]> git.proxmox.com Git - qemu-server.git/log
qemu-server.git
6 years agoAdd a migration lock to avoid a replication on rollback-time.
Wolfgang Link [Mon, 12 Jun 2017 08:38:22 +0000 (10:38 +0200)]
Add a migration lock to avoid a replication on rollback-time.

6 years agobuild: remove fakeroot from dpkg-buildpackage
Fabian Grünbichler [Mon, 12 Jun 2017 08:00:25 +0000 (10:00 +0200)]
build: remove fakeroot from dpkg-buildpackage

6 years agofix typo in previous changelog
Fabian Grünbichler [Mon, 12 Jun 2017 07:50:22 +0000 (09:50 +0200)]
fix typo in previous changelog

6 years agobuild: add substition variables
Fabian Grünbichler [Mon, 12 Jun 2017 07:49:21 +0000 (09:49 +0200)]
build: add substition variables

6 years agobuild: reformat b-d and depends
Fabian Grünbichler [Mon, 12 Jun 2017 07:47:01 +0000 (09:47 +0200)]
build: reformat b-d and depends

6 years agobuild: make control static
Fabian Grünbichler [Mon, 12 Jun 2017 07:44:04 +0000 (09:44 +0200)]
build: make control static

6 years agochange from dpkg-deb to dpkg-buildpackage
Thomas Lamprecht [Fri, 9 Jun 2017 16:37:03 +0000 (18:37 +0200)]
change from dpkg-deb to dpkg-buildpackage

Add debian directory and move the respective files there and add
missing (rules, compat, docs).

Add a Source section to the control.in file and "build" (sed
control.in >control) in the $DEB target as done in most other
packages

Move the qm verifyapi check to the new "test" target, which gets
executed before the dh_auto_install target.

Remove the obsolete vzsyscalls.ph target, vzsyscalls.h is gone for a
while now

Cleanup "clean" target.

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
6 years agoPVE::QemuConfig - implement $cleanup flag for get_replicatable_volumes
Dietmar Maurer [Mon, 12 Jun 2017 07:09:53 +0000 (09:09 +0200)]
PVE::QemuConfig - implement $cleanup flag for get_replicatable_volumes

6 years agofix #1405: sort pci ids by functions
Wolfgang Bumiller [Thu, 8 Jun 2017 07:18:25 +0000 (09:18 +0200)]
fix #1405: sort pci ids by functions

QemuServer::lspci() iterates over /sys/bus/pci/devices which
doesn't guarantee any order which means functions sometimes
ended up in the wrong order and it was never clear which
one would get the additional options such as x-vga passed
to them.

6 years agomigrate: migration_type setting moved to pve-guest-common
Wolfgang Bumiller [Tue, 6 Jun 2017 08:03:59 +0000 (10:03 +0200)]
migrate: migration_type setting moved to pve-guest-common

6 years agoimport disk: add \n to die
Fabian Grünbichler [Thu, 1 Jun 2017 09:40:06 +0000 (11:40 +0200)]
import disk: add \n to die

6 years agoimport disk: downgrade hotplug failure to warning
Fabian Grünbichler [Thu, 1 Jun 2017 09:15:14 +0000 (11:15 +0200)]
import disk: downgrade hotplug failure to warning

6 years agoAdd new qm command 'importdisk' to import external disk images
Emmanuel Kasper [Thu, 1 Jun 2017 08:26:39 +0000 (10:26 +0200)]
Add new qm command 'importdisk' to import external disk images

The VM must be already existing, and the syntax is

qm importdisk 421 minix204.img pve4tank

where 421 is an already existing VM

6 years agoAdd ImportDisk module to import external disk images into a VM
Emmanuel Kasper [Thu, 1 Jun 2017 08:26:38 +0000 (10:26 +0200)]
Add ImportDisk module to import external disk images into a VM

Each disk passed as paramater is add as 'unused[n]' in the vm.conf
(the default) or ide[n]|scsi[n]|sata[n]
We rely on qemu-img(1) convert heuristics to detect the file type,
as this works in most case.

