[v5,2/6] remus: don't call stream_continue() when doing failover

Commit Message

Wen Congyang Jan. 19, 2016, 7:17 a.m. UTC

stream_continue() is used for migration to read emulator
xenstore data and emulator context. For remus, if we do
failover, we have read it in the checkpoint cycle, and
we only need to complete the stream.

Signed-off-by: Wen Congyang <[e-mail]>
Reviewed-by: Andrew Cooper <[e-mail]>
---
 tools/libxl/libxl_stream_read.c | 36 +++++++++++++++++++++++++++++++-----
 1 file changed, 31 insertions(+), 5 deletions(-)

Comments

Ian Campbell Jan. 19, 2016, 10:50 a.m. UTC | #1

On Tue, 2016-01-19 at 15:17 +0800, Wen Congyang wrote:
> stream_continue() is used for migration to read emulator
> xenstore data and emulator context. For remus, if we do
> failover, we have read it in the checkpoint cycle, and
> we only need to complete the stream.
> 
> Signed-off-by: Wen Congyang <[e-mail]>
> Reviewed-by: Andrew Cooper <[e-mail]>

Acked-by: Ian Campbell <[e-mail]>

Patch

diff --git a/tools/libxl/libxl_stream_read.c b/tools/libxl/libxl_stream_read.c
index 258dec4..dac134e 100644
--- a/tools/libxl/libxl_stream_read.c
+++ b/tools/libxl/libxl_stream_read.c
@@ -104,6 +104,20 @@ 
  * Depending on the contents of the stream, there are likely to be several
  * parallel tasks being managed.  check_all_finished() is used to join all
  * tasks in both success and error cases.
+ *
+ * Failover for remus
+ *  - We buffer all records until a CHECKPOINT_END record is received
+ *  - We will consume the buffered records when a CHECKPOINT_END record
+ *    is received
+ *  - If we find some internal error, then rc or retval is not 0 in
+ *    libxl__xc_domain_restore_done(). In this case, we don't resume the
+ *    guest
+ *  - If we need to do failover from primary, then rc and retval are both
+ *    0 in libxl__xc_domain_restore_done(). In this case, the buffered
+ *    state will be dropped, because we haven't received a CHECKPOINT_END
+ *    record, and therefore the buffered state is inconsistent. In
+ *    libxl__xc_domain_restore_done(), we just complete the stream and
+ *    stream->completion_callback() will be called to resume the guest
  */
 
 /* Success/error/cleanup handling. */
@@ -758,6 +772,9 @@  void libxl__xc_domain_restore_done(libxl__egc *egc, void *dcs_void,
     libxl__stream_read_state *stream = &dcs->srs;
     STATE_AO_GC(dcs->ao);
 
+    /* convenience aliases */
+    const int checkpointed_stream = dcs->restore_params.checkpointed_stream;
+
     if (rc)
         goto err;
 
@@ -777,11 +794,20 @@  void libxl__xc_domain_restore_done(libxl__egc *egc, void *dcs_void,
      * If the stream is not still alive, we must not continue any work.
      */
     if (libxl__stream_read_inuse(stream)) {
-        /*
-         * Libxc has indicated that it is done with the stream.  Resume reading
-         * libxl records from it.
-         */
-        stream_continue(egc, stream);
+        if (checkpointed_stream) {
+            /*
+             * Failover from primary. Domain state is currently at a
+             * consistent checkpoint, complete the stream, and call
+             * stream->completion_callback() to resume the guest.
+             */
+            stream_complete(egc, stream, 0);
+        } else {
+            /*
+             * Libxc has indicated that it is done with the stream.
+             * Resume reading libxl records from it.
+             */
+            stream_continue(egc, stream);
+        }
     }
 }