Comment 10 for bug 1353962

Revision history for this message
Matt Riedemann (mriedem) wrote : Re: Test job failes with FixedIpLimitExceeded with nova network

Looking at tests that extend BaseV2ComputeAdminTest because tempest.api.compute.admin.test_servers creates 2 test instances for the JSON run alone (and there is an xml version also). Those would be on the same admin tenant, so if those are running concurrently we could have at least 4 instances running at the same time under that tenant.

We have 4 test workers, so we could have 2 other tests running concurrently.

There are 36 test classes that extend BaseV2ComputeAdminTest, not all of them create test servers though.

Looking at the console log when our test fails:

http://logs.openstack.org/98/125398/11/check/check-tempest-dsvm-postgres-full/f0d9495/console.html#_2014-10-06_08_00_28_464

And looking at what's running before it, we see:

tempest.api.compute.admin.test_quotas_negative.QuotasAdminNegativeTestXML

and

tempest.api.compute.admin.test_quotas.QuotaClassesAdminTestJSON

We've had race issues with QuotaClassesAdminTestJSON in the past where it would set the default quota class values to 0 so we'd hit over-quota. That was changed awhile back to take the default quotas and add 100 to the limit, so that shouldn't be causing issues.

QuotasAdminNegativeTestXML is running with force_tenant_isolation=False

So the quota admin tests should be running in tenant isolation and not impact ServersAdminTestJSON.