6 years agoRefactor the resolution of target disk format in own sub
Emmanuel Kasper [Thu, 1 Jun 2017 08:26:37 +0000 (10:26 +0200)]
Refactor the resolution of target disk format in own sub

This will allow use to reuse the code in ImportDisk.pm.

6 years agomigration: implement insecure offline migration
Wolfgang Bumiller [Tue, 30 May 2017 13:30:13 +0000 (15:30 +0200)]
migration: implement insecure offline migration

6 years agocheck also pending changes when reverting/deleting
Dominik Csapak [Mon, 29 May 2017 08:37:23 +0000 (10:37 +0200)]
check also pending changes when reverting/deleting

otherwise we are not able to revert/delete pending changes which
introduce a new config line

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
6 years agobump version to 5.0-6
Dietmar Maurer [Wed, 31 May 2017 07:21:50 +0000 (09:21 +0200)]
bump version to 5.0-6

6 years agofix burst length parameter names and pass them to qemu
Wolfgang Bumiller [Tue, 30 May 2017 13:30:15 +0000 (15:30 +0200)]
fix burst length parameter names and pass them to qemu

bps_max_length & friends were wrongly named and were only
passed to qemu when hot-applying changes. They can only
be passed via the command line with their new names. For
consistency let's rename them all, that way they're all in
one place.

Fixes #1195 (for real this time).

6 years agofix #1229: more explicit spice port allocation
Wolfgang Bumiller [Tue, 30 May 2017 13:30:14 +0000 (15:30 +0200)]
fix #1229: more explicit spice port allocation

We cannot look for ports on "any" wildcard address while
letting qemu bind to "localhost", this may lead to a qemu
process occupying ::1 while the next search successfully
finds the same port available for IPv4's '*' address.

Instead, we now lookup the IP of the desired family for
'localhost'. Note that while we could simply be hardcoding
::1 or 127.0.0.1, with this code we are protocol agnostic.

6 years agomigrate: aquire guest_migration_lock during migration
Dietmar Maurer [Wed, 24 May 2017 07:20:06 +0000 (09:20 +0200)]
migrate: aquire guest_migration_lock during migration

To block replication jobs.

6 years agodo not allow destroy if there are replication jobs
Dietmar Maurer [Tue, 23 May 2017 16:01:13 +0000 (18:01 +0200)]
do not allow destroy if there are replication jobs

6 years agoremove 'replicate' configuration - we will store that in another file
Dietmar Maurer [Tue, 9 May 2017 14:45:52 +0000 (16:45 +0200)]
remove 'replicate' configuration - we will store that in another file

6 years agoimprove error on '{full, linked} clone not available' error
Thomas Lamprecht [Tue, 23 May 2017 08:39:45 +0000 (10:39 +0200)]
improve error on '{full, linked} clone not available' error

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
6 years agoprint_cpu_device: remove unused variable
Fabian Grünbichler [Tue, 23 May 2017 12:23:52 +0000 (14:23 +0200)]
print_cpu_device: remove unused variable

6 years agoFix #1384: add missing decrement to calculation of socket-id
Tobias Böhm [Mon, 15 May 2017 14:16:12 +0000 (16:16 +0200)]
Fix #1384: add missing decrement to calculation of socket-id

For calculation of "current_core" the input variable id is decremented.
For calculation of "current_socket" this decrement was missing resulting
in a wrong value when "cores" is set to 1.

Signed-off-by: Tobias Böhm <tb@robhost.de>
6 years agomigrate: pass ssh_info to storage_migrate
Wolfgang Bumiller [Tue, 23 May 2017 07:04:46 +0000 (09:04 +0200)]
migrate: pass ssh_info to storage_migrate

6 years agobuild: run tests on build
Fabian Grünbichler [Wed, 17 May 2017 10:40:58 +0000 (12:40 +0200)]
build: run tests on build

