Accept --ignoreskippedcluster with pvs, vgs, lvs, pvdisplay, vgdisplay,
lvdisplay, vgchange and lvchange to avoid the 'Skipping clustered
VG' errors when requesting information about a clustered VG
without using clustered locking and still exit with success.
Peter Rajnoha [Mon, 30 Sep 2013 09:19:09 +0000 (11:19 +0200)]
libdm: export DM_UDEV_SUBSYSTEM_FLAG names for subystem udev flags
Just like we have symbolic names assigned to general DM udev flags
(DM_UDEV_* flags), we have the same for any subsystem flags now
(DM_SUBSYSTEM_UDEV_FLAG*), making it easier to use.
Peter Rajnoha [Mon, 30 Sep 2013 09:11:18 +0000 (11:11 +0200)]
udev: make subsystem rules responsible for importing subsystem flags
Each subsystem rule that needs to import any of DM_SUBSYSTEM_UDEV_FLAG*
flags is responsible for doing so. This simply moves control of these
flags from general 10-dm.rules to any subsystem rule using these flags
as each subsystem knows better how to handle these flags on its own.
Peter Rajnoha [Fri, 27 Sep 2013 11:58:55 +0000 (13:58 +0200)]
fix: make it possible to compile with --disable-devmapper again
Some code has been added recently which makes it impossible to compile
when "configure --disable-devmapper" is used. This patch just shuffles
the code around so it's under proper #ifdef DEVMAPPER_SUPPORT.
Jonathan Brassow [Thu, 26 Sep 2013 16:30:07 +0000 (11:30 -0500)]
RAID: Fix _sufficient_pes_free calculation for RAID
lib/metadata/lv_manip.c:_sufficient_pes_free() was calculating the
required space for RAID allocations incorrectly due to double
accounting. This resulted in failure to allocate when available
space was tight.
When RAID data and metadata areas are allocated together, the total
amount is stored in ah->new_extents and ah->alloc_and_split_meta is
set. '_sufficient_pes_free' was adding the necessary metadata extents
to ah->new_extents without ever checking ah->alloc_and_split_meta.
This often led to double accounting of the metadata extents. This
patch checks 'ah->alloc_and_split_meta' to perform proper calculations
for RAID.
This error is only present in the function that checks for the needed
space, not in the functions that do the actual allocation.
Jonathan Brassow [Thu, 26 Sep 2013 03:25:43 +0000 (22:25 -0500)]
mirror/RAID: Honor mirror_segtype_default when converting from linear
1) When converting from an x-way mirror/raid1 to a y-way mirror/raid1,
the default behaviour should be to stay the same segment type.
2) When converting from linear to mirror or raid1, the default behaviour
should honor the mirror_segtype_default.
3) When converting and the '--type' argument is specified, the '--type'
argument should be honored.
catch such conditions, but errors in the tests caused the issue to go
unnoticed. The code has been fixed to perform #2 properly, the tests
have been corrected to properly test for #2, and a few other tests
were changed to explicitly specify the '--type mirror' when necessary.
Peter Rajnoha [Wed, 25 Sep 2013 14:01:47 +0000 (16:01 +0200)]
thin: calculate thin pool chunk size based on device IO hints
If "default" thin pool chunk size calculation method is selected,
use minimum_io_size, otherwise optimal_io_size for "performance"
device hint exposed in sysfs. If there appear to be PVs with
different hints presented, use their least common multiple.
If the hint is less than the default value defined for the
calculation method, use the default value instead.
Add allocation/thin_pool_chunk_size_calculation lvm.conf
option to select a method for calculating thin pool chunk
sizes and define two possible values - "default" and "performance".
Jonathan Brassow [Wed, 25 Sep 2013 02:32:10 +0000 (21:32 -0500)]
RAID: Fix broken allocation policies for parity RAID types
A previous commit (b6bfddcd0a830d0c9312bc3ab906cb3d1b7a6dd9) which
was designed to prevent segfaults during lvextend when trying to
extend striped logical volumes forgot to include calculations for
RAID4/5/6 parity devices. This was causing the 'contiguous' and
'cling_by_tags' allocation policies to fail for RAID 4/5/6.
