In my first approach in 4b480d10768c I overlooked the comment in
qemuMigrationRunIncoming stating that during actual migration the
qemuMigrationRunIncoming does not wait until the migration is complete
but rather offloads that to the Finish phase of migration.

This means that during actual migration qemuProcessRefreshState was
called prior to qemu actually transferring the full state and thus the
queries did not get the correct information. The approach worked only
for restore, where we wait for the migration to finish during qemu
startup.

Fix the issue by calling qemuProcessRefreshState both from
qemuProcessStart if there's no incomming migration and from
qemuMigrationFinish so that the code actually works as expected.
---
 src/qemu/qemu_migration.c |  9 +++++++++
 src/qemu/qemu_process.c   | 17 ++++++++++-------
 src/qemu/qemu_process.h   |  4 ++++
 3 files changed, 23 insertions(+), 7 deletions(-)

diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c
index 1854900c9a..ea8b275601 100644
--- a/src/qemu/qemu_migration.c
+++ b/src/qemu/qemu_migration.c
@@ -5373,6 +5373,15 @@ qemuMigrationFinish(virQEMUDriverPtr driver,
             goto endjob;
     }

+    /* Now that the state data was transferred we can refresh the actual state
+     * of the devices */
+    if (qemuProcessRefreshState(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
+        /* Similarly to the case above v2 protocol will not be able to recover
+         * from this. Let's ignore this and perhaps stuff will not break. */
+        if (v3proto)
+            goto endjob;
+    }
+
     if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY)
         inPostCopy = true;

diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c
index bcd4ac8ad6..fe12d94d37 100644
--- a/src/qemu/qemu_process.c
+++ b/src/qemu/qemu_process.c
@@ -6084,7 +6084,7 @@ qemuProcessLaunch(virConnectPtr conn,
  * function is called after a deferred migration finishes so that we can update
  * state influenced by the migration stream.
  */
-static int
+int
 qemuProcessRefreshState(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob)
@@ -6125,9 +6125,6 @@ qemuProcessFinishStartup(virConnectPtr conn,
     virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
     int ret = -1;

-    if (qemuProcessRefreshState(driver, vm, asyncJob) < 0)
-        goto cleanup;
-
     if (startCPUs) {
         VIR_DEBUG("Starting domain CPUs");
         if (qemuProcessStartCPUs(driver, vm, conn,
@@ -6231,11 +6228,17 @@ qemuProcessStart(virConnectPtr conn,
                                  VIR_DOMAIN_PAUSED_USER) < 0)
         goto stop;

-    /* Keep watching qemu log for errors during incoming migration, otherwise
-     * unset reporting errors from qemu log. */
-    if (!incoming)
+    if (!incoming) {
+        /* Keep watching qemu log for errors during incoming migration, otherwise
+         * unset reporting errors from qemu log. */
         qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);

+        /* Refresh state of devices from qemu. During migration this needs to
+         * happen after the state information is fully transferred. */
+        if (qemuProcessRefreshState(driver, vm, asyncJob) < 0)
+            goto stop;
+    }
+
     ret = 0;

  cleanup:
diff --git a/src/qemu/qemu_process.h b/src/qemu/qemu_process.h
index cd9a720313..b383ff309b 100644
--- a/src/qemu/qemu_process.h
+++ b/src/qemu/qemu_process.h
@@ -129,6 +129,10 @@ int qemuProcessFinishStartup(virConnectPtr conn,
                              bool startCPUs,
                              virDomainPausedReason pausedReason);

+int qemuProcessRefreshState(virQEMUDriverPtr driver,
+                            virDomainObjPtr vm,
+                            qemuDomainAsyncJob asyncJob);
+
 typedef enum {
     VIR_QEMU_PROCESS_STOP_MIGRATED      = 1 << 0,
     VIR_QEMU_PROCESS_STOP_NO_RELABEL    = 1 << 1,
-- 
2.15.0

--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list

On Thu, Feb 01, 2018 at 15:18:59 +0100, Peter Krempa wrote:
> In my first approach in 4b480d10768c I overlooked the comment in
> qemuMigrationRunIncoming stating that during actual migration the
> qemuMigrationRunIncoming does not wait until the migration is complete
> but rather offloads that to the Finish phase of migration.
> 
> This means that during actual migration qemuProcessRefreshState was
> called prior to qemu actually transferring the full state and thus the
> queries did not get the correct information. The approach worked only
> for restore, where we wait for the migration to finish during qemu
> startup.
> 
> Fix the issue by calling qemuProcessRefreshState both from
> qemuProcessStart if there's no incomming migration and from
> qemuMigrationFinish so that the code actually works as expected.
> ---
>  src/qemu/qemu_migration.c |  9 +++++++++
>  src/qemu/qemu_process.c   | 17 ++++++++++-------
>  src/qemu/qemu_process.h   |  4 ++++
>  3 files changed, 23 insertions(+), 7 deletions(-)

ACK

Jirka

--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list