01_enforce_dom0_cpus_basic_pos reset vcpu count on failures.
PATCH 1/1 01_enforce_dom0_cpus_basic_pos reset vcpu count on failures.
If 01_enforce_dom0_cpus_basic_pos fails, the vcpu count for dom0 does not get restored causing subsequent boots to come up with only a single vcpu on dom0.
01_enforce_dom0_cpus_basic_pos.py | 29 +++++++++++++++-------------- 1 file changed, 15 insertions(+), 14 deletions(-)
# HG changeset patch
# User [email protected]
# Node ID 654b58e2c44d50ffcb7eaabfbaf41d4c5e9b09d7
# Parent 606db2959bd6dc7fc178e9792f0ef345e48dd35e
Fix a problem where the vcpu count for dom0 doesn't get reset back to the original value when 01_enforce_dom0_cpus_basic_pos fails.
tools/xm-test/tests/enforce_dom0_cpus/01_enforce_dom0_cpus_basic_pos.py +15 -14
@@ -28,6 +28,12 @@ check_status = 1 max_tries = 10 +def reset_vcpu_count(): + status, output = traceCommand("xm vcpu-set 0 %s"%(dom0_online_vcpus)) + if status != 0: + print "WARNING!!! Unable to set vcpus back to %s, please set manually"\ + %(dom0_online_vcpus) + # 1) Make sure we have a multi cpu system and dom0 has at least 2 vcpus online. if smpConcurrencyLevel() <= 1:
@@ -57,6 +63,7 @@ # 4) restart xend with new config os.putenv("XEND_CONFIG", "/tmp/xend-config.sxp") status = restartXend() +os.unsetenv("XEND_CONFIG") if check_status and status != 0: ns, no = restartXend() if ns != 0:
@@ -75,7 +82,7 @@ cmd = "grep \"^processor\" /proc/cpuinfo | wc -l" status, output = traceCommand(cmd) if check_status and status != 0: - os.unsetenv("XEND_CONFIG") + reset_vcpu_count() restartXend() FAIL("\"%s\" returned invalid %i != 0" %(cmd,status)) # Has it succeeded? If so, we can leave the loop
@@ -84,7 +91,7 @@ # Sleep for 1 second before trying again time.sleep(1) if output != str(enforce_dom0_cpus): - os.unsetenv("XEND_CONFIG") + reset_vcpu_count() restartXend() FAIL("/proc/cpuinfo says xend didn't enforce dom0_cpus (%s != %s)"%(output, enforce_dom0_cpus))
@@ -92,34 +99,28 @@ # 6) count number of online cpus and see that it matches enforce value num_online = int(getDomInfo("Domain-0", "VCPUs")) if num_online != enforce_dom0_cpus: - os.unsetenv("XEND_CONFIG") + reset_vcpu_count() restartXend() FAIL("xm says xend didn't enforce dom0_cpus (%s != %s)" %(num_online, enforce_dom0_cpus)) # 7) restore dead processors -status, output = traceCommand("xm vcpu-set 0 %s"%(dom0_online_vcpus)) -if check_status and status != 0: - os.unsetenv("XEND_CONFIG") - restartXend() - FAIL("\"%s\" returned invalid %i != 0" %(cmd,status)) +reset_vcpu_count() # check restore worked # Since this also takes time, we will do it in a loop with a 20 second timeout. timeout=20 starttime=time.time() while timeout + starttime > time.time(): - num_online = int(getDomInfo("Domain-0", "VCPUs")) - if num_online == dom0_online_vcpus: - break - time.sleep(1) + num_online = int(getDomInfo("Domain-0", "VCPUs")) + if num_online == dom0_online_vcpus: + break + time.sleep(1) if num_online != dom0_online_vcpus: - os.unsetenv("XEND_CONFIG") restartXend() FAIL("failed to restore dom0's VCPUs") # 8) Restart xend with default config -os.unsetenv("XEND_CONFIG") restartXend()
_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel
Quoted 2 lines · Paul Larson
If 01_enforce_dom0_cpus_basic_pos fails, the vcpu count for dom0 does not get restored causing subsequent boots to come up with only a single vcpu on dom0.
Though I've never seen this mentioned on the list, I thought we wanted to leave the system in the failed state if a test failed? That's certainly debatable. Thoughts?
tools/xm-test/tests/enforce_dom0_cpus/01_enforce_dom0_cpus_basic_pos.py line 118
Quoted 8 lines
tools/xm-test/tests/enforce_dom0_cpus/01_enforce_dom0_cpus_basic_pos.py +4
-4- num_online = int(getDomInfo("Domain-0", "VCPUs")) - if num_online == dom0_online_vcpus: - break - time.sleep(1) + num_online = int(getDomInfo("Domain-0", "VCPUs")) + if num_online == dom0_online_vcpus: + break + time.sleep(1)
White-space damage.
Signature · 5 lines
Ryan Harper Software Engineer; Linux Technology Center IBM Corp., Austin, Tx (512) 838-9253 T/L: 678-9253 [email protected]
PATCH 1/1 (corrected) 01_enforce_dom0_cpus_basic_pos reset vcpu count on failures.