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
6 years agotests: fix broken snapshot create tests
Fabian Grünbichler [Wed, 17 May 2017 10:40:57 +0000 (12:40 +0200)]
tests: fix broken snapshot create tests

by mocking de/activate vmstate volumes. these were broken by
3a8deb551f2c7648865437961b3593d11bf0591b, which introduced
activating/deactivating vmstate volumes for KRBD

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
6 years agotests: fix broken snapshot delete tests
Fabian Grünbichler [Wed, 17 May 2017 10:40:56 +0000 (12:40 +0200)]
tests: fix broken snapshot delete tests

these broke when the additional "snapshot-delete" lock was
introduced

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
6 years agotests: exit with -1 in case of failures
Fabian Grünbichler [Wed, 17 May 2017 10:40:55 +0000 (12:40 +0200)]
tests: exit with -1 in case of failures

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
6 years agouse ReuseAddr for vncproxy
Dominik Csapak [Tue, 16 May 2017 14:21:01 +0000 (16:21 +0200)]
use ReuseAddr for vncproxy

to avoid "Socket already in use" errors, when quickly opening/closing
console sessions on the webinterface

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
6 years agoFix #1361: create disk: stricter parsing of storage:size
Wolfgang Bumiller [Mon, 24 Apr 2017 09:12:29 +0000 (11:12 +0200)]
Fix #1361: create disk: stricter parsing of storage:size

This fixes paths such as
/dev/disk/by-id/some-ID:0.5
being interpreted as a request to create a 0.5G-sized disk
in the storage named '/dev/disk/by-id/some-ID'.

6 years agomove get_replicatable_volumes from QemuServer.pm to QemuConfig.pm
Dietmar Maurer [Sat, 6 May 2017 15:13:31 +0000 (17:13 +0200)]
move get_replicatable_volumes from QemuServer.pm to QemuConfig.pm

6 years agoPVE::QemuServer::create_disks - run code inside eval
Dietmar Maurer [Sat, 6 May 2017 15:00:20 +0000 (17:00 +0200)]
PVE::QemuServer::create_disks - run code inside eval

So that we can free allocated volumes ins something fails.

Thi bug was introduced with commit ae57f6b352242fed69da851f15f976095d290833

6 years agoget_replicatable_volumes: replacemnet for PVE::ReplicationTools::get_syncable_guestdi...
Dietmar Maurer [Sat, 6 May 2017 13:11:46 +0000 (15:11 +0200)]
get_replicatable_volumes: replacemnet for PVE::ReplicationTools::get_syncable_guestdisks()

correctly scan for volumes inside snapshots.

This will also solve the cyclic package dependency propblem.

6 years agorename drive 'replica' property to 'replicate'
Dietmar Maurer [Sat, 6 May 2017 08:46:21 +0000 (10:46 +0200)]
rename drive 'replica' property to 'replicate'

6 years agoRevert "Integrate replica in the qemu migration."
Dietmar Maurer [Sat, 6 May 2017 08:39:43 +0000 (10:39 +0200)]
Revert "Integrate replica in the qemu migration."

This reverts commit 63d02c707429efcd509ddb14d3fd560cfcf4979e.

The commit changes the configuration before the VM is actually
migrated, so it is possible to have a wrong configuration when
migration fails for some reason. Also, I am quite unsure if
this automatic target change is really wanted. The patch also
contains wrong refereces to $self->{opts}->{node}.

6 years agoRevert "migrate: cleanup replica volume skip condition"
Dietmar Maurer [Sat, 6 May 2017 08:38:06 +0000 (10:38 +0200)]
Revert "migrate: cleanup replica volume skip condition"

This reverts commit 6e8044dcea4187d076f637b9e5d3e153d080087f.

6 years agouse a single replicate property (format pve-replicate)
Dietmar Maurer [Sat, 6 May 2017 08:35:08 +0000 (10:35 +0200)]
use a single replicate property (format pve-replicate)

7 years agoadd description of read/writes statistics in vzdump output
Emmanuel Kasper [Tue, 2 May 2017 09:26:53 +0000 (11:26 +0200)]
add description of read/writes statistics in vzdump output

since this statistics are note really described, sometimes users
miss that their slow backups are caused by slow reads

7 years agostart vncproxy worker in the background
Dominik Csapak [Mon, 23 Jan 2017 12:51:21 +0000 (13:51 +0100)]
start vncproxy worker in the background

this allows it to use the vncproxy api call to use from pvesh,
because before we waited until the task finished to get the vncticket
which was too late

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
7 years agobump version to 5.0-5
Wolfgang Bumiller [Fri, 28 Apr 2017 12:02:38 +0000 (14:02 +0200)]
bump version to 5.0-5

7 years agoapi: replica: safer config update
Wolfgang Bumiller [Fri, 28 Apr 2017 08:40:22 +0000 (10:40 +0200)]
api: replica: safer config update

Configuration updates should only happen if and after their
change has successfully taken effect.

7 years agomigrate: cleanup replica volume skip condition
Wolfgang Bumiller [Fri, 28 Apr 2017 08:34:46 +0000 (10:34 +0200)]
migrate: cleanup replica volume skip condition

7 years agowhitespace fixup
Wolfgang Bumiller [Fri, 28 Apr 2017 08:14:26 +0000 (10:14 +0200)]
whitespace fixup

7 years agoDestroy all remote and local replication datasets when a VM will destroyed.
Wolfgang Link [Mon, 24 Apr 2017 15:15:42 +0000 (17:15 +0200)]
Destroy all remote and local replication datasets when a VM will destroyed.

7 years agoIntegrate replica in the qemu migration.
Wolfgang Link [Mon, 24 Apr 2017 15:15:41 +0000 (17:15 +0200)]
Integrate replica in the qemu migration.

Now it is possible to migrate a VM offline when replica is enabled.
It will reduce replication to an minimal amount.

7 years agoInsert new properties in the Qemu config for the PVE Replica.
Wolfgang Link [Mon, 24 Apr 2017 15:15:40 +0000 (17:15 +0200)]
Insert new properties in the Qemu config for the PVE Replica.

This patch will include all necessary properties for the replication.
Also will it enable and disable a replication job
when appointed flags are set or deleted.

7 years agoadd rerror for all drives
Alexandre Derumier [Fri, 28 Apr 2017 07:30:55 +0000 (09:30 +0200)]
add rerror for all drives

it was missing for scsi, and it's supported for all kind of drive

Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
7 years agobump version to 5.0-4
Wolfgang Bumiller [Fri, 21 Apr 2017 10:06:34 +0000 (12:06 +0200)]
bump version to 5.0-4

7 years agolive storage migration : fix check of target storage availability
Alexandre Derumier [Fri, 21 Apr 2017 01:39:34 +0000 (03:39 +0200)]
live storage migration : fix check of target storage availability

if we define a different target storeid for remote node,
and that storage is not available on source node

Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
7 years agobump version to 5.0-3
Wolfgang Bumiller [Fri, 21 Apr 2017 09:49:32 +0000 (11:49 +0200)]
bump version to 5.0-3

7 years agomigrate: clarify comment
Fabian Grünbichler [Wed, 19 Apr 2017 13:19:50 +0000 (15:19 +0200)]
migrate: clarify comment

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 years agomigrate: cleanup nbd source disks earlier
Fabian Grünbichler [Wed, 19 Apr 2017 13:19:49 +0000 (15:19 +0200)]
migrate: cleanup nbd source disks earlier

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 years agofix #1338: migrate: stop nbd before resuming
Fabian Grünbichler [Wed, 19 Apr 2017 13:19:48 +0000 (15:19 +0200)]
fix #1338: migrate: stop nbd before resuming

since Qemu 2.9, block device write access is limited to one
writer unless shared_rw is set to true. there is an
exception for live-migrating local disks via NBD as long as
the VM is suspended.

stop the NBD server before resuming the VM accordingly to
unbreak local disk live-migration.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 years agostart: keep VM paused when migrating via unix socket
Fabian Grünbichler [Wed, 19 Apr 2017 13:19:47 +0000 (15:19 +0200)]
start: keep VM paused when migrating via unix socket

this (correctly!) errored out with Qemu 2.9 when live-migrating
local disks, because the NBD server blocks the VM from being
resumed. was probably missed when migrating via unix domains
was originally introduced..

