From: Fei Li <fli@suse.com>
To: qemu-devel@nongnu.org
Cc: quintela@redhat.com, famz@redhat.com, armbru@redhat.com,
peterx@redhat.com, dgilbert@redhat.com
Subject: Re: [Qemu-devel] [PATCH RFC v6 5/7] migration: fix the multifd code when receiving less channels
Date: Wed, 31 Oct 2018 21:50:21 +0800 [thread overview]
Message-ID: <5cd70ffc-a761-f7c5-aaab-32e408fb253c@suse.com> (raw)
In-Reply-To: <20181029125818.28720-6-fli@suse.com>
Hi all,
I create a new thread to inquiry one live migration issue when using
multifd :)
I am not so sure with the rule that when and how to use the multifd is
correct,
so I'd like to confirm. This is because when I use the current upstream qemu
code and run into a failed case: "Migration status: failed (Unable to
write to
socket: Connection reset by peer)".
The detailed is as follows and the "failed" situation can not reproduce
100%.
But as far as I tested, if I do the live migration using multifd* just
after the
guest started for less than one minute*, I almost can reproduce this for
100%.
My steps are:
1. start the vm in the src side;
2. start the -incoming in the dst side;
3. after the vm started for a little while (After I open a file inside
the vm),
I begin the live migration, steps are:
- on src: migrate_set_capability x-multifd on
- on src: migrate_set_parameter x-multifd-channels 4
- on dst: migrate_set_capability x-multifd on
- on dst: migrate_set_parameter x-multifd-channels 4
- on src: migrate -d tcp:192.168.120.5:4444
Errors are:
[src]
linux-50ts:/mnt/live-migration # ./sle12-source.sh
QEMU 3.0.50 monitor - type 'help' for more information
(qemu) Running QEMU with SDL 1.2 is deprecated, and will be removed
in a future release. Please switch to SDL 2.0 instead
migrate_set_capability x-multifd on
(qemu) migrate_set_parameter x-multifd-channels 4
(qemu) migrate -d tcp:192.168.120.5:4444
(qemu) info migrate
globals:
store-global-state: on
only-migratable: off
send-configuration: on
send-section-footer: on
decompress-error-check: on
capabilities: xbzrle: off rdma-pin-all: off auto-converge: off
zero-blocks: off compress: off events: off postcopy-ram: off x-colo: off
release-ram: off block: off return-path: off pause-before-switchover:
off x-multifd: on dirty-bitmaps: off postcopy-blocktime: off
late-block-activate: off
Migration status: failed (Unable to write to socket: Connection reset by
peer)
total time: 0 milliseconds
[dst]
linux-p6v6:/mnt/live-migration # ./sle12-dest.sh
QEMU 3.0.50 monitor - type 'help' for more information
(qemu) migrate_set_capability x-multifd on
(qemu) migrate_set_parameter x-multifd-channels 4
(qemu) info migrate
globals:
store-global-state: on
only-migratable: off
send-configuration: on
send-section-footer: on
decompress-error-check: on
Hope this does not bother you too much. ;)
Have a nice day, thanks again
Fei
On 10/29/2018 08:58 PM, Fei Li wrote:
> In our current code, when multifd is used during migration, if there
> is an error before the destination receives all new channels, the
> source keeps running, however the destination does not exit but keeps
> waiting until the source is killed deliberately.
>
> Fix this by simply killing the destination when it fails to receive
> packet via some channel.
