dev_names() cannot be called within activation context,
add new parametr which allows to select if the sync call is needed
before executing new command.
Stop calling fs_unlock() from lv_de/activate().
Start using internal lvm fs cookie for dm_tree.
Stop directly calling dm_udev_wait() and
dm_tree_set/get_cookie() from activate code -
it's now called through fs_unlock() function.
Add lvm_do_fs_unlock()
Call fs_unlock() when unlocking vg where implicit unlock solves the
problem also for cluster - thus no extra command for clustering
environment is required - only lvm_do_fs_unlock() function is added
to call lvm's fs_unlock() while holding lvm_lock mutex in clvmd.
Add fs_unlock() also to set_lv() so the command waits until devices
are ready for regular open (i.e. wiping its begining).
Move fs_unlock() prototype to activation.h to keep fs.h private
in lib/activate dir and not expose other functions from this header.
Change function import_vg_from_buffer() to import_vg_from_config_tree().
Instead of creating config tree inside the function allow config tree to
be passed as parameter - usable later for caching.
If some allocation for peristent filter fails its memory reference
was lost, fix it by calling filter's destructor.
Fix log_error messages for failing allocation.
a mirror image (or removing a log while adding a mirror). Advise the
user to use two separate commands instead.
This issue become especially problematic when PVs are specified, as they
tend to mean different things when adding vs removing. In a command that
mixes adding and removing, it is impossible to decern exactly what the
user wants.
This change prevents bug 603912.
Add checks for clonning allocation a fail-out when something is
not allocated correctly.
Also move var declaration to the begining of the function
and fix log_error messages.
To have better control were the config tree could be modified use more
const pointers and very carefully downcast them back to non-const
(for config tree merge).
As ternary operator has lower priority then add operation, this check
was not doing what seemed to be expected.
So enclose the test in braces and check for NULL in *buf.
We need to be sure that /var/run and /var/lock is always there.
(E.g. these two directories could be using tmpfs which then loose
all the content after reboot.)
Detect existence of new SELinux selabel interface during configure.
Use new dm_prepare_selinux_context instead of dm_set_selinux_context.
We should set the SELinux context before the actual file system object creation.
The new dm_prepare_selinux_context function sets this using the selabel_lookup
fn in conjuction with the setfscreatecon fn. If selinux/label.h interface
(that should be a part of the selinux library) is not found during configure,
we fallback to the original matchpathcon function instead.
Set cmd->independent_metadata_areas if metadata/dirs or disk_areas in use.
- Identify and record this state.
Don't skip full scan when independent mdas are present even if memlock is set.
- Clusters and OOM aren't supported, so no problem doing the proper scans.
Avoid revalidating the label cache immediately after scanning.
- A simple optimisation.
Support scanning for a single VG in independent mdas.
- Not used by the fix but I left it in anyway as later patches might use it.
This reset of vgmem pointer causes access of already released memory.
(_vg_make_handle allocates vg from vgmem pool itself - which is a bit tricky)
Interestingly this memory fault was missed by our test suite.
Add log_error message for lv_info failure and exit from futher
processing.
Replace 'leg' occurence in debug message with 'image' which
is used in other messages.
Add test for NULL from dm_poll_create.
Reorder dm_pool_destroy() before file close and add label out:.
Avoid leaking file descriptor if the allocation fails.
Nicely hidden memory leak in outf macro error path.
This macro is using out_text() and does automagical return_0.
That would leak tag_buffer allocated memory.
As there was same code for tags output - create _out_tags() function.
Set vg to NULL after releasing it as the following memlock() test may
lead to goto for the second call of vg_release() with the already
released vg pointer.
Fix for the last commit as $MOUNTED is not only used as bool flag,
but also store mounted location for remount - so parsing output
from mount differently then from /proc/mounts.
Prefix calls of 'tunefs' tools with LANG=C to be sure we always do get
some nonlocalized strings.
Avoid using forced 'resize2fs' for cleanly unmounted filesystems and
run regular fsck -f for this case as required by resize2fs.
'fsadm check' uses date difference for extX filesystems between
the last mount and last check of 'fsck -f' execution and if the mount
was later run 'fsck' with -f so resize2fs is happy and user does not
need to pass '-f' flag.