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 years agoturn off alarm timer after accepting a vnc connection
Dominik Csapak [Tue, 18 Apr 2017 07:24:40 +0000 (09:24 +0200)]
turn off alarm timer after accepting a vnc connection

else we always get a failed task when the vnc connection is
running longer than $timeout

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
7 years agobump version to 5.0-2
Fabian Grünbichler [Mon, 10 Apr 2017 14:28:46 +0000 (16:28 +0200)]
bump version to 5.0-2

7 years agodon't compare booleans
Wolfgang Bumiller [Mon, 3 Apr 2017 12:08:19 +0000 (14:08 +0200)]
don't compare booleans

a) because perl isn't THAT bad of a language that you'd need
to do this and b) because JSON::XS version 2.90 "dropped"
this "feature"

7 years agoupdate_vm: check whether opt is set before deleting
Fabian Grünbichler [Fri, 17 Mar 2017 10:01:32 +0000 (11:01 +0100)]
update_vm: check whether opt is set before deleting

otherwise we get lots uninitialized warnings:

update VM 600: -delete unused7
Use of uninitialized value $data in split at /usr/share/perl5/PVE/JSONSchema.pm line 533.
Use of uninitialized value in concatenation (.) or string at /usr/share/perl5/PVE/API2/Qemu.pm line 1012.
Use of uninitialized value $volid in pattern match (m//) at /usr/share/perl5/PVE/QemuServer.pm line 1824.
Use of uninitialized value $volid in pattern match (m//) at /usr/share/perl5/PVE/Storage/Plugin.pm line 201.
Use of uninitialized value $volid in concatenation (.) or string at /usr/share/perl5/PVE/Storage/Plugin.pm line 205.

vs:

update VM 600: -delete unused7
cannot delete 'unused7' - not set in current configuration!

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 years agobump version to 5.0-1
Wolfgang Bumiller [Mon, 13 Mar 2017 11:00:25 +0000 (12:00 +0100)]
bump version to 5.0-1

7 years agobump version to 4.0-110
Wolfgang Bumiller [Mon, 13 Mar 2017 10:57:13 +0000 (11:57 +0100)]
bump version to 4.0-110

7 years agoUse backticks instead of single quotes in qm resize description
Emmanuel Kasper [Wed, 1 Mar 2017 09:31:54 +0000 (10:31 +0100)]
Use backticks instead of single quotes in qm resize description

Single quoted text in asciidoc is rendered in man pages
with underlines, which makes the '+' symbol very similar to '+/-'

Backticks are rendered with monospace text in HTML, normal text
in man pages, and still readable in raw format.

7 years agodefined() style cleanup
Fabian Grünbichler [Tue, 28 Feb 2017 11:46:45 +0000 (12:46 +0100)]
defined() style cleanup

7 years agoCheck array existed before use.
Wolfgang Link [Thu, 23 Feb 2017 13:44:54 +0000 (14:44 +0100)]
Check array existed before use.

This triggers if a qemu guest has a local unused disk.
The disk will migrate by offline disk migration, so it is not in the target_drives.

7 years agoremove netcat6 dependency
Wolfgang Bumiller [Tue, 7 Feb 2017 11:18:34 +0000 (12:18 +0100)]
remove netcat6 dependency

By simply using a socket.

7 years agoadd Skylake to CPU Models
Dominik Csapak [Fri, 17 Feb 2017 14:54:06 +0000 (15:54 +0100)]
add Skylake to CPU Models

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
7 years agoImprove error message when trying to shrink a volume, fix typo
Emmanuel Kasper [Thu, 16 Feb 2017 09:39:35 +0000 (10:39 +0100)]
Improve error message when trying to shrink a volume, fix typo

7 years agofix bad scsihw default value check
Wolfgang Bumiller [Mon, 13 Feb 2017 11:00:26 +0000 (12:00 +0100)]
fix bad scsihw default value check

This was checking for scsihw being set in both branches
whereas lsi is also the default. Added the missing 'not'.