Quoted 8 lines · Ryan Harper
Quoted 3 lines · Paul Larson
If 01_enforce_dom0_cpus_basic_pos fails, the vcpu count for dom0 does not get restored causing subsequent boots to come up with only a single vcpu on dom0.
Though I've never seen this mentioned on the list, I thought we wanted to leave the system in the failed state if a test failed? That's certainly debatable. Thoughts?
Generally, that's probably a good rule but in this case it's better to go with the FAIL() output and put it back to a reasonable state because the altered state can persist not only out the end of the test run, but across xend restarts and even reboots. The reason this came up is because a bug was filed where a few systems were only showing one cpu in /proc/cpuinfo, but they were configured for smp and had previously booted with all their processors.
Quoted 1 line
White-space damage.
Doh! Fixed version attached.
01_enforce_dom0_cpus_basic_pos.py | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-)
# HG changeset patch
# User [email protected]
# Node ID 74b14312d90ba444a9e697c64adb10f60233261a
# Parent 606db2959bd6dc7fc178e9792f0ef345e48dd35e
Fix a problem where the vcpu count for dom0 doesn't get reset back to the original value when 01_enforce_dom0_cpus_basic_pos fails.
tools/xm-test/tests/enforce_dom0_cpus/01_enforce_dom0_cpus_basic_pos.py +11 -10
@@ -28,6 +28,12 @@ check_status = 1 max_tries = 10 +def reset_vcpu_count(): + status, output = traceCommand("xm vcpu-set 0 %s"%(dom0_online_vcpus)) + if status != 0: + print "WARNING!!! Unable to set vcpus back to %s, please set manually"\ + %(dom0_online_vcpus) + # 1) Make sure we have a multi cpu system and dom0 has at least 2 vcpus online. if smpConcurrencyLevel() <= 1:
@@ -57,6 +63,7 @@ # 4) restart xend with new config os.putenv("XEND_CONFIG", "/tmp/xend-config.sxp") status = restartXend() +os.unsetenv("XEND_CONFIG") if check_status and status != 0: ns, no = restartXend() if ns != 0:
@@ -75,7 +82,7 @@ cmd = "grep \"^processor\" /proc/cpuinfo | wc -l" status, output = traceCommand(cmd) if check_status and status != 0: - os.unsetenv("XEND_CONFIG") + reset_vcpu_count() restartXend() FAIL("\"%s\" returned invalid %i != 0" %(cmd,status)) # Has it succeeded? If so, we can leave the loop
@@ -84,7 +91,7 @@ # Sleep for 1 second before trying again time.sleep(1) if output != str(enforce_dom0_cpus): - os.unsetenv("XEND_CONFIG") + reset_vcpu_count() restartXend() FAIL("/proc/cpuinfo says xend didn't enforce dom0_cpus (%s != %s)"%(output, enforce_dom0_cpus))
@@ -92,17 +99,13 @@ # 6) count number of online cpus and see that it matches enforce value num_online = int(getDomInfo("Domain-0", "VCPUs")) if num_online != enforce_dom0_cpus: - os.unsetenv("XEND_CONFIG") + reset_vcpu_count() restartXend() FAIL("xm says xend didn't enforce dom0_cpus (%s != %s)" %(num_online, enforce_dom0_cpus)) # 7) restore dead processors -status, output = traceCommand("xm vcpu-set 0 %s"%(dom0_online_vcpus)) -if check_status and status != 0: - os.unsetenv("XEND_CONFIG") - restartXend() - FAIL("\"%s\" returned invalid %i != 0" %(cmd,status)) +reset_vcpu_count() # check restore worked # Since this also takes time, we will do it in a loop with a 20 second timeout.
@@ -114,12 +117,10 @@ break time.sleep(1) if num_online != dom0_online_vcpus: - os.unsetenv("XEND_CONFIG") restartXend() FAIL("failed to restore dom0's VCPUs") # 8) Restart xend with default config -os.unsetenv("XEND_CONFIG") restartXend()
_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel
Ewan Mellor
AppliedRe: [PATCH] (corrected) 01_enforce_dom0_cpus_basic_pos reset vcpu count on failures.
Quoted 17 lines · Paul Larson
Quoted 8 lines · Ryan Harper
Quoted 3 lines · Paul Larson
If 01_enforce_dom0_cpus_basic_pos fails, the vcpu count for dom0 does not get restored causing subsequent boots to come up with only a single vcpu on dom0.
Though I've never seen this mentioned on the list, I thought we wanted to leave the system in the failed state if a test failed? That's certainly debatable. Thoughts?
Generally, that's probably a good rule but in this case it's better to go with the
FAIL()output and put it back to a reasonable state because the altered state can persist not only out the end of the test run, but across xend restarts and even reboots. The reason this came up is because a bug was filed where a few systems were only showing one cpu in/proc/cpuinfo, but they were configured for smp and had previously booted with all their processors.
Quoted 5 lines
Quoted 1 line
White-space damage.
Doh! Fixed version attached.
01_enforce_dom0_cpus_basic_pos.py | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-)Signed-off-by Paul Larson <[email protected]>
Applied, thank you.
Ewan.
_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel