]> sourceware.org Git - lvm2.git/log
lvm2.git
9 years agolibdaemon: fix typo in last log enable commit
David Teigland [Wed, 21 Oct 2015 16:22:22 +0000 (11:22 -0500)]
libdaemon: fix typo in last log enable commit

9 years agolibdaemon: allow all debug levels for stderr
David Teigland [Wed, 21 Oct 2015 16:16:59 +0000 (11:16 -0500)]
libdaemon: allow all debug levels for stderr

The parsing code for log levels was missing
some levels.  Also ignore an unrecognized
level name rather than failing.

9 years agotests: add lvmetad shutdown on idle test
Ondrej Kozina [Tue, 20 Oct 2015 13:41:42 +0000 (15:41 +0200)]
tests: add lvmetad shutdown on idle test

9 years agolvmetad.8.in: update man page with timeout info
Ondrej Kozina [Tue, 20 Oct 2015 11:33:29 +0000 (13:33 +0200)]
lvmetad.8.in: update man page with timeout info

9 years agolvmetad: add optional timeout option
Ondrej Kozina [Tue, 20 Oct 2015 11:24:50 +0000 (13:24 +0200)]
lvmetad: add optional timeout option

9 years agotests: add test for missing mirror in-sync info
Ondrej Kozina [Tue, 20 Oct 2015 09:55:51 +0000 (11:55 +0200)]
tests: add test for missing mirror in-sync info

9 years agopolldaemon: fix missing mirror in-sync info with lvmpolld
Ondrej Kozina [Mon, 19 Oct 2015 14:56:45 +0000 (16:56 +0200)]
polldaemon: fix missing mirror in-sync info with lvmpolld

CONVERTING status flag is a tricky one. It's not set when converting
a non-mirror LV type to the mirror type, i.e.: linear -> two leg mirror.
Also the conversion itself is instant and doesn't require to be polled.
When mirror reaches sync state there's no final update on VG metadata
for lvmpolld to be made thereby report_progress in fact doesn't report
percentage of mirror being converted but percentage of mirror
being in sync. Perhaps we should reword the lvconvert output here.

On the other hand CONVERTING is set while we upconvert the mirror
from i.e. two leg mirror to four leg mirror. In such case the operation
is required to be polled so that lvmpolld can cleanup temporary
conversion log when the conversion is over.

Ignore CONVERTING lv_type for the moment and match LVs only by uuids
during 'mirror conversion'/'waiting for a sync to finish'.

9 years agoconf: clarify compact output settings
Peter Rajnoha [Tue, 20 Oct 2015 09:01:09 +0000 (11:01 +0200)]
conf: clarify compact output settings

9 years agopvscan: use process_each_pv
David Teigland [Mon, 19 Oct 2015 18:58:43 +0000 (13:58 -0500)]
pvscan: use process_each_pv

The old code made two loops through the PVs: in the first
loop it found the max PV and VG name lengths, and in the
second loop it printed each PV using the name lengths as
field widths for aligning columns.

The new code uses process_each_pv() which makes one loop
through the PVs.  In the *first* call to pvscan_single(),
the max name lengths are found by looping through the
lvmcache entries which have been populated by the generic
process_each code prior to calling any _single functions.
Subsequent calls to pvscan_single() reuse the max lengths
that were found by the first call.

9 years agoconf: regenerate
Peter Rajnoha [Mon, 19 Oct 2015 08:01:14 +0000 (10:01 +0200)]
conf: regenerate

9 years agocleanup: typo in comment: compact_output_fields should be compact-output_cols
Peter Rajnoha [Mon, 19 Oct 2015 07:59:05 +0000 (09:59 +0200)]
cleanup: typo in comment: compact_output_fields should be compact-output_cols

9 years agoconfig: add report/compact_output_cols to control which columns to compact in report...
Peter Rajnoha [Fri, 16 Oct 2015 14:25:51 +0000 (16:25 +0200)]
config: add report/compact_output_cols to control which columns to compact in report output

The new report/compact_output_cols setting has exactly the same effect
as report/compact_output setting. The difference is that with the new
setting it's possible to define which cols should be compacted exactly
in contrast to all cols in case of report/compact_output.