Fixes a bug where a VM with a disk with a scsi index >= 7
refused to start due to an invalid scsi id.

Reported-by: Friedrich Ramberger <f.ramberger@proxmox.com>
7 years agofix dependency
Wolfgang Bumiller [Thu, 9 Feb 2017 10:48:03 +0000 (11:48 +0100)]
fix dependency

7 years agobump version to 4.0-109
Wolfgang Bumiller [Thu, 9 Feb 2017 10:41:01 +0000 (11:41 +0100)]
bump version to 4.0-109

7 years agoskiplock standard option was moved to pve-common
Wolfgang Bumiller [Tue, 7 Feb 2017 13:55:48 +0000 (14:55 +0100)]
skiplock standard option was moved to pve-common

this now need libpve-common-perl >= 4.0-92

7 years agodon't pull HA modules for doc-generation
Wolfgang Bumiller [Mon, 6 Feb 2017 15:13:30 +0000 (16:13 +0100)]
don't pull HA modules for doc-generation

7 years agobuildsys: use fakeroot for dpkg-deb
Wolfgang Bumiller [Mon, 6 Feb 2017 15:13:08 +0000 (16:13 +0100)]
buildsys: use fakeroot for dpkg-deb

7 years agobuildsys: use gzip -n to disable timestamps
Wolfgang Bumiller [Mon, 6 Feb 2017 15:12:48 +0000 (16:12 +0100)]
buildsys: use gzip -n to disable timestamps

7 years agobuildsys: set PVE_GENERATING_DOCS for completions
Wolfgang Bumiller [Mon, 6 Feb 2017 15:12:18 +0000 (16:12 +0100)]
buildsys: set PVE_GENERATING_DOCS for completions

7 years agobump version to 4.0-108
Wolfgang Bumiller [Tue, 31 Jan 2017 12:54:56 +0000 (13:54 +0100)]
bump version to 4.0-108

7 years agobuildsys: job-safe deb target
Wolfgang Bumiller [Tue, 31 Jan 2017 12:39:17 +0000 (13:39 +0100)]
buildsys: job-safe deb target

7 years agochange TLS cipher suite to HIGH
Fabian Grünbichler [Wed, 11 Jan 2017 14:51:42 +0000 (15:51 +0100)]
change TLS cipher suite to HIGH

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 years agofix #1260: convert moved template disk to base
Fabian Grünbichler [Wed, 25 Jan 2017 13:58:14 +0000 (14:58 +0100)]
fix #1260: convert moved template disk to base

this was already possible manually via "qm template", but
doing it automatically when moving a disk of a template
makes more sense.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 years agobump version to 4.0-107
Fabian Grünbichler [Wed, 25 Jan 2017 08:48:55 +0000 (09:48 +0100)]
bump version to 4.0-107

7 years agoadd dependencies for common packages
Fabian Grünbichler [Wed, 25 Jan 2017 08:46:27 +0000 (09:46 +0100)]
add dependencies for common packages

libpve-guest-common-perl
libpve-common-perl

7 years agobump version to 4.0-106
Dietmar Maurer [Wed, 25 Jan 2017 08:19:33 +0000 (09:19 +0100)]
bump version to 4.0-106

7 years agoonly use scsi-block with explicit opt-in
Fabian Grünbichler [Tue, 24 Jan 2017 10:25:52 +0000 (11:25 +0100)]
only use scsi-block with explicit opt-in

since it can cause I/O errors and data corruption in low
memory or highly fragmented memory situations since Qemu 2.7

use scsi-hd by default instead

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 years agobump version to 4.0-105
Dietmar Maurer [Thu, 19 Jan 2017 08:20:30 +0000 (09:20 +0100)]
bump version to 4.0-105

7 years agouse new PVE::Storage::check_volume_access()
Dietmar Maurer [Wed, 18 Jan 2017 16:22:34 +0000 (17:22 +0100)]
use new PVE::Storage::check_volume_access()

7 years agoSet zero $size and continue if volume_resize() returns false
Dmitry Petuhov [Mon, 16 Jan 2017 07:45:10 +0000 (10:45 +0300)]
Set zero $size and continue if volume_resize() returns false

Needed for some storages with backing block devices to do oonlinr resize.

