The function already takes two bool arguments, switching to flags makes
it a lot easier to read. Especially in case we need to add another
boolean in the future.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
In post-copy mode none of the hosts has a complete guest state and
rolling back migration is impossible. Thus aborting it would be
equivalent to destroying the domain.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
When migration fails in the post-copy mode, it's impossible to just kill
the destination domain and resume the source since the source no longer
contains current guest state. Let's mark domains on both sides as
VIR_DOMAIN_PAUSED_POSTCOPY_FAILED to let the upper layer decide what to
do with them.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
When destination libvirtd is restarted during migration in Finish phase
just after the point we started guest CPUs, we should not kill the
domain.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Migration enters "postcopy-active" state after QEMU switches to
post-copy and pauses guest CPUs. From libvirt's point of view this state
is similar to "completed" because we need to transfer guest execution to
the destination host.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
VIR_DOMAIN_EVENT_SUSPENDED_POSTCOPY and VIR_DOMAIN_PAUSED_POSTCOPY are
used on the source host once migration enters post-copy mode (which
means the domain gets paused on the source. After the destination host
takes over the execution of the domain, its virtual CPUs are resumed and
the domain enters VIR_DOMAIN_RUNNING_POSTCOPY state and
VIR_DOMAIN_EVENT_RESUMED_POSTCOPY event is emitted.
In case migration fails during post-copy mode and none of the hosts have
complete state of the domain, both domains will remain paused with
VIR_DOMAIN_PAUSED_POSTCOPY_FAILED reason and an upper layer may decide
what to do.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
The VIR_DOMAIN_EVENT_ID_JOB_COMPLETED event will be triggered once a job
(such as migration) finishes and it will contain statistics for the job
as one would get by calling virDomainGetJobStats. Thanks to this event
it is now possible to get statistics of a completed migration of a
transient domain on the source host.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
We would happily report and free statistics of a completed migration
even before it actually completed (on the source host while migration is
in the Finish phase).
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Computing a total downtime during a migration requires us to store a
time stamp when guest CPUs get stopped. The value (and all other
statistics) is then transferred to the destination to compute the
downtime. Because the stopped time stamp is stored by a STOP event
handler while the statistics which will be sent over to the destination
are copied synchronously within qemuMigrationWaitForCompletion.
Depending on the timing of STOP and MIGRATION events, we may end up
copying (and transferring) statistics without the stopped time stamp
set. Let's make sure we always use the correct time stamp.
https://bugzilla.redhat.com/show_bug.cgi?id=1282744
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
With a very old QEMU which doesn't support events we need to explicitly
call qemuMigrationSetOffline at the end of migration to update our
internal state. On the other hand, if we talk to QEMU using QMP, we
should just wait for the STOP event and let the event handler update the
state and trigger a libvirt event.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
We should not overwrite all migration statistics on the source with the
numbers sent by the destination since the source may have an updated
view in some cases (such as post-copy migration). It's safer to update
just the timing info we need to get from the destination and be prepared
for the future. And we should only do all this after a successful
migration.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Statistics for a completed migration only make sense if the migration
was successful. Let's not store them in priv->job.completed until we
are sure it was a success.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
When SPICE graphics is configured for a domain but we did not ask the
client to switch to the destination, we should not wait for
SPICE_MIGRATE_COMPLETED event (which will never come).
https://bugzilla.redhat.com/show_bug.cgi?id=1151723
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Migration statistics are not available on the destination host and
starting a query job during incoming migration is not allowed. Trying to
do that would result in
Timed out during operation: cannot acquire state change lock (held
by remoteDispatchDomainMigratePrepare3Params)
error. We should not even try to start the job.