In case both compact_output and compact_output_cols is enabled/set,
the compact_output prevails.

For example:

$ lvmconfig --type full report/compact_output report/compact_output_cols
compact_output=0
compact_output_cols=""

$ lvs vg
  LV    VG   Attr       LSize Pool Origin Data%  Meta%  Move Log Cpy%Sync Convert
  lvol0 vg   -wi-a----- 4.00m

---

$ lvmconfig --type full report/compact_output report/compact_output_cols
compact_output=0
compact_output_cols="data_percent,metadata_percent,pool_lv,move_pv,origin"

$ lvs vg
  LV    VG   Attr       LSize Log Cpy%Sync Convert
  lvol0 vg   -wi-a----- 4.00m

---

$ lvmconfig --type full report/compact_output report/compact_output_cols
compact_output=1
compact_output_cols="data_percent,metadata_percent,pool_lv,move_pv,origin"

$ lvs vg
  LV    VG   Attr       LSize
  lvol0 vg   -wi-a----- 4.00m

9 years agolibdm: add dm_report_compact_given_fields
Peter Rajnoha [Fri, 16 Oct 2015 13:50:13 +0000 (15:50 +0200)]
libdm: add dm_report_compact_given_fields

dm_report_compact_given_fields is the same as dm_report_compact_fields,
but it processes only given fields, not all the fields in the report
like dm_report_compact_field does.

9 years agolvmlockd: use flag to avoid blocking in sanlock_acquire
David Teigland [Wed, 14 Oct 2015 19:36:46 +0000 (14:36 -0500)]
lvmlockd: use flag to avoid blocking in sanlock_acquire

If a host failed while holding a sanlock lease,
sanlock_acquire will by default block and wait
for the lease to expire before returning.  We
want it to return with an error so we can retry
instead of blocking, which allows us to process
other lock operations.

(Enclose this in an ifdef until the new flag
appears in a sanlock release.)

9 years agodmeventd: use matching function
Zdenek Kabelac [Wed, 14 Oct 2015 12:19:47 +0000 (14:19 +0200)]
dmeventd: use matching function

Respect lvm2_log_fn prototype. The idea of 'reusing' print_log with
plain cast is causing very strange crashes with some older 'gcc'  compilers.
So just do it cleanly...

9 years agothin: enforce local activation when creation new thin
Zdenek Kabelac [Tue, 13 Oct 2015 23:00:35 +0000 (01:00 +0200)]
thin: enforce local activation when creation new thin

As we need to check how full thin-pool is - require thin-pool is
locally active.

9 years agomakefiles: switch to rpath-link
Zdenek Kabelac [Tue, 13 Oct 2015 22:51:55 +0000 (00:51 +0200)]
makefiles: switch to rpath-link

Plain rpath is to invasive and gets into binary.
We only want to provide hint for linker.

9 years agodmeventd: fix missing '!'
Zdenek Kabelac [Tue, 13 Oct 2015 20:10:47 +0000 (22:10 +0200)]
dmeventd: fix missing '!'

During recent code changes '!' was badly converted.

9 years agomakefiles: older gcc needs hint with rpath
Zdenek Kabelac [Tue, 13 Oct 2015 20:02:17 +0000 (22:02 +0200)]
makefiles: older gcc needs hint with rpath

gcc 4.3 seems not to be able to find linked library without
specifying -rpath to linker  (plain -L) is not enough.

9 years agodmeventd: compilable without DEBUG CFLAG
Zdenek Kabelac [Tue, 13 Oct 2015 18:57:59 +0000 (20:57 +0200)]
dmeventd: compilable without DEBUG CFLAG

Missed compilability without DEBUG.

9 years agoRevert "log: no file for external logging"
Alasdair G Kergon [Tue, 13 Oct 2015 14:31:57 +0000 (15:31 +0100)]
Revert "log: no file for external logging"

This reverts commit 1b1c01a27b359f8e91c3c9e08684c435d8fd51c2.

This caused messages to get dropped instead of logged into the log file.

(The log file and log function are independent at the moment.)

