[pve-devel] [RFC qemu-server 24/31] block job: add blockdev mirror

[Fiona Ebner]

With blockdev-mirror, it is possible to change the aio setting on the
fly and this is useful for migrations between storages where one wants
to use io_uring by default and the other doesn't.

Already mock blockdev_mirror in the tests.

Co-developed-by: Alexandre Derumier <alexandre.derumier at groupe-cyllene.com>
Signed-off-by: Fiona Ebner <f.ebner at proxmox.com>
---
 src/PVE/QemuServer/BlockJob.pm            | 162 ++++++++++++++++++++++
 src/PVE/QemuServer/Blockdev.pm            |   2 +-
 src/test/MigrationTest/QemuMigrateMock.pm |   8 ++
 3 files changed, 171 insertions(+), 1 deletion(-)

diff --git a/src/PVE/QemuServer/BlockJob.pm b/src/PVE/QemuServer/BlockJob.pm
index 0013bde6..89cd1312 100644
--- a/src/PVE/QemuServer/BlockJob.pm
+++ b/src/PVE/QemuServer/BlockJob.pm
@@ -4,12 +4,14 @@ use strict;
 use warnings;
 
 use JSON;
+use Storable qw(dclone);
 
 use PVE::Format qw(render_duration render_bytes);
 use PVE::RESTEnvironment qw(log_warn);
 use PVE::Storage;
 
 use PVE::QemuServer::Agent qw(qga_check_running);
+use PVE::QemuServer::Blockdev;
 use PVE::QemuServer::Drive qw(checked_volume_format);
 use PVE::QemuServer::Monitor qw(mon_cmd);
 use PVE::QemuServer::RunState;
@@ -340,6 +342,166 @@ sub qemu_drive_mirror_switch_to_active_mode {
     }
 }
 
+=pod
+
+=head3 blockdev_mirror
+
+    blockdev_mirror($source, $dest, $jobs, $completion, $options)
+
+Mirrors the volume of a running VM specified by C<$source> to destination C<$dest>.
+
+=over
+
+=item C<$source>
+
+The source information consists of:
+
+=over
+
+=item C<< $source->{vmid} >>
+
+The ID of the running VM the source volume belongs to.
+
+=item C<< $source->{drive} >>
+
+The drive configuration of the source volume as currently attached to the VM.
+
+=item C<< $source->{bitmap} >>
+
+(optional) Use incremental mirroring based on the specified bitmap.
+
+=back
+
+=item C<$dest>
+
+The destination information consists of:
+
+=over
+
+=item C<< $dest->{volid} >>
+
+The volume ID of the target volume.
+
+=item C<< $dest->{vmid} >>
+
+(optional) The ID of the VM the target volume belongs to. Defaults to C<< $source->{vmid} >>.
+
+=item C<< $dest->{'zero-initialized'} >>
+
+(optional) True, if the target volume is zero-initialized.
+
+=back
+
+=item C<$jobs>
+
+(optional) Other jobs in the transaction when multiple volumes should be mirrored. All jobs must be
+ready before completion can happen.
+
+=item C<$completion>
+
+Completion mode, default is C<complete>:
+
+=over
+
+=item C<complete>
+
+Wait until all jobs are ready, block-job-complete them (default). This means switching the orignal
+drive to use the new target.
+
+=item C<cancel>
+
+Wait until all jobs are ready, block-job-cancel them. This means not switching the original drive
+to use the new target.
+
+=item C<skip>
+
+Wait until all jobs are ready, return with block jobs in ready state.
+
+=item C<auto>
+
+Wait until all jobs disappear, only use for jobs which complete automatically.
+
+=back
+
+=item C<$options>
+
+Further options:
+
+=over
+
+=item C<< $options->{'guest-agent'} >>
+
+If the guest agent is configured for the VM. It will be used to freeze and thaw the filesystems for
+consistency when the target belongs to a different VM.
+
+=item C<< $options->{'bwlimit'} >>
+
+The bandwidth limit to use for the mirroring operation, in KiB/s.
+
+=back
+
+=back
+
+=cut
+
+sub blockdev_mirror {
+    my ($source, $dest, $jobs, $completion, $options) = @_;
+
+    my $vmid = $source->{vmid};
+
+    my $drive_id = PVE::QemuServer::Drive::get_drive_id($source->{drive});
+    my $device_id = "drive-$drive_id";
+
+    my $storecfg = PVE::Storage::config();
+
+    # Need to replace the format node below the top node.
+    my $source_node_name = PVE::QemuServer::Blockdev::get_node_name(
+        'fmt', $drive_id, $source->{drive}->{file},
+    );
+
+    # Copy original drive config (aio, cache, discard, ...):
+    my $dest_drive = dclone($source->{drive});
+    $dest_drive->{file} = $dest->{volid};
+
+    my $generate_blockdev_opts = {};
+    $generate_blockdev_opts->{'zero-initialized'} = 1 if $dest->{'zero-initialized'};
+
+    # Note that if 'aio' is not explicitly set, i.e. default, it can change if source and target
+    # don't both allow or both not allow 'io_uring' as the default.
+    my $target_drive_blockdev = PVE::QemuServer::Blockdev::generate_drive_blockdev(
+        $storecfg, $dest_drive, $generate_blockdev_opts,
+    );
+    # Top node is the throttle group, must use the file child.
+    my $target_blockdev = $target_drive_blockdev->{file};
+
+    PVE::QemuServer::Monitor::mon_cmd($vmid, 'blockdev-add', $target_blockdev->%*);
+    my $target_nodename = $target_blockdev->{'node-name'};
+
+    $jobs = {} if !$jobs;
+    my $jobid = "mirror-$drive_id";
+    $jobs->{$jobid} = {};
+
+    my $qmp_opts = common_mirror_qmp_options(
+        $device_id, $target_nodename, $source->{bitmap}, $options->{bwlimit},
+    );
+
+    $qmp_opts->{'job-id'} = "$jobid";
+    $qmp_opts->{replaces} = "$source_node_name";
+
+    # if a job already runs for this device we get an error, catch it for cleanup
+    eval { mon_cmd($vmid, "blockdev-mirror", $qmp_opts->%*); };
+    if (my $err = $@) {
+        eval { qemu_blockjobs_cancel($vmid, $jobs) };
+        log_warn("unable to cancel block jobs - $@");
+        eval { mon_cmd($vmid, 'blockdev-del', 'node-name' => $target_nodename) };
+        log_warn("unable to delete blockdev '$target_nodename' - $@");
+        die "error starting blockdev mirrror - $err";
+    }
+    qemu_drive_mirror_monitor(
+        $vmid, $dest->{vmid}, $jobs, $completion, $options->{'guest-agent'}, 'mirror',
+    );
+}
+
 sub mirror {
     my ($source, $dest, $jobs, $completion, $options) = @_;
 