Signed-off-by: Dmitry Petuhov <mityapetuhov@gmail.com>
7 years agobump version to 4.0-104
Dietmar Maurer [Thu, 12 Jan 2017 13:15:27 +0000 (14:15 +0100)]
bump version to 4.0-104

7 years agoadd setup_environment hook to CLIHandler classes
Dietmar Maurer [Thu, 12 Jan 2017 13:10:24 +0000 (14:10 +0100)]
add setup_environment hook to CLIHandler classes

7 years agocleanup: drop superfluous condition in assignment
Wolfgang Bumiller [Mon, 9 Jan 2017 14:33:10 +0000 (15:33 +0100)]
cleanup: drop superfluous condition in assignment

Its use later on already uses itself as condition.

7 years agoadd romfile option to hostpci
Alexandre Derumier [Mon, 9 Jan 2017 13:36:24 +0000 (14:36 +0100)]
add romfile option to hostpci

Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
7 years agoadd with-local-disks option for live storage migration
Alexandre Derumier [Fri, 6 Jan 2017 09:15:07 +0000 (10:15 +0100)]
add with-local-disks option for live storage migration

As Fabian as required,
add an extra flag "with-local-disks"  to enable live storage migration with localdisk.

default target storage is same sid than source, this can be overrided with
"targetstorage" option.

I will try improve this later, with optionnal mapping, disk by disk.

Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
7 years agocleanup: error messages
Wolfgang Bumiller [Thu, 5 Jan 2017 08:54:07 +0000 (09:54 +0100)]
cleanup: error messages

7 years agodrive-mirror: bump timeout to 5s, add 30s inactivity timeout
Wolfgang Bumiller [Thu, 5 Jan 2017 08:40:44 +0000 (09:40 +0100)]
drive-mirror: bump timeout to 5s, add 30s inactivity timeout

7 years agodrive-mirror: warn use POSIX::_exit on exec failure
Wolfgang Bumiller [Thu, 5 Jan 2017 08:39:13 +0000 (09:39 +0100)]
drive-mirror: warn use POSIX::_exit on exec failure

7 years agocleanup: whitespaces & style
Wolfgang Bumiller [Thu, 5 Jan 2017 08:16:28 +0000 (09:16 +0100)]
cleanup: whitespaces & style

7 years agolive clone_vm : suspend or freezefs before block-job-cancel
Alexandre Derumier [Tue, 3 Jan 2017 14:03:19 +0000 (15:03 +0100)]
live clone_vm : suspend or freezefs before block-job-cancel

if qga is enabled, we try to freeze the fs before cancelling block job.
if not , we pause the vm.

Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
7 years agoadd socat and unix socket for storage migration
Alexandre Derumier [Tue, 3 Jan 2017 14:03:18 +0000 (15:03 +0100)]
add socat and unix socket for storage migration

This is a workaround for nbd infinite timeout connect

Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
7 years agoadd live storage migration with vm migration
Alexandre Derumier [Tue, 3 Jan 2017 14:03:17 +0000 (15:03 +0100)]
add live storage migration with vm migration

This allow to migrate disks on local storage  to a remote node storage.

When the target node start, a new volumes are created and exposed through qemu embedded nbd server.

qemu drive-mirror is launch on source vm for each disk with nbd server as target.

when drive-mirror reach 100% of 1 disk, we don't complete the block jobs and begin mirror of next disk.
(mirroring are parralel, but we try to mirroring them 1 by 1 to avoid storage && network overload)

Then we live migrate the vm to destination node. (drive-mirror still occur at the same time).

We the vm is livemigrate (source vm paused, target vm pause), we complete the block jobs mirror.

When is done we stop the source vm and resume the target vm

Signed-off-by: Alexandre Derumier <aderumier@odiso.com>