Updated patch from Florian Haas from Linux-HA project.
User needs to 'configure --enable-ocf' to get file installed
by 'make install' target by default.
User can also use 'make install_ocf' to get only ocf files installed.
With disabled (default) ocf support - no ocf files are installed.
FIXME: ocf installation path needs to be kept in sync with pacemaker.
find better way and possible also better location.
Patch updates exec_cmd() and adds 3rd parameter with pointer for
status value, so caller might examine returned status code.
If the passed pointer is NULL, behavior is unmodified.
Patch allows to confinue with lvresize if the failure from fsadm check is
caused by mounted filesystem as many of filesystem resize tools do support
online filesystem resize. (originally user had to use flag '-n' to bypass
this filesystem check)
Return status code 3 for fsadm check of mounted filesystem - used later with
lvresize update patch to better support online filesystem resize.
Also makes a more consistent user interruption and returns status code 2
in this case.
Simultaneous -a and --refresh is not valid.
poll+monitor are valid together with or without -ay* (but not with -an*)
No longer print polling results summary if no LVs in the VG were polled.
We cast (char*) to (uint32_t*) that changes alignment requierements.
For our case the code has been correct as alloca() returns properly
aligned buffer, however this patch make it cleaner and more readable
and avoids warning generation.
A merged snapshot's DM device is made to use the "error" target as part
of lvm's transaction to merge a snapshot. This snapshot merge use-case
aside, any device using the error target shouldn't be scanned.
Problem:
When both legs of a mirrored log fail, neither the log nor the parent
mirror can proceed. The repair code must be careful to replace the
log with an error target before operating on the parent - otherwise,
the parent can get stuck trying to suspend because it can't push through
any writes. The steps to replace the log device with an error target
were incomplete and resulted in the replacement not happening at all!
The code originally had all the necessary logic to complete the
replacement task, but was pulled out in a effort to clean-up that
section of code, while fixing another bug:
<offending commit msg>
In addition, I added following three changes.
- Removed tmp_orphan_lvs handling procedure
It seems that _delete_lv() can handle detached_log_lv properly
without adding mirror legs in mirrored log to tmp_orphan_lvs.
Therefore, I removed the procedure.
- Removed vg_write()/vg_commit()
Metadata is saved by vg_write()/vg_commit() just after detached_log_lv
is handled. Therefore, I removed vg_write()/vg_commit().
</offending commit msg>
http://sources.redhat.com/cgi-bin/cvsweb.cgi/LVM2/lib/metadata/mirror.c?cvsroot=lvm2&f=h#rev1.130
I've reverted the "clean-up" changes associated with that fix, but not what
that commit was actually fixing.
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Petr Rockai <prockai@redhat.com>
page.
Add ->target_name to segtype_handler to allow a more specific target
name to be returned based on the state of the segment.
Result of trying to merge a snapshot using a kernel that doesn't have
the snapshot-merge target:
Before:
# lvconvert --merge vg/snap
Can't expand LV lv: snapshot target support missing from kernel?
Failed to suspend origin lv
After:
# lvconvert --merge vg/snap
Can't process LV lv: snapshot-merge target support missing from kernel?
Failed to suspend origin lv
Unable to merge LV "snap" into it's origin.
Try to distinguish between the case of using interactive shell and non
interactive running - different combinations of '-y' and '-p' option
needs to be used for fsck.
Update the way how fsadm detects mounted filesystem.
With udev /dev/dm-XXX paths are now returned - but mount or /proc/mounts
prints names in form of /dev/mapper/vg-lv - so the match was not found.
Fixex RHBZ #638050.
Current solution uses same trick as mount and detects vg-lv name through
/sys where available - this should be reasonable safe.
Instead of calling mount without parameter to get actual mount table,
switch to use /proc/mounts directly.
Fix missing 'dry' execution of lvresize - fixing problem where resize
command were 'dry-run' executed - but lvresize has been executed for real.
Also adapt code slightly to support better recursive execution of fsadm
through lvresize call.
Under certain conditions it was possible to break (^C) fsadm before actually
resizing filesystem, but lvresize which executed fsadm will think resize
was succesful and shrinks partitions with unresized filesystem on it.