9 years agotests: more extend testing
Zdenek Kabelac [Tue, 13 Oct 2015 12:33:24 +0000 (14:33 +0200)]
tests: more extend testing

9 years agotests: wait for initial sync
Zdenek Kabelac [Tue, 13 Oct 2015 09:21:40 +0000 (11:21 +0200)]
tests: wait for initial sync

Raid should be in-sync ATM for any gaming.

9 years agotests: start logging dmeventd output
Zdenek Kabelac [Tue, 13 Oct 2015 07:44:08 +0000 (09:44 +0200)]
tests: start logging dmeventd output

Log dmeventd output from test - keep output logging in test
and avoid sending messages to syslog.

9 years agotests: better regex check
Zdenek Kabelac [Fri, 9 Oct 2015 19:30:37 +0000 (21:30 +0200)]
tests: better regex check

Enhance check for matching prefix.

9 years agotests: show status in in_sync
Zdenek Kabelac [Wed, 7 Oct 2015 13:07:15 +0000 (15:07 +0200)]
tests: show status in in_sync

Make visible actual status line when returing 'raid' is in sync.
Useful to prove when kernel is wrong.

9 years agomakefiles: src dirs have depth 5
Zdenek Kabelac [Sun, 11 Oct 2015 22:10:37 +0000 (00:10 +0200)]
makefiles: src dirs have depth 5

For tags generation scan depth 5.

9 years agomakefiles: drop explicit linking
Zdenek Kabelac [Fri, 9 Oct 2015 19:33:59 +0000 (21:33 +0200)]
makefiles: drop explicit linking

Nothing in /sscripts needs devmapper-event library so drop its linking.

9 years agomakefiles: generate Local
Zdenek Kabelac [Fri, 9 Oct 2015 19:31:41 +0000 (21:31 +0200)]
makefiles: generate Local

Use same  exported symbols list for other generated symfile
as it appeared to be the right way for libdm.

Local {
local:
*;
};

9 years agodmeventd: rework raid plugin
Zdenek Kabelac [Tue, 13 Oct 2015 09:37:14 +0000 (11:37 +0200)]
dmeventd: rework raid plugin

Switch to per-thread mem-pool and lock only for lvm2 calls.
Use libdm parser for raid status line.

9 years agodmeventd: less locking mirror
Zdenek Kabelac [Tue, 13 Oct 2015 09:35:44 +0000 (11:35 +0200)]
dmeventd: less locking mirror

Rework mirror plugin to use per-thread mem-pool.
This allows to use locking only for lvm2 library calls.

9 years agodmeventd: less locking for snapshot and thin
Zdenek Kabelac [Tue, 13 Oct 2015 09:35:03 +0000 (11:35 +0200)]
dmeventd: less locking for snapshot and thin

Use lvm lock for lvm library calls.

dm functions in-use are 'thread-safe' since we use
local per-thread mem pool.

9 years agodmeventd: add local mempool for raid and mirror
Zdenek Kabelac [Tue, 13 Oct 2015 09:28:57 +0000 (11:28 +0200)]
dmeventd: add local mempool for raid and mirror

Using local mempools allows to drop locks when such memory is needed.

9 years agodmeventd: use new macros to init mempool
Zdenek Kabelac [Tue, 13 Oct 2015 09:30:37 +0000 (11:30 +0200)]
dmeventd: use new macros to init mempool

For thin and snapshot use new macros
to simplify mempool init and destroy

9 years agodmeventd: introduce macro for init
Zdenek Kabelac [Mon, 12 Oct 2015 09:37:36 +0000 (11:37 +0200)]
dmeventd: introduce macro for init

Simplify commonly used pool creation.

9 years agodmeventd: new initialization of plugin threads
Zdenek Kabelac [Tue, 13 Oct 2015 07:43:09 +0000 (09:43 +0200)]
dmeventd: new initialization of plugin threads

Rework thread creation code to better use resources.

New code will not leak 'timeout' registered thread on error path.

Also if the thread already exist - avoid creation of thread
object and it's later destruction.

If the race is noticed during adding new monitoring thread,
such thread is put on cleanup list and -EEXIST is reported.