https://bugzilla.redhat.com/show_bug.cgi?id=1278727
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
virDomainObjWait is designed to be called in a loop. Make sure we break
the loop in case the domain dies to avoid waiting for an event which
will never happen.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Calling qemuProcessStop without a job opens a way to race conditions
with qemuDomainObjExitMonitor called in another thread. A real world
example of such a race condition:
- migration thread (A) calls qemuMigrationWaitForSpice
- another thread (B) starts processing qemuDomainAbortJob API
- thread B signals thread A via qemuDomainObjAbortAsyncJob
- thread B enters monitor (qemuDomainObjEnterMonitor)
- thread B calls qemuMonitorSend
- thread A awakens and calls qemuProcessStop
- thread A calls qemuMonitorClose and sets priv->mon to NULL
- thread B calls qemuDomainObjExitMonitor with priv->mon == NULL
=> monitor stays ref'ed and locked
Depending on how lucky we are, the race may result in a memory leak or
it can even deadlock libvirtd's event loop if it tries to lock the
monitor to process an event received before qemuMonitorClose was called.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Stopping a domain without a job risks a race condition with another
thread which started a job a which does not expect anyone else to be
messing around with the same domain object.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Only a small portion of processGuestPanicEvent was enclosed within a
job, let's make sure we use the job for all operations to avoid race
conditions.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
When destroying a domain we need to make sure we will be able to start a
job no matter what other operations are running or even stuck in a job.
This is done by killing the domain before starting the destroy job.
Let's introduce qemuProcessBeginStopJob which combines killing a domain
and starting a job in a single API which can be called everywhere we
need a job to stop a domain.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Ending a nested job is no different from ending any other (non-async)
job, after all the code in qemuDomainBeginJobInternal does not handle
them differently either. Thus we should call qemuDomainObjEndJob to stop
nested jobs.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
The VIR_DOMAIN_EVENT_ID_MIGRATION_ITERATION event will be triggered
whenever VIR_DOMAIN_JOB_MEMORY_ITERATION changes its value, i.e.,
whenever a new iteration over guest memory pages is started during
migration.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
A device tree binary file specified by /domain/os/dtb element is a
read-only resource similar to kernel and initrd files. We shouldn't
restore its label when destroying a domain to avoid breaking other
domains configure with the same device tree.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Kernel/initrd files are essentially read-only shareable images and thus
should be handled in the same way. We already use the appropriate label
for kernel/initrd files when starting a domain, but when a domain gets
destroyed we would remove the labels which would make other running
domains using the same files very unhappy.
https://bugzilla.redhat.com/show_bug.cgi?id=921135
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
memory_dirty_rate corresponds to dirty-pages-rate in QEMU and
memory_iteration is what QEMU reports in dirty-sync-count.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
The structure actually contains migration statistics rather than just
the status as the name suggests. Renaming it as
qemuMonitorMigrationStats removes the confusion.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
A migration is in "setup" state after it was "inactive" and before it
becomes "active". Let's reflect this in our migration status enum.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
My commit 674afcb09e moved computing the
default listen address from qemuMigrationPrepareAny to
qemuMigrationPrepareIncoming. However, I didn't notice listenAddress was
later passed to qemuMigrationStartNBDServer. Thus, it would be called
with the original value of listenAddress (NULL).
Let's add the updated listen address to qemuProcessIncomingDef and use
it when starting NBD servers.
Reported-by: Michael Chapman <mike@very.puzzling.org>
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
The *event --loop commands would keep running even though a connection
to libvirtd is lost. This doesn't make a lot of sense since clearly we
won't get any new events from the closed connection.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Using qemuProcess{Init,Launch,FinishStartup} allows us to run
pre-migration commands on destination before asking QEMU to wait for
incoming migration data.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
NBD storage migration will not work with offline migration anyway and we
already checked that the user did not ask for it. Thus it doesn't make
sense to keep the code after 'done' label where we jump in case of
offline migration.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Some failure paths in qemuMigrationPrepareAny forgot to kill the just
started QEMU process. This patch fixes this by combining 'stop' and
'endjob' label into a new label 'stopjob'. This name was chosen to avoid
confusion with the most common semantics of 'endjob'. Normally, 'endjob'
is always called at the end of an API to stop the job we entered at the
beginning. In qemuMigrationPrepareAny we only want to stop the job in
failure path; on success we need to carry the job over to the Finish
phase.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Once qemuProcessInit was called, qemuProcessLaunch will launch a new
QEMU process with stopped virtual CPUs.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
qemuProcessStart is going to be split in three parts: qemuProcessInit,
qemuProcessLaunch, and qemuProcessFinish so that migration Prepare phase
can insert additional code in the process. qemuProcessStart will be a
small wrapper for all other callers.
qemuProcessInit prepares the domain up to the point when priv->qemuCaps
is initialized.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Remembering to call qemuMonitorSetDomainLog in the right paths before
calling qemuProcessStop is annoying and easy to forget. And I already
forgot to do so in commit v1.2.8-52-g0389060: logfd may be leaked if
QEMU process dies between Prepare and Finish migration phases.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
qemuProcessStart is so big that any nontrivial code should be moved to
dedicated functions to make the code easier to read and maintain.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
qemuProcessStart is so big that any nontrivial code should be moved to
dedicated functions to make the code easier to read and maintain.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
qemuProcessStart is so big that any nontrivial code should be moved to
dedicated functions to make the code easier to read and maintain.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
qemuProcessStart is so big that any nontrivial code should be moved to
dedicated functions to make the code easier to read and maintain.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Traditionally, we pass incoming migration URI on QEMU command line,
which has some drawbacks. Depending on the URI QEMU may initialize its
migration state immediately without giving us a chance to set any
additional migration parameters (this applies mainly for fd: URIs). For
some URIs the monitor may be completely blocked from the beginning until
migration is finished, which means we may be stuck in qmp_capabilities
command without being able to send any QMP commands.
QEMU solved this by introducing "defer" parameter for -incoming command
line option. This will tell QEMU to prepare for an incoming migration
while the actual incoming URI is sent using migrate-incoming QMP
command. Before calling this command we can normally talk to the
monitor and even set any migration parameters which will be honored by
the incoming migration.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
We only started an async job for incoming migration from another host.
When we were starting a domain from scratch or restoring from a saved
state (migration from file) we didn't set any async job. Let's introduce
a new QEMU_ASYNC_JOB_START for these cases.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Incoming migration may require quite a few parameters (URI, fd, path) to
be considered while starting QEMU and we will soon add another one.
Let's group all of them in a single struct.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Make callers of qemuBuildCommandLine responsible for providing the URI
which should be passed as a parameter for -incoming.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Logging current async job while in BeginJob is useful, but the async job
we want to start is even more interesting.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
It's just a copy&paste of qemuxml2argv test anyway. We can test most of
them (except for qemuxmlns-qemu-ns-domain.xml which fails to validate
against our schema) by qemuxml2argv test.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Fixes several style issues and removes "DEF" (what is it supposed to
mean anyway?) from debug messages.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
VIR_DEBUG and VIR_WARN will automatically add a new line to the message,
having "\n" at the end or at the beginning of the message results in
empty lines.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
nodeset should be freed in both success and failure paths.
While tmppath is freed immediately after it's consumed, moving it from
error to cleanup label is a bit more consistent and robust.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Generally, we use "ret" variable for storing the value we are going to
return at the and of a function, but this is not the case in
qemuProcessStart. Let's rename "ret" as "rv".
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
qemuProcessStart was passing char * migrateFrom as the third argument to
qemuPrepareNVRAM. We should explicitly convert the pointer to bool which
is what the function expects.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Even though QEMU on the source host reports completed migration and thus
we move to the Finish phase, QEMU on the destination host may still be
processing migration data. Thus before we can start guest CPUs on the
destination, we have to wait for a completed migration event.
https://bugzilla.redhat.com/show_bug.cgi?id=1265902
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
With new QEMU which supports migration events,
qemuMigrationCheckJobStatus needs to explicitly query QEMU for migration
statistics once migration is completed to make sure the caller sees
up-to-date statistics with both old and new QEMU. However, some callers
are not interested in the statistics at all and once we start waiting
for a completed migration on the destination host too, checking the
statistics would even fail. Let's push the decision whether to update
the statistics or not to the caller.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
The function already has two bool parameters and we will need to add a
new one. Let's switch to flags to make the callers readable.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
The destination host gets detailed statistics about the current
migration form the source host via migration cookie and copies them to
the domain object so that they can be queried using
virDomainGetJobStats. However, we should only copy statistics to the
domain object when migration finished successfully.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Even if we are migrating a domain with VIR_MIGRATE_PAUSED flag set, we
should still update the total time of the migration. Updating downtime
doesn't hurt either, even though we don't actually start guest CPUs.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
As of commit 6992994, we set graphics/@listen attribute according to the
first listen child element even if that element is of type='network'.
This was done for backward compatibility with applications which only
support the original listen attribute. However, by doing so we broke
migration to older libvirt which tried to check that the listen
attribute matches one of the listen child elements but which did not
take type='network' elements into account.
We are not concerned about compatibility with old applications when
formatting domain XML for migration for two reasons. The XML is consumed
only by libvirtd and the IP address associated with type='network'
listen address on the source host is just useless on the destination
host. Thus, we can safely avoid propagating the type='network' IP
address to graphics/@listen attribute when creating migratable XML.
https://bugzilla.redhat.com/show_bug.cgi?id=1265111
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
After my "client rpc: Report proper error for keepalive disconnections"
patch, virsh would no long print a warning when it closes a connection
to a daemon after a keepalive timeout. Although the warning
virsh # 2015-09-15 10:59:26.729+0000: 642080: info :
libvirt version: 1.2.19
2015-09-15 10:59:26.729+0000: 642080: warning :
virKeepAliveTimerInternal:143 : No response from client
0x7efdc0a46730 after 1 keepalive messages in 2 seconds
was pretty ugly, it was still useful. This patch brings the useful part
back while making it much nicer:
virsh # error: Disconnected from qemu:///system due to keepalive timeout
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Even though we hit an error in client's IO loop, we still want to
process any pending data. So instead of reporting the error right away,
we can finish the current iteration and report the error once we're done
with it. Note that the error is stored in client->error by
virNetClientMarkClose so we don't need to worry about it being reset or
rewritten by any API we call in the meantime.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Whenever a connection was closed due to keepalive timeout, we would log
a warning but the interrupted API would return rather useless generic
error:
internal error: received hangup / error event on socket
Let's report a proper keepalive timeout error and make sure it is
propagated to all pending APIs. The error should be better now:
internal error: connection closed due to keepalive timeout
Based on an old patch from Martin Kletzander.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
When persistently migrating a domain to a destination host where the
same domain already exists (i.e., it is persistent and shutdown at the
destination), we would happily throw away the original persistent
definition without properly freeing it. And when updating the definition
fails for some reason we don't properly revert to the original state
leaving the domain broken.
In addition to fixing these issues, the patch also makes sure the domain
definition parsed from a migration cookie is either used or freed.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
For quite a long time we don't need to postpone queueing events until
the end of the function since we no longer have the big driver lock.
Let's make the code of qemuMigrationFinish simpler by queuing events at
the time we generate them.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Every single call to qemuDomainEventQueue() uses the following pattern:
if (event)
qemuDomainEventQueue(driver, event);
Let's move the check for valid event to qemuDomainEventQueue and
simplify all callers.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Finish is the final state in v2 of our migration protocol. If something
fails, we have no option to abort the migration and resume the original
domain. Non fatal errors (such as failure to start guest CPUs or make
the domain persistent) has to be treated as success. Keeping the domain
running while reporting the failure was just asking for trouble.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Jiri Denemark