Fix by returning error (1) for this case - this stops lvresize from futher
proceding in resize operation.
In other LVM memory structures such as volume_group, the field
used to store flags is called "status", and on-disk fields are called
'flags', so rename the one inside metadata_area to be consistent.
Not only is it more consistent with existing code but is cleaner
to say "the status of this mda is ignored".
Background for this patch - prajnoha pinged me on IRC this morning
about a fix he was working on related to metadataignore when
metadata/dirs was set. I was reviewing my patches from this year
and realized the 'flags' field was probably not the best choice
when I originally did the metadataignore patches.
Current lvm1 allocation code seems to not properly
map segments on missing PVs.
For now disable this functionality.
(It never worked and previous commit just introduced segfault here.)
So the partial mode in lvm1 can only process missing PVs
with no LV segments only.
Also do not use random PV UUID for missing part but use fixed
string derived from VG UUID (to not confuse clvmd tests).
Read complete content of /proc/self/maps into one buffer without
realocation in the middle of reading and before doing any m/unlock
operation with these lines - as some of them gets change.
With previous implementation we've read some mappings twice ([stack])
If some lvm1 device is missing, lvm fails on all operations
# vgcfgbackup -f bck -P vg_test
Partial mode. Incomplete volume groups will be activated read-only.
3 PV(s) found for VG vg_test: expected 4
PV segment VG free_count mismatch: 152599 != 228909
PV segment VG extent_count mismatch: 152600 != 228910
Internal error: PV segments corrupted in vg_test.
Volume group "vg_test" not found
Allow loading of lvm1 partial VG by allocating "new" missing PV,
which covers lost space. Also this fake mising PV inform code
that it is partial VG.
https://bugzilla.redhat.com/show_bug.cgi?id=501390
Revert to old glibc behaviour for vsnprintf used in emit_to_buffer fn.
Otherwise, the check that follows would be wrong for new glibc versions.
This caused the rh bug #633033 to be undetected and pass throught the check,
corrupting the metadata!
In certain configurations, we're not under a VG rw lock while trying to write
a new archive file with VG metadata. A common example is using "vgs" while
having the content of backup and archive directories empty. The code scans the
content of these directories and tries to determine the final index that should
be used in archive name. Since we're not under a lock, we can get into a race
while choosing the index which could end up showing errors about not being able
to rename to final archive name. Let's add random number suffix to these archive
file names so we can avoid the race.
For example, when using '--config "backup { ... }"' line, the values from
lvm.conf (or default values) should be overridden. This patch adds
reinitialisation of archive and backup handling on toolcontext refresh
which makes these settings to be applied.
can be opprobriously slow if created with '--nosync'.
One of the ways cluster mirrors coordinate I/O and recovery
amoung the different machines is by the use of the log
function 'is_remote_recovering()' which lets nodes know if
a region they wish to perform a write on is currently being
recovered on another node. If the region is being recovered,
the I/O is delayed.
The 'is_remote_recovering' routine has been optimized to
avoid the deluge of requests that would be issued to the
userspace log server by maintaining a marker of how far
the recovery has gotten. It can then immediately return
'not recovering' if the region being inquired about is
less than this mark. Additionally, if the region of
concern is greater than the mark, the function will
limit the number of transmissions to userspace by assuming
the region /is/ being recovered when skipping the
transmission. This limits the amount of processing
and updates the mark in 1/4 sec time steps.
This patch fixes a problem where 'the mark' is not being
updated because of faulty logic in the userspace log
daemon. When '--nosync' is used to create a cluster
mirror, the userspace log daemon never has a chance
to update the mark in the normal way. The fix is to set
the mark to "complete" if the mirror was created with
the --nosync flag.
may never complete.
If you convert from a linear to a mirror and then convert that
mirror back to linear /while/ the previous (up)convert is
taking place, the mirror polling process will never complete.
This is because the function that polls the mirror for
completion doesn't check if it is still polling a mirror and
the copy_percent that it gets back from the linear device is
certainly never 100%.
The fix is simply to check if the daemon is still looking at
a mirror device - if not, return PROGRESS_CHECK_FAILED.
