Method get_cmd_pids is retrying calling list_pids if enough
PIDs for specified process aren't found.
In such case there is no need to check every time all PIDs which
were already checked and don't match command filter.
So now it will "remember" about PIDs already checked and will only
check cmdline for new pids found in each iteration.
Change-Id: Icf873812aec80a35a988b32e31d9199e248b201e
In later OSP releases, kexec-tools package is not installed on overcloud
images - Due to this, controller nodes are not recovered after
a triggering a kernel panic on more then one (in a setup with
three controllers, two of them need to be healthy in order to have quorum)
Change-Id: I32a8cb7d803eefa89695bdde0b04cb22180dde66
In OSP versions below 16.2 there is no 'Processes' column in the
header of `ss` command line tool output. While processes themselves
exist in the output body it still failed to parse results of the
command execution.
As a solution we may provide expected header as a parameter to the
parser as the output itself is identical for all versions
Change-Id: I2896cdbd2aa47e7bbb4d793b746a8abe0cc019ca
When the main_vip is an IPv6, the pacemaker status command output
replaces ":" by "."
The same replacement needs to be applied in tobiko
Change-Id: I03f90d689f93e5bd9e2df541238f259eea7694e1
This method calls tobiko.fail when it detects something wrong wrt
containers
tobiko.fail input parameters should be a string and an Exception, but
the code wrongly passes two strings
Change-Id: I8e25276e423e3fbc02b2a486c1f5f15066fd6037
This test often fails because it tries to connect via ssh immediately
after the reboot command is sent
paramiko gets stuck until it times out after 30 minutes
Change-Id: I8027fd76f8dcb0abdfefff0e3946a5046e97ef78
Recently, the default overcloud username value has changed in the
downstream jobs
After that, some tests using the podman API failed because they
configured the podman sockets service using a hardcoded value that
doesn't apply anymore
This patch removes the harcoded value from the podman code and also from
the cloud_disruptions module
The proper overcloud ssh username needs to be configured in the
tobiko.conf file (its default configurable value is still heat-admin)
Change-Id: I38af9de21caf4eb2a8b222bb8f849da0cea5a323
There are different places where we try to get information from ss
command so it make sense to have a common module to capture info
In current patch there is only tcp_listening handler added. More
functionality in following patches.
Change-Id: I15e8d1fc68946672d599863150da9406908522a4
Package sphinx_rtd_theme is constrainted in the global
upper constraints to version 0.5.1 so we need to have at least
the same version in our lower constraints for it.
Change-Id: I1f1b55c614a32ee35388895be240743729db505e
This job will be run in the post, release and pre-release queues
in zuul and will mirror or opendev.org/x/tobiko repository to the
redhat-openstack/tobiko project on github.
Change-Id: Ie657db33295ca8b70579203a5806889ddbef6237
Pre-commit's check-yaml hook by default loads all yaml files to
check them. This works fine so far but as we want to have zuul
job to mirror Tobiko code to the github.com we will need to use
"secret" with custom tag "!encrypted/pkcs1-oaep". That is causing
failure of the check-yaml hook while loading yaml files.
To avoid that failure, lets use "--unsafe" option in the check-yaml
hook. With this option hook instead of loading the files,
simply parse them for syntax.
Change-Id: I9a27da6c096bae933bc9254fbad5fc5c22269560
This module contained only one test "test_recreate_physical_bridge"
which was deleting physical bridges from the controller node, creating
those bridges again and ensuring that neutron-ovs-agent did
reconfiguration of that physical bridges properly and connectivity
through the bridges is restored.
Problem with this test is that it is "too much" intrusive. Physical
bridge needs to have some physical interface added to it to have
connectivity to the physical network. When bridge is deleted, such
port from the bridge is gone too. And Tobiko was adding bridge again,
but it wasn't adding interfaces again at all. That caused problems with
connectivity of the FIPs as well as some other stuff, like access to the
rabbitmq for services on that node when the same bridge was used for the
control plane networks.
The problem is that it would be very hard for tobiko to restore all the
configuration of the bridges as it can be some vlan interace (like
vlan50 in many Tripleo based virtual deployments used in CI), some
physical NIC or patch port to other bridge (like to br-infra in the u/s
CI configurations).
Taking all of that into account I think that it's better to simply
remove that test which in fact don't have that much value and make
Tobiko test suite more stable in testing other stuff.
Change-Id: I1f87cd74936809fe391cd8841d29ece46f9c7afe
test_1_dvr_router_without_server passes during create-resources but it
fails during verify-resources. The reason is that it cleans up some
resources and tries to create them again, but the fact verify-resources
is called with TOBIKO_PREVENT_CREATE=True prevents their creation
Change-Id: Ided072d74eb2b089d1e23d2823e540ca28b6c234
The output of these commands is printed either in stdout or in stderr
depending on the Openstack version - due to this, some faults tests fail
on later versions
This patch redirects stderr output to stdin
Change-Id: Ic14d69e6a96261d2ba61a11d8ff4dcdb40982bbe
Verifying following parameters:
- Single leader node presents
- There are connections between leader and followers
- Connections are faster then election timer
Change-Id: Ibd4264a435b3ac27dee3ab76ee18244d10a2c1f5