>
> Cc: Dr. David Alan Gilbert <dgilbert@redhat.com>
> Cc: Peter Xu <peterx@redhat.com>
> Signed-off-by: Fei Li <fli@suse.com>
> ---
> migration/channel.c | 7 ++++++-
> migration/migration.c | 9 +++++++--
> migration/migration.h | 2 +-
> migration/ram.c | 17 ++++++++++++++---
> migration/ram.h | 2 +-
> 5 files changed, 29 insertions(+), 8 deletions(-)
>
> diff --git a/migration/channel.c b/migration/channel.c
> index 33e0e9b82f..572be4245a 100644
> --- a/migration/channel.c
> +++ b/migration/channel.c
> @@ -44,7 +44,12 @@ void migration_channel_process_incoming(QIOChannel *ioc)
> error_report_err(local_err);
> }
> } else {
> - migration_ioc_process_incoming(ioc);
> + Error *local_err = NULL;
> + migration_ioc_process_incoming(ioc, &local_err);
> + if (local_err) {
> + error_report_err(local_err);
> + exit(EXIT_FAILURE);
> + }
> }
> }
>
> diff --git a/migration/migration.c b/migration/migration.c
> index 8b36e7f184..87dfc7374f 100644
> --- a/migration/migration.c
> +++ b/migration/migration.c
> @@ -541,7 +541,7 @@ void migration_fd_process_incoming(QEMUFile *f)
> migration_incoming_process();
> }
>
> -void migration_ioc_process_incoming(QIOChannel *ioc)
> +void migration_ioc_process_incoming(QIOChannel *ioc, Error **errp)
> {
> MigrationIncomingState *mis = migration_incoming_get_current();
> bool start_migration;
> @@ -563,9 +563,14 @@ void migration_ioc_process_incoming(QIOChannel *ioc)
> */
> start_migration = !migrate_use_multifd();
> } else {
> + Error *local_err = NULL;
> /* Multiple connections */
> assert(migrate_use_multifd());
> - start_migration = multifd_recv_new_channel(ioc);
> + start_migration = multifd_recv_new_channel(ioc, &local_err);
> + if (local_err) {
> + error_propagate(errp, local_err);
> + return;
> + }
> }
>
> if (start_migration) {
> diff --git a/migration/migration.h b/migration/migration.h
> index f7813f8261..7df4d426d0 100644
> --- a/migration/migration.h
> +++ b/migration/migration.h
> @@ -229,7 +229,7 @@ struct MigrationState
> void migrate_set_state(int *state, int old_state, int new_state);
>
> void migration_fd_process_incoming(QEMUFile *f);
> -void migration_ioc_process_incoming(QIOChannel *ioc);
> +void migration_ioc_process_incoming(QIOChannel *ioc, Error **errp);
> void migration_incoming_process(void);
>
> bool migration_has_all_channels(void);
> diff --git a/migration/ram.c b/migration/ram.c
> index 4db3b3e8f4..8f03afe228 100644
> --- a/migration/ram.c
> +++ b/migration/ram.c
> @@ -1072,6 +1072,7 @@ out:
> static void multifd_new_send_channel_async(QIOTask *task, gpointer opaque)
> {
> MultiFDSendParams *p = opaque;
> + MigrationState *s = migrate_get_current();
> QIOChannel *sioc = QIO_CHANNEL(qio_task_get_source(task));
> Error *local_err = NULL;
>
> @@ -1080,6 +1081,7 @@ static void multifd_new_send_channel_async(QIOTask *task, gpointer opaque)
> }
>
> if (qio_task_propagate_error(task, &local_err)) {
> + migrate_set_state(&s->state, s->state, MIGRATION_STATUS_FAILED);
> if (multifd_save_cleanup(&local_err) != 0) {
> migrate_set_error(migrate_get_current(), local_err);
> }
> @@ -1337,16 +1339,20 @@ bool multifd_recv_all_channels_created(void)
> }
>
> /* Return true if multifd is ready for the migration, otherwise false */
> -bool multifd_recv_new_channel(QIOChannel *ioc)
> +bool multifd_recv_new_channel(QIOChannel *ioc, Error **errp)
> {
> + MigrationIncomingState *mis = migration_incoming_get_current();
> MultiFDRecvParams *p;
> Error *local_err = NULL;
> int id;
>
> id = multifd_recv_initial_packet(ioc, &local_err);
> if (id < 0) {
> + error_propagate_prepend(errp, local_err,
> + "failed to receive packet via multifd channel %x: ",
> + multifd_recv_state->count);
> multifd_recv_terminate_threads(local_err, false);
> - return false;
> + goto fail;
> }
>
> p = &multifd_recv_state->params[id];
> @@ -1354,7 +1360,8 @@ bool multifd_recv_new_channel(QIOChannel *ioc)
> error_setg(&local_err, "multifd: received id '%d' already setup'",
> id);
> multifd_recv_terminate_threads(local_err, true);
> - return false;
> + error_propagate(errp, local_err);
> + goto fail;
> }
> p->c = ioc;
> object_ref(OBJECT(ioc));
> @@ -1366,6 +1373,10 @@ bool multifd_recv_new_channel(QIOChannel *ioc)
> QEMU_THREAD_JOINABLE);
> atomic_inc(&multifd_recv_state->count);
> return multifd_recv_state->count == migrate_multifd_channels();
> +fail:
> + qemu_fclose(mis->from_src_file);
> + mis->from_src_file = NULL;
> + return false;
> }
>
> /**
> diff --git a/migration/ram.h b/migration/ram.h
> index 83ff1bc11a..046d3074be 100644
> --- a/migration/ram.h
> +++ b/migration/ram.h
> @@ -47,7 +47,7 @@ int multifd_save_cleanup(Error **errp);
> int multifd_load_setup(void);
> int multifd_load_cleanup(Error **errp);
> bool multifd_recv_all_channels_created(void);
> -bool multifd_recv_new_channel(QIOChannel *ioc);
> +bool multifd_recv_new_channel(QIOChannel *ioc, Error **errp);
>
> uint64_t ram_pagesize_summary(void);
> int ram_save_queue_pages(const char *rbname, ram_addr_t start, ram_addr_t len);
next prev parent reply other threads:[~2018-10-31 13:50 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-10-29 12:58 [Qemu-devel] [PATCH RFC v6 0/7] qemu_thread_create: propagate errors to callers to check Fei Li
2018-10-29 12:58 ` [Qemu-devel] [PATCH RFC v6 1/7] Fix segmentation fault when qemu_signal_init fails Fei Li
2018-10-29 12:58 ` [Qemu-devel] [PATCH RFC v6 2/7] qemu_init_vcpu: add a new Error parameter to propagate Fei Li
2018-10-29 12:58 ` [Qemu-devel] [PATCH RFC v6 3/7] qemu_thread_join: fix segmentation fault Fei Li
2018-10-29 12:58 ` [Qemu-devel] [PATCH RFC v6 4/7] migration: fix some segmentation faults when using multifd Fei Li
2018-10-29 12:58 ` [Qemu-devel] [PATCH RFC v6 5/7] migration: fix the multifd code when receiving less channels Fei Li
2018-10-30 6:05 ` Peter Xu
2018-10-30 10:05 ` Fei Li
2018-10-30 22:18 ` Peter Xu
2018-10-31 12:26 ` Fei Li
2018-10-31 13:50 ` Fei Li [this message]
2018-10-29 12:58 ` [Qemu-devel] [PATCH RFC v6 6/7] migration: fix some error handling Fei Li
2018-10-30 19:49 ` Dr. David Alan Gilbert
2018-10-31 11:25 ` Fei Li
2018-10-31 16:30 ` Dr. David Alan Gilbert
2018-11-01 5:20 ` Fei Li
2018-10-29 12:58 ` [Qemu-devel] [PATCH RFC v6 7/7] qemu_thread_create: propagate the error to callers to handle Fei Li
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=5cd70ffc-a761-f7c5-aaab-32e408fb253c@suse.com \
--to=fli@suse.com \
--cc=armbru@redhat.com \
--cc=dgilbert@redhat.com \
--cc=famz@redhat.com \
--cc=peterx@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=quintela@redhat.com \
/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).