[pve-devel] [PATCH-SERIES v2 qemu-server] Cleanup migration code and improve migration disk cleanup

Fabian Ebner f.ebner at proxmox.com
Mon Apr 19 08:49:32 CEST 2021


Ping for this series.

Patch #13 conflicts with Mira's proposed change for copying the 
conntrack information[0].

Patch #1 and Patch #5 are the only ones *not* conflicting with Fabian's 
remote migration RFC[1].

[0]: https://lists.proxmox.com/pipermail/pve-devel/2021-February/046873.html

[1]: https://lists.proxmox.com/pipermail/pve-devel/2021-April/047600.html

Am 29.01.21 um 16:11 schrieb Fabian Ebner:
> This series intends to make the migration code more readable by
> simplyfing/unifying how we keep track of local volumes and splitting up
> sync_disks into multiple subroutines.
> 
> This is done by keeping more information within the hash of local_volumes we
> obtain in the very beginning and re-using it later. Also a method to filter by
> migration/replication type is introduced, making it possible to get rid of some
> special-case handling when iterating over local volumes.
> 
> Still more should be possible, but the series is getting big, so I stopped here.
> Namely, $self->{local_volumes}, $self->{volume_map} and $self->{target_drive}
> still have some overlap and it might be possible to merge them somehow.
> 
> Also it might make sense to put abstract more things to the common code, one
> example is to make target storage mapping available for containers too.
> 
> 
> Changes from v1:
>      * dropped already applied patches
>      * rebased
>      * prefix commit messages with 'migration' (and 'test')
>      * increased test coverage a little, see patch #1 and patch #13
>      * reordered patches #8+#9 to avoid temporary breakage, see
>        patches #8 and patch #13 (added test-case) for the details
>      * add patch to fix bandwidth limit calculation
>      * add patch to split out run_replication from scan_local_volumes
>      * add patch to improve error handling when block job finishing fails
> 
> 
> Fabian Ebner (13):
>    test: migration: add parse_volume_id calls
>    migration: split sync_disks into two functions
>    migration: avoid re-scanning all volumes
>    migration: split out config_update_local_disksizes from
>      scan_local_volumes
>    migration: fix calculation of bandwith limit for non-disk migration
>    migration: save targetstorage and bwlimit in local_volumes hash and
>      re-use information
>    migration: add nbd migrated volumes to volume_map earlier
>    migration: simplify removal of local volumes and get rid of
>      self->{volumes}
>    migration: cleanup_remotedisks: simplify and include more disks
>    migration: use storage_migration for checks instead of
>      online_local_volumes
>    migration: keep track of replicated volumes via local_volumes
>    migration: split out replication from scan_local_volumes
>    migration: move finishing block jobs to phase2 for better/uniform
>      error handling
> 
>   PVE/QemuMigrate.pm                    | 401 ++++++++++++++------------
>   PVE/QemuServer.pm                     |   2 +
>   test/MigrationTest/QemuMigrateMock.pm |   9 +
>   test/MigrationTest/QmMock.pm          |   2 +
>   test/MigrationTest/Shared.pm          |   4 +
>   test/run_qemu_migrate_tests.pl        |  40 ++-
>   6 files changed, 267 insertions(+), 191 deletions(-)
> 




More information about the pve-devel mailing list