9 years agocleanup: typo in comment
Zdenek Kabelac [Mon, 12 Oct 2015 09:37:04 +0000 (11:37 +0200)]
cleanup: typo in comment

9 years agocleanup: move system defines before structs
Zdenek Kabelac [Mon, 12 Oct 2015 09:35:33 +0000 (11:35 +0200)]
cleanup: move system defines before structs

9 years agocleanup: avoid using private
Zdenek Kabelac [Mon, 12 Oct 2015 09:40:51 +0000 (11:40 +0200)]
cleanup: avoid using private

Switch private to user.

9 years agocleanup: remove multilog
Zdenek Kabelac [Tue, 13 Oct 2015 07:40:39 +0000 (09:40 +0200)]
cleanup: remove multilog

9 years agocleanup: always set nsec
Zdenek Kabelac [Tue, 13 Oct 2015 07:39:18 +0000 (09:39 +0200)]
cleanup: always set nsec

9 years agocleanup: more readable code
Zdenek Kabelac [Fri, 9 Oct 2015 19:37:12 +0000 (21:37 +0200)]
cleanup: more readable code

9 years agodmeventd: support logging on stdout
Zdenek Kabelac [Tue, 13 Oct 2015 13:14:11 +0000 (15:14 +0200)]
dmeventd: support logging on stdout

Add new supported option '-l' log to stdout/stderr.

It has to be paired with '-f' (foreground run).

9 years agocleanup: static with _
Zdenek Kabelac [Thu, 8 Oct 2015 09:59:11 +0000 (11:59 +0200)]
cleanup: static with _

Document ? is printing to stderr.

9 years agodmeventd: handle snapshot overflow
Zdenek Kabelac [Fri, 9 Oct 2015 19:41:50 +0000 (21:41 +0200)]
dmeventd: handle snapshot overflow

When snapshot reports overflow, handle it in the same way as Invalid.
Until better ideas are implemented.

9 years agoman: dmeventd update
Zdenek Kabelac [Mon, 12 Oct 2015 09:35:53 +0000 (11:35 +0200)]
man: dmeventd update

9 years agodmeventd: add more debug messages
Zdenek Kabelac [Sat, 10 Oct 2015 14:36:18 +0000 (16:36 +0200)]
dmeventd: add more debug messages

9 years agodmeventd: drop changing logging
Zdenek Kabelac [Fri, 9 Oct 2015 20:08:01 +0000 (22:08 +0200)]
dmeventd: drop changing logging

As we now use 'unified' logging macro system - we no longer need
to protect from change of logging function pointer - it's set
once at the start of dmeventd and not change anymore
(as lvm2 library no longer interferers here).

9 years agodmeventd: enable new logging
Zdenek Kabelac [Tue, 13 Oct 2015 13:13:28 +0000 (15:13 +0200)]
dmeventd: enable new logging

9 years agodmeventd: enable new logging
Zdenek Kabelac [Sat, 10 Oct 2015 16:32:53 +0000 (18:32 +0200)]
dmeventd: enable new logging

9 years agodmeventd: use log_ macros in plugins
Zdenek Kabelac [Sat, 10 Oct 2015 14:58:31 +0000 (16:58 +0200)]
dmeventd: use log_ macros in plugins

Convert from syslog to standard log_() macros in plugins.

9 years agodmeventd: syslog -> log_xxxxx
Zdenek Kabelac [Fri, 9 Oct 2015 19:57:48 +0000 (21:57 +0200)]
dmeventd: syslog -> log_xxxxx

Convert logging of dmeventd to libdm logging macros.

9 years agolibdmevent: add dm_event_log
Zdenek Kabelac [Sat, 10 Oct 2015 14:40:52 +0000 (16:40 +0200)]
libdmevent: add dm_event_log

Add dm_event_log() logging function.
Controled by dm_event_log_set().

9 years agodmeventd: log all commands with CMD: prefix
Zdenek Kabelac [Tue, 13 Oct 2015 09:19:59 +0000 (11:19 +0200)]
dmeventd: log all commands with CMD: prefix

For easier grep of all processed command - add CMD: prefix to them.

Use  >>>  for entring command processing
and  <<<  when command is processed.

9 years agodmeventd: fix loging report
Zdenek Kabelac [Fri, 9 Oct 2015 20:16:33 +0000 (22:16 +0200)]
dmeventd: fix loging report

msg.cmd holds return value from processing after _do_process_request,
so save it before call.

