On 05/10/2009 08:15 PM, sudhir kumar wrote:
Hi Uri,
Any comments?
---------- Forwarded message ----------
From: sudhir kumar<smalikphy@xxxxxxxxx>
The kvm-autotest shows the following PASS results for migration,
while the VM was crashed and test should have failed.
Here is the sequence of test commands and results grepped from
kvm-autotest output.
/root/sudhir/regression/test/kvm-autotest-phx/client/tests/kvm_runtest_2/qemu
-name 'vm1' -monitor
unix:/tmp/monitor-20090508-055624-QSuS,server,nowait -drive
file=/root/sudhir/regression/test/kvm-autotest-phx/client/tests/kvm_runtest_2/images/rhel5-32.raw,if=ide,boot=on
-net nic,vlan=0 -net user,vlan=0 -m 8192
-smp 4 -redir tcp:5000::22 -vnc :1
/root/sudhir/regression/test/kvm-autotest-phx/client/tests/kvm_runtest_2/qemu
-name 'dst' -monitor
unix:/tmp/monitor-20090508-055625-iamW,server,nowait -drive
file=/root/sudhir/regression/test/kvm-autotest-phx/client/tests/kvm_runtest_2/images/rhel5-32.raw,if=ide,boot=on
-net nic,vlan=0 -net user,vlan=0 -m 8192
-smp 4 -redir tcp:5001::22 -vnc :2 -incoming tcp:0:5200
2009-05-08 05:58:43,471 Configuring logger for client level
GOOD
kvm_runtest_2.raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.1
END GOOD
kvm_runtest_2.raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.1
GOOD
kvm_runtest_2.raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2
kvm_runtest_2.raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2
timestamp=1241762371
localtime=May 08 05:59:31 completed successfully
Persistent state variable __group_level now set to 1
END GOOD
kvm_runtest_2.raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2
kvm_runtest_2.raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2
timestamp=1241762371
localtime=May 08 05:59:31
From the test output it looks that the test was succesful to
log into the guest after migration:
20090508-055926 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: Migration
finished successfully
20090508-055926 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
send_monitor_cmd: Sending monitor command: screendump
/root/sudhir/regression/test/kvm-autotest-phx/client/results/default/kvm_runtest_2.raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2/debug/migration_post.ppm
20090508-055926 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
send_monitor_cmd: Sending monitor command: screendump
/root/sudhir/regression/test/kvm-autotest-phx/client/results/default/kvm_runtest_2.raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2/debug/migration_pre.ppm
20090508-055926 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
send_monitor_cmd: Sending monitor command: quit
20090508-055926 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
is_sshd_running: Timeout
20090508-055926 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: Logging into
guest after migration...
20090508-055926 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
remote_login: Trying to login...
20090508-055927 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
remote_login: Got 'Are you sure...'; sending 'yes'
20090508-055927 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
remote_login: Got password prompt; sending '123456'
20090508-055928 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
remote_login: Got shell prompt -- logged in
20090508-055928 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: Logged in
after migration
20090508-055928 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
get_command_status_output: Sending command: help
20090508-055930 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
postprocess_vm: Postprocessing VM 'vm1'...
20090508-055930 raw.8gb_mem.smp4.RHEL.5.3.i386.migrate.2: DEBUG:
postprocess_vm: VM object found in environment
When I did vnc to the final migrated VM was crashed with a call trace
as shown in the attachment.
Quite less possible that the call trace appeared after the test
finished as migration with memory
more than 4GB is already broken [BUG 52527]. This looks a false PASS
to me. Any idea how can we handle
such falso positive results? Shall we wait for sometime after
migration, log into the vm, do some work or run some good test,
get output and report that the vm is alive?
I don't think it's a False PASS.
It seems the test was able to ssh into the guest, and run a command on the guest.
Currently we only run migration once (round-trip). I think we should run
migration more than once (using iterations). If the guest crashes due to
migration, it would fail following rounds of migration.
Sorry for the late reply,
Uri.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