The solution is to remember that while we can compare
ah->area_count == prev_lvseg->area_count
for non-RAID, we should compare
(ah->area_count + ah->parity_count) == prev_lvseg->area_count
for a general solution.
Peter Rajnoha [Tue, 24 Sep 2013 12:03:42 +0000 (14:03 +0200)]
lvmconf: use_lvmetad=0 on --enable-cluster, reset to default on --disable-cluster
lvmetad is not yet supported in clustered environment so
disable it automatically if using lvmconf --enable-cluster
and reset it to default value if using lvmconf --disable-cluster.
Also, add a few comments in lvm.conf about locking_type vs. use_lvmetad
if setting it for clustered environment.
clvmd: Fix node up/down handing in corosync module
The corosync cluster interface for clvmd did not correctly
deal with node up/down events so that when a node was removed
from the cluster clvmd would prevent remote operations
from happening, as it thought the node was up but not
running clvmd.
This patch fixes that code by simplifying the case to node
being up or down - which was the original intention
and is supported by pacemaker and CPG in the higher layers.
Jonathan Brassow [Fri, 20 Sep 2013 16:33:29 +0000 (11:33 -0500)]
test: Blacklist fedora-19 kernel for RAID4/5/6 dev replace tests
A know issue with kmem_cach is causing failures while testing
RAID 4/5/6 device replacement. Blacklist the offending kernel
so that these tests are not performed there.
Add internal devtypes reporting command to display built-in recognised
block device types. (The output does not include any additional
types added by a configuration file.)
> lvm devtypes -o help
Device Types Fields
-------------------
devtype_all - All fields in this section.
devtype_name - Name of Device Type exactly as it appears in /proc/devices.
devtype_max_partitions - Maximum number of partitions. (How many device minor numbers get reserved for each device.)
devtype_description - Description of Device Type.
> lvm devtypes
DevType MaxParts Description
aoe 16 ATA over Ethernet
ataraid 16 ATA Raid
bcache 1 bcache block device cache
blkext 1 Extended device partitions
...
The CC override in commit f42b2d4bbf16345e5b5457f4298e751d5c134776
caused the built-in value to be used instead of the configured value
when it wasn't being overridden.
The behaviour is explained here:
http://stackoverflow.com/questions/18007326/how-to-change-default-values-of-variables-like-cc-in-makefile
Since our current vgcfgbackup/restore doesn't deal
with difference of active volumes between current and
restored set of volumes - run test with inactive LVs.
Peter Rajnoha [Mon, 16 Sep 2013 09:47:09 +0000 (11:47 +0200)]
systemd: run lvm2-activation-net.service after lvm2-activation.service
The lvm2-activation-net.service was ordered only with respect to iscsi
and fcoe service before. In addition to that, we also need ordering
with respect to lvm2-activation.service to prevent parallel vgchange -aay
runs which may cause some problems during activation.
See also https://bugs.gentoo.org/show_bug.cgi?id=480066.
With this patch, the ordering is firmly set to:
lvm2-activation-early.service -> lvm2-activation.service -> lvm2-activation-net.service
Thanks to Alexander Tsoy for the original patch (modified a bit here):
https://www.redhat.com/archives/lvm-devel/2013-September/msg00049.html
Rewrite check lv_on and add new lv_tree_on
Move more pvmove test unrelated code out to check & get sections
(so they do not obfuscate trace output unnecesserily)
Use new lv_tree_on()
NOTE: unsure how the snapshot origin should be accounted here.
Split pmove-all-segments into separate tests for raid and thins
(so the test output properly shows what has been skipped in test)
Update usage of "" around shell vars.
trim needs to trim both sides now.
trim also removes debug.log since it's only called when lvm command
has finished properly (so if something fails afterward, there
is no missleading debug trace in the log)
Jonathan Brassow [Thu, 12 Sep 2013 18:23:53 +0000 (13:23 -0500)]
TEST: Unaccounted possible output causing failure
lvchange-raid.sh checks to ensure that the 'p'artial flag takes
precedence over the 'w'ritemostly flag by disabling and reenabling
a device in the array. Most of the time this works fine, but
sometimes the kernel can notice the device failure before it is
reenabled. In that case, the attr flag will not return to 'w', but
to 'r'efresh. This is because 'r'efresh also takes precedence over
the 'w'ritemostly flag. So, we also do a quick check for 'r' and
not just 'w'.