9 years agolog: no file for external logging
Zdenek Kabelac [Tue, 13 Oct 2015 07:44:33 +0000 (09:44 +0200)]
log: no file for external logging

When external logging is set do not bother creating log file,
as it's not going to be used.

9 years agotoolcontext: do not change already set dm_log
Zdenek Kabelac [Fri, 9 Oct 2015 19:38:50 +0000 (21:38 +0200)]
toolcontext: do not change already set dm_log

If the dm_log is already set to non-default logger,
avoid to change it.

9 years agothin: enhance logging
Zdenek Kabelac [Fri, 9 Oct 2015 19:35:05 +0000 (21:35 +0200)]
thin: enhance logging

Add debug message with more details about threshold overflow.

9 years agodmeventd: correct ret code for fail
Zdenek Kabelac [Tue, 13 Oct 2015 07:49:59 +0000 (09:49 +0200)]
dmeventd: correct ret code for fail

As  _do_register_device() calls  register_device() function that returns
1 success & 0 fail -  ret = -ENOMEM for fail case instead of 0.

9 years agothin: check for overflown pool earlier
Zdenek Kabelac [Wed, 7 Oct 2015 13:10:03 +0000 (15:10 +0200)]
thin: check for overflown pool earlier

Check for pool early before we actually start to modify metadata.
This requires locally active thin-pool.

9 years agolibdm: enforce writethrough mode for cleaner
Zdenek Kabelac [Tue, 13 Oct 2015 12:06:15 +0000 (14:06 +0200)]
libdm: enforce writethrough mode for cleaner

With "cleaner" policy always use 'writethrough' mode.

9 years agowiping: make libblkid detect all copies of the same signature if use_blkid_wiping=1
Peter Rajnoha [Tue, 13 Oct 2015 10:18:04 +0000 (12:18 +0200)]
wiping: make libblkid detect all copies of the same signature if use_blkid_wiping=1

Some signatures are spread around the disk in several copies, mainly for
backup. Make libblkid to detect these extra copies - there was missing
"blkid_probe_step_back" fn call after successful wipe of previous signature
copy.

An example with FAT table which has copies:

$ mkfs.vfat /dev/sda1

Before this patch:

$ pvcreate /dev/sda1
WARNING: vfat signature detected on /dev/sda1 at offset 54. Wipe it? [y/n]: y
  Wiping vfat signature on /dev/sda1.
  Physical volume "/dev/sda1" successfully created

With this patch applied:

$ pvcreate /dev/sda1
WARNING: vfat signature detected on /dev/sda1 at offset 54. Wipe it? [y/n]: y
  Wiping vfat signature on /dev/sda1.
WARNING: vfat signature detected on /dev/sda1 at offset 0. Wipe it? [y/n]: y
  Wiping vfat signature on /dev/sda1.
WARNING: vfat signature detected on /dev/sda1 at offset 510. Wipe it? [y/n]: y
  Wiping vfat signature on /dev/sda1.
  Physical volume "/dev/sda1" successfully created

9 years agotests: test vgimportclone and -n option
Peter Rajnoha [Mon, 12 Oct 2015 10:54:35 +0000 (12:54 +0200)]
tests: test vgimportclone and -n option

9 years agovgimportclone: make sure there's no prefix in VG name list
Peter Rajnoha [Mon, 12 Oct 2015 10:51:13 +0000 (12:51 +0200)]
vgimportclone: make sure there's no prefix in VG name list

Make sure log/prefix is set to "" when getting the list of VG names.
We need this for the format to be correct so it's properly searched
through later on.