The user sees the following output from the first (up)convert
if someone else sneaks in and does a down-convert shortly
after their convert:
[root@bp-01 ~]# lvconvert -m1 vg/lv
vg/lv: Converted: 43.4%
ABORTING: Mirror percentage check failed.
to block when a mirror under a snapshot suffers a failure.
The problem has to do with label scanning. When a mirror suffers
a failure, the kernel blocks I/O to prevent corruption. When
LVM attempts to repair the mirror, it scans the devices on the
system for LVM labels. While mirrors are skipped during this
scanning process, snapshot-origins are not. When the origin is
scanned, it kicks up I/O to the mirror (which is blocked)
underneath - causing the label scan (an thus the repair operation)
to hang.
This patch simply bypasses snapshot-origin devices when doing
labels scans (while ignore_suspended_devices() is set). This
fixes the issue.
introduced in commit b16b4d92a7
"Improve various log messages."
fixes a lot of
../include/metadata.h:148: warning: type qualifiers ignored on function return type
set appropriate Required-Start and Required-Stop at configure time.
Reorder the checks for user selected cluster managers to match auto
detected ones, to be consistent in the output.
Add special case for qdiskd that´s started after cman/lock_gulmd for
RHEL-4/RHEL-5.
If pvmove crashed and metadata contains pvmove LV
but without miorrored segments, pvmove --abort
will not repair the situation (and finish wth success!).
Fix it by allowing metadata update if aborting
(thus removing pvmove LV) even if no moved LVs detected.
(Tested on real metadata provided by an lvm user:-)
Add "devices/default_data_alignment" to lvm.conf to control the internal
default that LVM2 uses: 0==64k, 1==1MB, 2==2MB, etc.
If --dataalignment (or lvm.conf's "devices/data_alignment") is specified
then it is always used to align the start of the data area. This means
the md_chunk_alignment and data_alignment_detection are disabled if set.
(Same now applies to pvcreate --dataalignmentoffset, the specified value
will be used instead of the result from data_alignment_offset_detection)
set_pe_align() still looks to use the determined default alignment
(based on lvm.conf's default_data_alignment) if the default is a
multiple of the MD or topology detected values.
In all top vg read functions only LCK_VG_READ/WRITE can be used.
All other vg lock definitions are low-level backend machinery.
Moreover, LCK_WRITE cannot be tested through bitmask.
This patch fixes these mistakes.
For _recover_vg() we do not need lock_flags, it can be only
two of above and we always upgrading to LCK_VG_WRITE lock there.
(N.B. that code is racy)
There is no functional change in code (despite wrong masking
it produces correct bits:-)
One shiny day we should use libblkid here. But now using LUKS is
very common together with LVM and pvcreate destroys LUKS completely.
So for user's convenience, try to detect LUKS signature and allow abort.
This is not only undocumented but is is also in violation with --help
documentation.
Using --yes without --force is useful in pvcreate when it detects
old signature.
pvcreate detects MD and swap signature.
The logic hidden there is not only documented but it is also
user unfriendly. Who invented this logic should run pvcreate
on its own critical MD device to see why;-)
This patch
- creates one function instead of duplication code
- asks if user want to overwrite signature
- allows aborting (!)
(Please note that writing LVM signatute without wiping old
is wrong, it confuses blkid, MD will not work anyway and
swap and LUKS is broken too.)
The lvm repair issues I believe are the superficial symptoms of this
bug - there are worse issues that are not as clearly seen. From my
inline comments:
* If the mirror was successfully recovered, we want to always
* force every machine to write to all devices - otherwise,
* corruption will occur. Here's how:
* Node1 suffers a failure and marks a region out-of-sync
* Node2 attempts a write, gets by is_remote_recovering,
* and queries the sync status of the region - finding
* it out-of-sync.
* Node2 thinks the write should be a nosync write, but it
* hasn't suffered the drive failure that Node1 has yet.
* It then issues a generic_make_request directly to
* the primary image only - which is exactly the device
* that has suffered the failure.
* Node2 suffers a lost write - which completely bypasses the
* mirror layer because it had gone through generic_m_r.
* The file system will likely explode at this point due to
* I/O errors. If it wasn't the primary that failed, it is
* easily possible in this case to issue writes to just one
* of the remaining images - also leaving the mirror inconsistent.