diff --git a/src/PVE/QemuServer/Blockdev.pm b/src/PVE/QemuServer/Blockdev.pm
index 6e6b9245..716a0ac9 100644
--- a/src/PVE/QemuServer/Blockdev.pm
+++ b/src/PVE/QemuServer/Blockdev.pm
@@ -12,7 +12,7 @@ use PVE::Storage;
 
 use PVE::QemuServer::Drive qw(drive_is_cdrom);
 
-my sub get_node_name {
+sub get_node_name {
     my ($type, $drive_id, $volid, $snap) = @_;
 
     my $info = "drive=$drive_id,";
diff --git a/src/test/MigrationTest/QemuMigrateMock.pm b/src/test/MigrationTest/QemuMigrateMock.pm
index 25a4f9b2..c52df84b 100644
--- a/src/test/MigrationTest/QemuMigrateMock.pm
+++ b/src/test/MigrationTest/QemuMigrateMock.pm
@@ -9,6 +9,7 @@ use Test::MockModule;
 use MigrationTest::Shared;
 
 use PVE::API2::Qemu;
+use PVE::QemuServer::Drive;
 use PVE::Storage;
 use PVE::Tools qw(file_set_contents file_get_contents);
 
@@ -167,6 +168,13 @@ $qemu_server_blockjob_module->mock(
 
         common_mirror_mock($vmid, $drive_id);
     },
+    blockdev_mirror => sub {
+        my ($source, $dest, $jobs, $completion, $options) = @_;
+
+        my $drive_id = PVE::QemuServer::Drive::get_drive_id($source->{drive});
+
+        common_mirror_mock($source->{vmid}, $drive_id);
+    },
     qemu_drive_mirror_monitor => sub {
         my ($vmid, $vmiddst, $jobs, $completion, $qga) = @_;
 
-- 
2.47.2