9 years agovgimportclone: fix vgimportclone with -n to not add number unnecessarily to base...
Peter Rajnoha [Mon, 12 Oct 2015 09:11:34 +0000 (11:11 +0200)]
vgimportclone: fix vgimportclone with -n to not add number unnecessarily to base VG name

$ vgcreate vgA /dev/sda
  Volume group "vgA" successfully created

$ dd if=/dev/sda of=/dev/sdb bs=1M
$ dd if=/dev/sda of=/dev/sdc bs=1M

(the new VG name is prefix of existing VG name)
$ vgimportclone -n vg /dev/sdb

(the new VG name is suffix of existing VG name)
$ vgimportclone -n gA /dev/sdc

Before this patch:
------------------
(we end up with "vg1" and "gA1" names with the "1" suffix which is not needed)
$ vgs -o vg_name
  VG
  gA1
  vg1
  vgA

With this patch applied:
------------------------
(we end up with "vg" and "gA" names as they're unique already and no extra suffix is added)
$ # vgs -o vg_name
  VG
  gA
  vg
  vgA

Of course, if the name supplied is not unique, the number is added correctly:
$ dd if=/dev/sda of=/dev/sdb bs=1M
$ vgimportclone -n vgA /dev/sdb
$ vgs -o vg_name
  VG
  vgA
  vgA1

9 years agoFix segfault when lvmlockd is running but not lvmetad
David Teigland [Fri, 9 Oct 2015 17:20:22 +0000 (12:20 -0500)]
Fix segfault when lvmlockd is running but not lvmetad

If lvmlockd is running, lvmetad is configured (use_lvmetad=1),
but lvmetad is not running, then commands will seg fault
when trying to send a message to lvmetad.

The difference is lvmetad being "active", not just "used".

9 years agovgimportclone: cleanup the script and remove dependency on awk, grep, cut and tr
Peter Rajnoha [Fri, 9 Oct 2015 14:26:27 +0000 (16:26 +0200)]
vgimportclone: cleanup the script and remove dependency on awk, grep, cut and tr

We can replace the expressions with awk/grep/cut/tr with --select now and
more suitable reporting options and modes. Also, we don't need to check
the temporary lvm.conf generated within vgimportclone script since we're
generating it ourselves now using lvmconfig, not using sed anymore like
it was before (so we can be pretty sure it's correct - we use lvmconfig
now even for generating the lvm.conf itself).

9 years agoreport: add vg_missing_pv_count field
Peter Rajnoha [Fri, 9 Oct 2015 14:20:29 +0000 (16:20 +0200)]
report: add vg_missing_pv_count field

We already have pv_count to report number of PVs that a VG has based
on metadata.

This patch exposes the information about how many of these PVs are
missing which is also useful information for a VG. Wwe could count
the sum of pv_missing reporting fields for each PV in the VG before,
but the new field is practical when reporting VG as a whole and there's
no need to process each PV from VG alone.

9 years agoman lvmsystemid: fix typos
David Teigland [Thu, 8 Oct 2015 18:21:42 +0000 (13:21 -0500)]
man lvmsystemid: fix typos

9 years agoman lvmsystemd: fix typo
David Teigland [Thu, 8 Oct 2015 18:14:28 +0000 (13:14 -0500)]
man lvmsystemd: fix typo

9 years agovgcreate: improve error message for multiple lock managers
David Teigland [Thu, 8 Oct 2015 15:38:35 +0000 (10:38 -0500)]
vgcreate: improve error message for multiple lock managers

If 'vgcreate --shared' finds both sanlock and dlm are running,
print a more accurate error message:
"Found multiple lock managers, select one with --lock-type."

When neither is running, we still print:
"Failed to detect a running lock manager to select lock type."

9 years agovgcreate: allow both --shared and --lock-type options
David Teigland [Thu, 8 Oct 2015 15:14:33 +0000 (10:14 -0500)]
vgcreate: allow both --shared and --lock-type options

Using --lock-type sanlock|dlm implies --shared.

Using --shared selects lock type sanlock|dlm
(by choosing the one that's running.)

Using both --shared and --lock-type sanlock|dlm should
also be allowed (--shared is just redundant information.)

9 years agometadata: format_text: provide more detailed error message when metadata too large...
Peter Rajnoha [Thu, 8 Oct 2015 14:25:10 +0000 (16:25 +0200)]
metadata: format_text: provide more detailed error message when metadata too large for PV mda

Also, leave out the note about "circular buffer" which is
an internal imeplementation detail anyway and not quite
informational for users:

Before this patch:
$ vgcreate vg1 /dev/sda
  VG vg1 metadata too large for circular buffer
  Failed to write VG vg1.