*
* We let in_sync() return 1 in a cluster regardless of what is
* in the bitmap once recovery has successfully completed on a
* mirror. This ensures the mirroring code will continue to
* attempt to write to all mirror images. The worst that can
* happen for reads is that additional read attempts may be
* taken.
Ignore snapshots when performing mirror recovery beneath an origin.
Pass LCK_ORIGIN_ONLY flag around cluster.
Add suspend_lv_origin and resume_lv_origin using LCK_ORIGIN_ONLY.
DM devices were not handled properly on nodes in a cluster that were not
where the splitmirrors command was issued. This was happening because
suspend_lv/resume_lv were being used in a place where activate_lv should
have been used.
When the suspend/resume are issued on (effectively) new LVs, their
'resource' (UUID) is not located in the lv_hash. Thus, both operations
turn into no-ops. You can see this from the output of clvmd from one
of the remote nodes:
<snip>
do_suspend_lv, lock not already held
<snip>
do_resume_lv, lock not already held
'activate_lv' enjoins the other nodes in the cluster to process the lock
and activate the new LV. clvmd output from remote node as follows:
do_lock_lv: resource 'zMseY7CBuO3Ty09vXlplPAHzD0Y0CovjrTdv0R1VcwggMwPdYhutHErRcwm5Nd2S', cmd = 0x19 LCK_LV_ACTIVATE (READ|LV|NONBLOCK), flags = 0x84 (DMEVENTD_MONITOR ), memlock = 1
sync_lock: 'zMseY7CBuO3Ty09vXlplPAHzD0Y0CovjrTdv0R1VcwggMwPdYhutHErRcwm5Nd2S' mode:1 flags=1
sync_lock: returning lkid 27b0001
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Petr Rockai <prockai@redhat.com>
clvmd daemon itself does the right thing when invoked as non-root, by
returning 4.
The patch removes the use daemon function from
/etc/rc.d/init.d/functions that´s unnecessary and has th bad habit to
mask the return codes from the real daemon.
Add a simple and generic check to see if clvmd is executed by root or not.
Our stop/reload/restart paths in the init script are complex and not all
the tools involved in the process are guaranteed to return 4 if executed
by non-root against a process that´s running as root (for example kill
-TERM will return -1 and parsing the output to catch the error is
suboptimal at best).
https://bugzilla.redhat.com/show_bug.cgi?id=553381
The new standard in the storage industry is to default alignment of data
areas to 1MB. fdisk, parted, and mdadm have all been updated to this
default.
Update LVM to align the PV's data area start (pe_start) to 1MB. This
provides a more useful default than the previous default of 64K (which
generally ended up being a 192K pe_start once the first metadata area
was created).
Before this patch:
# pvs -o name,vg_mda_size,pe_start
PV VMdaSize 1st PE
/dev/sdd 188.00k 192.00k
After this patch:
# pvs -o name,vg_mda_size,pe_start
PV VMdaSize 1st PE
/dev/sdd 1020.00k 1.00m
The heuristic for setting the default alignment for LVM data areas is:
- If the default value (1MB) is a multiple of the detected alignment
then just use the default.
- Otherwise, use the detected value.
In practice this means we'll almost always use 1MB -- that is unless:
- the alignment was explicitly specified with --dataalignment
- or MD's full stripe width, or the {minimum,optimal}_io_size exceeds
1MB
- or the specified/detected value is not a power-of-2
Introduce --norestorefile to allow user to override the new requirement.
This can also be overridden with "devices/require_restorefile_with_uuid"
in lvm.conf -- however the default is 1.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
We can already detect MD devices internally. But when using MD partitions,
these have "block extended major" (blkext) assigned (259). Blkext major
is also used in general, so we need to check whether the original device
is an MD device actually.
An incorrect fix on July 13, 2010 for an annoyance has caused a regression.
The offending check-in was part of the 2.02.71 release of LVM. That
check-in caused any PVs specified on the command line to be ignored when
performing a mirror split.
This patch reverses the aforementioned check-in (solving the regressions)
and posits a new solution to the list reversal problem. The original
problem was that we would always take the lowest mimage LVs from a mirror
when performing a split, but what we really want is to take the highest
mimage LVs. This patch accomplishes that by working through the list in
reverse order - choosing the higher numbered mimages first. (This also
reduces the amount of processing necessary.)