Peter Rajnoha [Thu, 12 Sep 2013 11:46:20 +0000 (13:46 +0200)]
udev: keep DM_ACTIVATION and DM_UDEV_PRIMARY_SOURCE_FLAG meaning as before commit 8d1d835
The DM_ACTIVATION and DM_UDEV_PRIMARY_SOURCE_FLAG needs to be kept the
way it was for backward compatibility (e.g. the old rules are still
in initramfs). This way the check in whether the device should be
scanned in 69-dm-lvmetad.rules is even easier.
tests: extend harness with output of /var/log/messages
Add a very simple hack for embeding /var/log/messages into
the tests output - it's not ideal since it sometimes breaks lines,
but still gives valuable info.
Add more 'realistic' simulation of dlm locking.
Previous version was not capable to maintain multiple locks.
Current version doesn't handle multiqueues for locks,
so the ordering is different.
Jonathan Brassow [Wed, 11 Sep 2013 20:58:44 +0000 (15:58 -0500)]
Mirror/Thin: Disallow thinpools on mirror logical volumes
The same corner cases that exist for snapshots on mirrors exist for
any logical volume layered on top of mirror. (One example is when
a mirror image fails and a non-repair LVM command is the first to
detect it via label reading. In this case, the LVM command will hang
and prevent the necessary LVM repair command from running.) When
a better alternative exists, it makes no sense to allow a new target
to stack on mirrors as a new feature. Since, RAID is now capable of
running EX in a cluster and thin is not active-active aware, it makes
sense to pair these two rather than mirror+thinpool.
As further background, here are some additional comments that I made
when addressing a bug related to mirror+thinpool:
(https://bugzilla.redhat.com/show_bug.cgi?id=919604#c9)
I am going to disallow thin* on top of mirror logical volumes.
Users will have to use the "raid1" segment type if they want this.
This bug has come down to a choice between:
1) Disallowing thin-LVs from being used as PVs.
2) Disallowing thinpools on top of mirrors.
The problem is that the code in dev_manager.c:device_is_usable() is unable
to tell whether there is a mirror device lower in the stack from the device
being checked. Pretty much anything layered on top of a mirror will suffer
from this problem. (Snapshots are a good example of this; and option #1
above has been chosen to deal with them. This can also be seen in
dev_manager.c:device_is_usable().) When a mirror failure occurs, the
kernel blocks all I/O to it. If there is an LVM command that comes along
to do the repair (or a different operation that requires label reading), it
would normally avoid the mirror when it sees that it is blocked. However,
if there is a snapshot or a thin-LV that is on a mirror, the above code
will not detect the mirror underneath and will issue label reading I/O.
This causes the command to hang.
Choosing #1 would mean that thin-LVs could never be used as PVs - even if
they are stacked on something other than mirrors.
Choosing #2 means that thinpools can never be placed on mirrors. This is
probably better than we think, since it is preferred that people use the
"raid1" segment type in the first place. However, RAID* cannot currently
be used in a cluster volume group - even in EX-only mode. Thus, a complete
solution for option #2 must include the ability to activate RAID logical
volumes (and perform RAID operations) in a cluster volume group. I've
already begun working on this.
Peter Rajnoha [Wed, 11 Sep 2013 10:42:29 +0000 (12:42 +0200)]
udev: override new udev default timeout of 30s to original 3min
New versions of udev changed the default event timeout to 30s
from original 3min. This causes problems with LVM processes that
starve because of the IO load caused by some LVM actions (e.g.
mirror/raid synchronization).
Reinstate the 3min udev timeout for now until we optimize this
in a way that even the 30s timeout is sufficient.
Jonathan Brassow [Tue, 10 Sep 2013 21:33:22 +0000 (16:33 -0500)]
RAID: Make RAID single-machine-exclusive capable in a cluster
Creation, deletion, [de]activation, repair, conversion, scrubbing
and changing operations are all now available for RAID LVs in a
cluster - provided that they are activated exclusively.