With this patch applied:
$ vgcreate vg1 /dev/sda
  VG vg1 metadata too large: size of metadata to write is 691 bytes while PV metadata area size on /dev/sda is 512 bytes.
  Failed to write VG vg1.

9 years agoreport: identify LV hodling sanlock locks as 'private,lockd,sanlock' within lv_role...
Peter Rajnoha [Thu, 8 Oct 2015 11:44:21 +0000 (13:44 +0200)]
report: identify LV hodling sanlock locks as 'private,lockd,sanlock' within lv_role report field

Before this patch:
$ lvs -a -o name,layout,role test/lvmlock
  LV        Layout     Role
  [lvmlock] linear     public

With this patch applied:
$ lvs -a -o name,layout,role test/lvmlock
  LV        Layout     Role
  [lvmlock] linear     private,lockd,sanlock

9 years agotests: check devices with prefix aren't in use
Zdenek Kabelac [Tue, 6 Oct 2015 12:55:24 +0000 (14:55 +0200)]
tests: check devices with prefix aren't in use

Avoid running tests, when prefix already exist in the system.
As prefix just uses  PID number, we may hit a case for long
running tests, where devices from some previous runs were not
properly cleared away - detect this and fail early.
(Such machine should be inspected and fixed).

9 years agoman: lvcreate continue
Zdenek Kabelac [Tue, 6 Oct 2015 12:55:09 +0000 (14:55 +0200)]
man: lvcreate continue

Finish remaing bits of updating pages for better rendering
with -Thtml, -Tps.

9 years agoman: fsadm
Zdenek Kabelac [Mon, 5 Oct 2015 10:28:00 +0000 (12:28 +0200)]
man: fsadm

9 years agocleanup: fix compiler warning
Zdenek Kabelac [Sat, 3 Oct 2015 20:10:31 +0000 (22:10 +0200)]
cleanup: fix compiler warning

9 years agoblkdeactivate: recognize and deactivate MD devices too
Peter Rajnoha [Tue, 6 Oct 2015 11:25:37 +0000 (13:25 +0200)]
blkdeactivate: recognize and deactivate MD devices too

9 years agolockd: add error message for EEXIST
David Teigland [Fri, 2 Oct 2015 20:41:23 +0000 (15:41 -0500)]
lockd: add error message for EEXIST

The EEXIST error for LV locks is unusual, and
was missing an explanatory error message.

9 years agoraid_manip: fix wrong image size allocation on raid10 "lvconvert --replace ..."
Heinz Mauelshagen [Fri, 2 Oct 2015 15:09:37 +0000 (17:09 +0200)]
raid_manip: fix wrong image size allocation on raid10 "lvconvert --replace ..."

9 years agoraid: Add metadata dev information to reports.
Alasdair G Kergon [Fri, 2 Oct 2015 09:09:28 +0000 (10:09 +0100)]
raid: Add metadata dev information to reports.

Add metadata_devices and seg_metadata_le_ranges report fields.
Currently only defined for raid, but should probably be extended
to all other segment types that don't report all their device
usage in the 'devices' field.

9 years agoman lvmcache: updates for mode and policy
David Teigland [Thu, 1 Oct 2015 16:39:07 +0000 (11:39 -0500)]
man lvmcache: updates for mode and policy

Correct some things, e.g. set mode and policy on
the cache lv, not the pool, lvm.conf field for
mode changed.

Add smq which was missing.

Make the sections on cache mode and cache policy
consistent in structure and style.

9 years agoconfig: improve cache_settings description
David Teigland [Thu, 1 Oct 2015 16:45:07 +0000 (11:45 -0500)]
config: improve cache_settings description

9 years agotests: ensure test will not block
Zdenek Kabelac [Thu, 1 Oct 2015 14:45:14 +0000 (16:45 +0200)]
tests: ensure test will not block

This test is somewhat 'stupid' and simulates mostly unsupported
situation - so avoid blocking here.

