qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Peter Xu <peterx@redhat.com>
To: qemu-devel@nongnu.org
Cc: peterx@redhat.com, Juraj Marcin <jmarcin@redhat.com>,
	Julia Suvorova <jusual@redhat.com>,
	Prasad Pandit <ppandit@redhat.com>,
	Fabiano Rosas <farosas@suse.de>
Subject: [PATCH 03/16] migration: Optimize postcopy on downtime by avoiding JSON writer
Date: Tue, 14 Jan 2025 18:07:33 -0500	[thread overview]
Message-ID: <20250114230746.3268797-4-peterx@redhat.com> (raw)
In-Reply-To: <20250114230746.3268797-1-peterx@redhat.com>

postcopy_start() is the entry function that postcopy is destined to start.
It also means QEMU source will not dump VM description, aka, the JSON
writer is garbage now.

We can leave that to be cleaned up when migration completes, however when
with the JSON writer object being present, vmstate_save() will still try to
construct the JSON objects for the VM descriptions, even though it'll never
be used later if it's postcopy.

To save those cycles, release the JSON writer earlier for postcopy. Then
vmstate_save() later will be smart enough to skip the JSON object
constructions completely.  It can logically reduce downtime because all
such JSON constructions happen during postcopy blackout.

Signed-off-by: Peter Xu <peterx@redhat.com>
---
 migration/migration.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/migration/migration.c b/migration/migration.c
index bad7b39293..0eb28e850d 100644
--- a/migration/migration.c
+++ b/migration/migration.c
@@ -1397,6 +1397,11 @@ void migrate_set_state(MigrationStatus *state, MigrationStatus old_state,
     }
 }
 
+static void migration_cleanup_json_writer(MigrationState *s)
+{
+    g_clear_pointer(&s->vmdesc, json_writer_free);
+}
+
 static void migrate_fd_cleanup(MigrationState *s)
 {
     MigrationEventType type;
@@ -1404,11 +1409,11 @@ static void migrate_fd_cleanup(MigrationState *s)
 
     trace_migrate_fd_cleanup();
 
+    migration_cleanup_json_writer(s);
+
     g_free(s->hostname);
     s->hostname = NULL;
 
-    g_clear_pointer(&s->vmdesc, json_writer_free);
-
     qemu_savevm_state_cleanup();
 
     close_return_path_on_source(s);
@@ -2501,6 +2506,14 @@ static int postcopy_start(MigrationState *ms, Error **errp)
     uint64_t bandwidth = migrate_max_postcopy_bandwidth();
     int cur_state = MIGRATION_STATUS_ACTIVE;
 
+    /*
+     * Now we're 100% sure to switch to postcopy, so JSON writer won't be
+     * useful anymore.  Free the resources early if it is there.  Clearing
+     * the vmdesc also means any follow up vmstate_save()s will start to
+     * skip all JSON operations, which can shrink postcopy downtime.
+     */
+    migration_cleanup_json_writer(ms);
+
     if (migrate_postcopy_preempt()) {
         migration_wait_main_channel(ms);
         if (postcopy_preempt_establish_channel(ms)) {
-- 
2.47.0



  parent reply	other threads:[~2025-01-14 23:08 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-01-14 23:07 [PATCH 00/16] migration: Switchover phase refactoring Peter Xu
2025-01-14 23:07 ` [PATCH 01/16] migration: Remove postcopy implications in should_send_vmdesc() Peter Xu
2025-01-14 23:07 ` [PATCH 02/16] migration: Do not construct JSON description if suppressed Peter Xu
2025-01-14 23:07 ` Peter Xu [this message]
2025-01-14 23:07 ` [PATCH 04/16] migration: Avoid two src-downtime-end tracepoints for postcopy Peter Xu
2025-01-14 23:07 ` [PATCH 05/16] migration: Drop inactivate_disk param in qemu_savevm_state_complete* Peter Xu
2025-01-14 23:07 ` [PATCH 06/16] migration: Synchronize all CPU states only for non-iterable dump Peter Xu
2025-01-14 23:07 ` [PATCH 07/16] migration: Adjust postcopy bandwidth during switchover Peter Xu
2025-01-14 23:07 ` [PATCH 08/16] migration: Adjust locking in migration_maybe_pause() Peter Xu
2025-01-14 23:07 ` [PATCH 09/16] migration: Drop cached migration state " Peter Xu
2025-01-14 23:07 ` [PATCH 10/16] migration: Take BQL slightly longer in postcopy_start() Peter Xu
2025-01-14 23:07 ` [PATCH 11/16] migration: Notify COMPLETE once for postcopy Peter Xu
2025-01-14 23:07 ` [PATCH 12/16] migration: Unwrap qemu_savevm_state_complete_precopy() in postcopy Peter Xu
2025-01-14 23:07 ` [PATCH 13/16] migration: Cleanup qemu_savevm_state_complete_precopy() Peter Xu
2025-01-14 23:07 ` [PATCH 14/16] migration: Always set DEVICE state Peter Xu
2025-01-14 23:07 ` [PATCH 15/16] migration: Merge precopy/postcopy on switchover start Peter Xu
2025-01-14 23:07 ` [PATCH 16/16] migration: Trivial cleanup on JSON writer of vmstate_save() Peter Xu
2025-01-15  9:12 ` [PATCH 00/16] migration: Switchover phase refactoring Jiri Denemark
2025-01-15 12:55   ` Peter Xu
2025-01-15 16:13 ` Juraj Marcin
2025-01-15 16:49 ` Fabiano Rosas

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20250114230746.3268797-4-peterx@redhat.com \
    --to=peterx@redhat.com \
    --cc=farosas@suse.de \
    --cc=jmarcin@redhat.com \
    --cc=jusual@redhat.com \
    --cc=ppandit@redhat.com \
    --cc=qemu-devel@nongnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).