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Takahiro Yasui <takahiro.yasui@hds.com>
corruption bug in cmirror. 'dm_bit' is only ever used as a boolean operation
within LVM, but it can return a range of values. If the bit is set, a power of
2 is returned. If the bit is unset, 0 is returned.
'log_test_bit' (a function in the cluster mirror log daemon code) has switched
to using the dm bit operations in rhel6. There are two places in the daemon
code where 'log_test_bit' is not used merely as a boolean, but rather the
return value is used as the return value for the log functions 'is_clean' and
'in_sync' - having assumed that 'dm_bit' was returning 0 or 1 only.
One place the 'in_sync' function is utilized is in 'dm_rh_get_state' - a
function that informs the mirroring code how to treat I/O and which devices to
read/write from. 'dm_rh_get_state' was checking if the return value of
'in_sync' was 1 to determine if the region was DM_RH_CLEAN. Since 'dm_bit'
(and by extension 'log_test_bit' and 'in_sync') was returning powers of 2,
DM_RH_CLEAN was rarely being reported as it should have been. Thinking the
region was out-of-sync, the mirroring code would write only to the primary
device. When the primary device was failed, all of those writes were lost -
leaving the entire mirror corrupted.
all but one mirror leg.
<patch header>
To handle a double failure of a mirrored log, Jon's two patches are
commited, however, lvconvert command can't still handle an error
when mirror leg and mirrored log got failure at the same time.
[Patch]: Handle both devices of a mirrored log failing (bug 607347)
posted: https://www.redhat.com/archives/lvm-devel/2010-July/msg00009.html
commit: https://www.redhat.com/archives/lvm-devel/2010-July/msg00027.html
[Patch]: Handle both devices of a mirrored log failing (bug 607347) -
additional fix
posted: https://www.redhat.com/archives/lvm-devel/2010-July/msg00093.html
commit: https://www.redhat.com/archives/lvm-devel/2010-July/msg00101.html
In the second patch, the target type of mirrored log is replaced with
error target when remove_log is set to 1, but this procedure should be
also used in other cases such as the number of mirror leg is 1. This
patch relocates the procedure to the main path.
In addition, I added following three changes.
- Removed tmp_orphan_lvs handling procedure
It seems that _delete_lv() can handle detached_log_lv properly
without adding mirror legs in mirrored log to tmp_orphan_lvs.
Therefore, I removed the procedure.
- Removed vg_write()/vg_commit()
Metadata is saved by vg_write()/vg_commit() just after detached_log_lv
is handled. Therefore, I removed vg_write()/vg_commit().
- With Jon's second patch, we think that we don't have to call
remove_mirror_log() in _lv_update_mirrored_log() because will be
handled remove_mirror_images() in _lvconvert_mirrors_repaire().
</patch header>
Signed-off-by: Takahiro Yasui <takahiro.yasui@hds.com>
Reviewed-by: Petr Rockai <prockai@redhat.com>
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
The cluster log daemon (cmirrord) is not multi-threaded and
can handle only one request at a time. When a log is stacked
on top of a mirror (which itself contains a 'core' log), it
creates a situation that cannot be solved without threading.
When the top level mirror issues a "resume", the log daemon
attempts to read from the log device to retrieve the log
state. However, the log is a mirror which, before issuing
the read, attempts to determine the 'sync' status of the
region of the mirror which is to be read. This sync status
request cannot be completed by the daemon because it is
blocked on a read I/O to the very mirror requesting the
sync status.
With mirror_log_fault_policy of 'remove' and mirror_image_fault_policy
of 'allocate', the log type of the mirror volume is converted from
'disk' or 'mirrored' to 'core' when all mirror legs but one in a mirror
volume broke.
Keep new_log_count as a number of valid log devices by using log_count
variable for a temporary usage in the first phase of error recovery
in _lvconvert_mirrors_repair().
Signed-off-by: Takahiro Yasui <takahiro.yasui@hds.com>
Reviewed-by: Petr Rockai <prockai@redhat.com>
mirrors, we must also check that the log daemon (cmirrord) is running.