9 years agotests: use proper target name for check
Zdenek Kabelac [Thu, 1 Oct 2015 14:05:51 +0000 (16:05 +0200)]
tests: use proper target name for check

9 years agoman: update lvm pages
Zdenek Kabelac [Wed, 23 Sep 2015 09:28:54 +0000 (11:28 +0200)]
man: update lvm pages

9 years agoman: update dmsetup and dmstats pages
Zdenek Kabelac [Wed, 23 Sep 2015 09:25:00 +0000 (11:25 +0200)]
man: update dmsetup and dmstats pages

Try to provide properly rendered pages no just with
plain 'man' but also for:

man -Tps
man -Thtml
man2html

9 years agoalloc: fix update or area_len
Zdenek Kabelac [Thu, 1 Oct 2015 12:59:21 +0000 (14:59 +0200)]
alloc: fix update or area_len

Commit: 192d9ad97769f2c16f3ebe67e51cfb287e67f67c
changed logic for area_len formula - so it returns
different values.

Placing () to restore previous behaviour and make it
explicit.

9 years agotest: Fix timeout handling
Marian Csontos [Thu, 1 Oct 2015 11:17:00 +0000 (13:17 +0200)]
test: Fix timeout handling

Timeouts were considered as warnings only. Signalling failure is
preferred.

9 years agolockd: add VG name to debug logging
David Teigland [Tue, 29 Sep 2015 19:07:08 +0000 (14:07 -0500)]
lockd: add VG name to debug logging

Add the VG name to the new debug logging in
the previous commit.

9 years agolockd: add debug logging for metadata error
David Teigland [Tue, 29 Sep 2015 18:40:52 +0000 (13:40 -0500)]
lockd: add debug logging for metadata error

When lvmetad_pvscan_vg() reads VG metadata from each PV,
it compares it to the last one to verify it matches.
If the VG metadata does not match on the PVs, an error
is printed and it fails to read the VG.  In this error
case, use log_debug to show the differences between
the two unmatching copies of the metadata.

9 years agolockd: fix rescanning VG metadata
David Teigland [Tue, 29 Sep 2015 16:07:59 +0000 (11:07 -0500)]
lockd: fix rescanning VG metadata

One host changes a VG, making the cached VG on another
host invalid.  The other host then rereads the VG from
disk to get the latest copy.  If the first host removed
a PV from the VG, the second host attempts to reread the
VG from old PV when rescanning.  Reading the VG from the
removed PV fails, causing vg_read to return "VG not found".
The fix is to simply not fail when a VG is not found while
rereading a PV and continue without it.

(This doesn't happen if the second host happens to first
run a command like 'vgs' that triggers a global revalidation
of metadata.)

9 years agoraid: Introduce DEFAULT_RAID_MAX_IMAGES
Heinz Mauelshagen [Mon, 28 Sep 2015 20:38:40 +0000 (21:38 +0100)]
raid: Introduce DEFAULT_RAID_MAX_IMAGES

Prepare to allow for the number of images in a raid set to be
larger than the limit for old-style dm raid1.

9 years agoraid: Fail if kernel status denominator is zero.
Alasdair G Kergon [Mon, 28 Sep 2015 19:45:44 +0000 (20:45 +0100)]
raid: Fail if kernel status denominator is zero.

9 years agolibdm: Move status fns from deptree to targets.
Alasdair G Kergon [Mon, 28 Sep 2015 19:28:31 +0000 (20:28 +0100)]
libdm: Move status fns from deptree to targets.

libdm-deptree is only for functions working with dm tree nodes.

9 years agoraid: Use loop for text uint32_t parameter import.
Heinz Mauelshagen [Mon, 28 Sep 2015 13:21:00 +0000 (14:21 +0100)]
raid: Use loop for text uint32_t parameter import.

Plus some other tidying up of the same file.

9 years agotest: Cope with stripe rounding message change.
Alasdair G Kergon [Mon, 28 Sep 2015 11:25:31 +0000 (12:25 +0100)]
test: Cope with stripe rounding message change.

9 years agosegtypes: Use flags in raid segtype macros.
Heinz Mauelshagen [Thu, 24 Sep 2015 19:26:33 +0000 (20:26 +0100)]
segtypes: Use flags in raid segtype macros.

9 years agoraid: Add a segtype flag for each raid type.
Heinz Mauelshagen [Thu, 24 Sep 2015 19:17:57 +0000 (20:17 +0100)]
raid: Add a segtype flag for each raid type.

This page took 0.069904 seconds and 5 git commands to generate.