Migration: Generate the completed event only when we complete
PATCH 1/1 Migration: Generate the completed event only when we complete
From: "Dr. David Alan Gilbert" <[email protected]>
The current migration-completed event is generated a bit too early, which means that an eager libvirt that's ready to go as soon as it sees the event ends up racing with the actual end of migration.
This corresponds to RH bug:
https://bugzilla.redhat.com/show_bug.cgi?id=1271145
--- migration/migration.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)
migration/migration.c +7 -1
@@ -294,12 +294,12 @@ static void process_incoming_migration_co(void *opaque)migrate_decompress_threads_join ();exit (EXIT_FAILURE ); } -migrate_generate_event (MIGRATION_STATUS_COMPLETED );qemu_announce_self ();/* Make sure all file formats flush their mutable metadata */ bdrv_invalidate_cache_all (&local_err);if (local_err) { +migrate_generate_event (MIGRATION_STATUS_FAILED );error_report_err (local_err);migrate_decompress_threads_join ();exit (EXIT_FAILURE );
@@ -320,6 +320,12 @@ static void process_incoming_migration_co(void *opaque)runstate_set (global_state_get_runstate ()); }migrate_decompress_threads_join (); +/* +* This must happen after any state changes since as soon as an external +* observer sees this event they might start to prod at the VM assuming +* it's ready to use. +*/ +migrate_generate_event (MIGRATION_STATUS_COMPLETED ); }void process_incoming_migration (QEMUFile *f)
Signature · 1 line
2.5.0
Quoted 10 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
The current migration-completed event is generated a bit too early, which means that an eager libvirt that's ready to go as soon as it sees the event ends up racing with the actual end of migration.
This corresponds to RH bug:
https://bugzilla.redhat.com/show_bug.cgi?id=1271145Signed-off-by Dr. David Alan Gilbert <[email protected]>
migration/migration.c static void process_incoming_migration_co(void *opaque) · line 297
Quoted 13 lines
migration/migration.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)migration/migration.c +0
-1@@ -294,12 +294,12 @@ static void process_incoming_migration_co(void *opaque)migrate_decompress_threads_join ();exit (EXIT_FAILURE ); } -migrate_generate_event (MIGRATION_STATUS_COMPLETED );qemu_announce_self ();
Related in this context (but independent of this patch): this announce_self should happen after we're through with all the error
cases: we don't want to announce our new IP and end up having run on src due to a failed migration.
migration/migration.c static void process_incoming_migration_co(void *opaque) · line 305
Quoted 7 lines
migration/migration.c +1
-0/* Make sure all file formats flush their mutable metadata */ bdrv_invalidate_cache_all (&local_err);if (local_err) { +migrate_generate_event (MIGRATION_STATUS_FAILED );error_report_err (local_err);migrate_decompress_threads_join ();exit (EXIT_FAILURE );
Amit
Quoted 17 lines · Amit Shah
Quoted 10 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
The current migration-completed event is generated a bit too early, which means that an eager libvirt that's ready to go as soon as it sees the event ends up racing with the actual end of migration.
This corresponds to RH bug:
https://bugzilla.redhat.com/show_bug.cgi?id=1271145Signed-off-by Dr. David Alan Gilbert <[email protected]>Reviewed-by Amit Shah <[email protected]>Quoted 3 lines
migration/migration.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)
Quoted 15 lines
migration/migration.c static void process_incoming_migration_co(void *opaque) · line 297
Quoted 10 lines
migration/migration.c +0
-1@@ -294,12 +294,12 @@ static void process_incoming_migration_co(void *opaque)migrate_decompress_threads_join ();exit (EXIT_FAILURE ); } -migrate_generate_event (MIGRATION_STATUS_COMPLETED );qemu_announce_self ();Related in this context (but independent of this patch): this
announce_selfshould happen after we're through with all the error
cases: we don't want to announce our new IP and end up having run on src due to a failed migration.
Yeap, sounds very reasonable.
Quoted 10 lines
migration/migration.c static void process_incoming_migration_co(void *opaque) · line 305
Quoted 8 lines
migration/migration.c +1
-0/* Make sure all file formats flush their mutable metadata */ bdrv_invalidate_cache_all (&local_err);if (local_err) { +migrate_generate_event (MIGRATION_STATUS_FAILED );error_report_err (local_err);migrate_decompress_threads_join ();exit (EXIT_FAILURE );Amit
Quoted 10 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
The current migration-completed event is generated a bit too early, which means that an eager libvirt that's ready to go as soon as it sees the event ends up racing with the actual end of migration.
This corresponds to RH bug:
https://bugzilla.redhat.com/show_bug.cgi?id=1271145Signed-off-by Dr. David Alan Gilbert <[email protected]>
Applied.
Quoted 35 lines
--- migration/migration.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)migration/migration.c +7
-1@@ -294,12 +294,12 @@ static void process_incoming_migration_co(void *opaque)migrate_decompress_threads_join ();exit (EXIT_FAILURE ); } -migrate_generate_event (MIGRATION_STATUS_COMPLETED );qemu_announce_self ();/* Make sure all file formats flush their mutable metadata */ bdrv_invalidate_cache_all (&local_err);if (local_err) { +migrate_generate_event (MIGRATION_STATUS_FAILED );error_report_err (local_err);migrate_decompress_threads_join ();exit (EXIT_FAILURE );@@ -320,6 +320,12 @@ static void process_incoming_migration_co(void *opaque)runstate_set (global_state_get_runstate ()); }migrate_decompress_threads_join (); +/* +* This must happen after any state changes since as soon as an external +* observer sees this event they might start to prod at the VM assuming +* it's ready to use. +*/ +migrate_generate_event (MIGRATION_STATUS_COMPLETED ); }void process_incoming_migration (QEMUFile *f)