The log module can be auto-loaded, but the daemon cannot be
"auto-started". Failing to check for the daemon produces cryptic
messages that customers have a hard time deciphering. (The system
messages do report that the log daemon is not running, but people
don't seem to find this message easily.)
Here are examples of what is printed when the module is available,
but the log daemon has not been started.
[root@bp-01 LVM2]# lvcreate -m1 -l1 -n lv vg
Shared cluster mirrors are not available.
[root@bp-01 LVM2]# lvcreate -m1 -l1 -n lv vg -v
Setting logging type to disk
Finding volume group "vg"
Archiving volume group "vg" metadata (seqno 3).
Creating logical volume lv
Executing: /sbin/modprobe dm-log-userspace
Cluster mirror log daemon is not running
Shared cluster mirrors are not available.
Creating volume group backup "/etc/lvm/backup/vg" (seqno 4).
The main problem with these bugs was that the newly split
off LV was not being suspended properly. This meant that
the memlock count was not being balanced, the DM devices
were not being renamed, and some DM devices which should
have been removed were not.
I've also renamed some of the variables and added comments
to make things clearer as to what is going on. (I can break
this patch in two if it means easier review.)
Switch dmeventd to use dm_create_lockfile and drop duplicate code.
Allow clvmd pidfile to be configurable.
Switch cmirrord and clvmd to use dm_create_lockfile.
This should bring less confusion when there are some settings left and
people just forgot about it and then they run into problems. These messages
should give them a hint of what's really going on.
A previous check-in added logic to handle the case where both images
of a mirrored log failed. It solved the problem by simply removing
the log entirely - leaving the parent mirror with a 'core' log. This
worked for most cases. However, if there was a small delay between
the failures of the two mirrored log devices, the mirror would hang,
LVM would hang, and no additional LVM commands could be issued.
When the first leg of the log fails, it signals the need for repair.
Before 'lvconvert --repair' is run by dmeventd, the second leg fails.
'lvconvert' would see both devices as failed and try to remove the
log entirely. When it came time to suspend the parent mirror to
update the configuration, the suspend would hang because it couldn't
get any I/O through the mirrored log, which was plugged waiting for
corrective action. The solution is to replace the log with an error
target to clear any pending writes before removing it. This allows
the parent mirror to suspend and make the proper changes.
When using vgmetadatacopies value other than "umanaged" (0), prompt
the user if the usage of --metadataignore would change the value of
vgmetadatacopies. The main 2 cases are:
1) pvchange --metadataignore
2) vgextend --metadataignore
We leave the prompt check in the tools, and do not change anything
if the user says 'n'.
Examples:
vgextend --metadataignore y vgtest /dev/loop0
Setting metadataignore will override preferred number of copies of VG vgtest metadata.
Are you sure? [y/n]: y
No physical volume label read from /dev/loop0
Physical volume "/dev/loop0" successfully created
Volume group "vgtest" successfully extended
pvchange --metadataignore y /dev/loop3
Setting metadataignore on /dev/loop3 will override preferred number of copies of VG vgtest metadata.
Are you sure? [y/n]: y
WARNING: Changing preferred number of copies of VG vgtest metadata from 3 to 2
Physical volume "/dev/loop3" changed
1 physical volume changed / 0 physical volumes not changed
Signed-off-by: Dave Wysochanski <dwysocha@redhat.com>
- If a PV contained empty mdas, the auto-recovery code was not kicking in.
- The 'inconsistent' state was getting lost when metadata was cached so
recovery didn't kick in. But leave the behaviour alone when using
precommitted metadata because of a warning in a confusing FIXME.
In my testing, pvs and vgs didn't repair inconsistent metadata like they
used to do. (How many other tools fail similarly now?)
And there should be no need to cache inconsistent metadata because it is
supposed to get repaired under the protection of a write lock immediately it is
discovered.
This code is in need of a redesign based on first principles.
I still see bugs in this code and this commit is risky.
Rather than attempting to remove all the images of a mirrored
log volume via remove_mirror_images, simply remove the log
if all its devices have failed.
Taka was the first to report that there is still an outstanding
issue with handling this case. I've managed to reproduce it
only very rarely, and am still working on identifying the problem.
Failing to handle the problem rarely is better than not handling
the scenario at all, so I'm checking this in.