The code has been changed to ensure that no LV or sub-LV activation
is attempted cluster-wide. This includes the often overlooked
operations of activating metadata areas for the brief time it takes
to clear them. Additionally, some 'resume_lv' operations were
replaced with 'activate_lv_excl_local' when sub-LVs were promoted
to top-level LVs for removal, clearing or extraction. This was
necessary because it forces the appropriate renaming actions the
occur via resume in the single-machine case, but won't happen in
a cluster due to the necessity of acquiring a lock first.
The *raid* tests have been updated to allow testing in a cluster.
For the most part, this meant creating devices with '-aey' if they
were to be converted to RAID. (RAID requires the converting LV to
be EX because it is a condition of activation for the RAID LV in
a cluster.)
Peter Rajnoha [Tue, 10 Sep 2013 13:49:05 +0000 (15:49 +0200)]
udev: fix pvscan --cache -aay to trigger on relevant events
This patch fixes the way the special devices are handled
(special in this context means that they're not usable
after the usual ADD event like other generic devices):
- DM and MD devices are pvscanned only when they are just set up.
This is the first CHANGE event that makes the device usable
(the DM_UDEV_PRIMARY_SOURCE_FLAG is set for DM and the
md/array_state sysfs attribute is present for MD).
Whether the device is activated is remembered via
DM_ACTIVATED (for DM) and LVM_MD_PV_ACTIVATED (for MD)
udev environment variable. This is then used to decide
whether we should fire the pvscan on ADD event to
support coldplugging. For any (artificial) ADD event
generated during coldplug, the device must be already
set up properly to fire the pvscan on it.
- Similar for loop devices. For loop devices, only CHANGE
events are relevant (so there's a CHANGE after the loop
device is set up as well as detached). Whether the loop
has just been activated is detected via loop/backing_file
sysfs attribute presence. The activation state is remembered
via LVM_LOOP_PV_ACTIVATED udev environment variable.
- Do not pvscan multipath device components (underlying paths).
- Do not pvscan RAID device components.
- Also, set LVM_SCANNED="1" udev environment variable for
debug purposes (it's visible in the lvmdump -u that takes
the current udev database). This variable is set once
the pvscan is triggered.
The table below summarises when the pvscan is triggered
(marked with X, X* means fire only if the special dev is properly set up):
| real ADD | real CHANGE | artificial ADD | artificial CHANGE | remove
=============================================================================
DM | | X | X* | | X
MD | | X | X* | |
loop | | X | X* | |
other | X | | X | | X
Misc/RAID: Enable resume_lv to handle some renaming conflicts.
When images and their associated metadata are removed from a RAID1 LV,
the remaining sub-LVs are "shifted" down to fill the gaps. For
example, if there is a 3-way mirror:
[0][1][2]
and we remove device#0, the devices will be shifted down
[1][2]
and renamed.
[0][1]
This can create a problem for resume_lv (specifically,
dm_tree_activate_children) during the renaming process though. This
is because it will attempt to rename the higher indexed sub-LVs first
and find that it cannot because there are currently other sub-LVs with
that name. The solution is to check for a conflicting name before
attempting to rename. If a conflict is found and that conflicting
sub-LV is also in the process of renaming, we can defer the current
rename until the conflicting sub-LV has renamed and cleared the
conflict.
Now that resume_lv can handle these types of rename conflicts, we can
remove the workaround in RAID that was attempting to resume a RAID1
LV from the bottom-up in order to force a proper rename in assending
order before attempting a resume on the top-level LV. This "hack"
only worked for single machine use-cases of LVM. Clearing this up
paves the way for exclusive activation of RAID LVs in a cluster.
Properly skip unmonitoring of thin pool volume in deactivation code
path. Code makes sure if there is just any thin pool user
it stays monitored with all its resources.
When the pool is created from non-linear target the more complex rules
have to be used and stacking needs to properly decode args for _tdata
LV. Also proper allocation policies are being used according to those
set in lvm2 metadata for data and metadata LVs.
Also properly check for active pool and extra code to active it
temporarily.