From: Ondrej Lichtner <olichtne(a)redhat.com&gt; When the --packet-capture argument was specified and tcpdump wasn't installed on one of the machines LNST would fail and report a very nondescriptive exception. This commit adds a is_installed check to the slave method "start_packet_capture" and the generated exception is then handled on the Controller, reporting a more useful error message. Fixes #161. Signed-off-by: Ondrej Lichtner <olichtne(a)redhat.com&gt; --- lnst/Controller/NetTestController.py | 11 +++++++++-- lnst/Slave/NetTestSlave.py | 4 ++++ 2 files changed, 13 insertions(+), 2 deletions(-) diff --git a/lnst/Controller/NetTestController.py b/lnst/Controller/NetTestController.py index 889dfa7..791dd95 100644 --- a/lnst/Controller/NetTestController.py +++ b/lnst/Controller/NetTestController.py @@ -651,8 +651,15 @@ class NetTestController: self._cleanup_slaves() raise - if self._packet_capture: - self._start_packet_capture() + try: + if self._packet_capture: + self._start_packet_capture() + except Exception as exc: + logging.error("Couldn't start packet capture.") + logging.error(str(exc)) + self._cleanup_slaves() + return {"passed": False, + "err_msg": str(exc)} err = None try: diff --git a/lnst/Slave/NetTestSlave.py b/lnst/Slave/NetTestSlave.py index 8d8fada..b18f47d 100644 --- a/lnst/Slave/NetTestSlave.py +++ b/lnst/Slave/NetTestSlave.py @@ -31,6 +31,7 @@ from lnst.Common.NetTestCommand import NetTestCommandContext from lnst.Common.NetTestCommand import NetTestCommand from lnst.Common.NetTestCommand import DEFAULT_TIMEOUT from lnst.Common.Utils import check_process_running +from lnst.Common.Utils import is_installed from lnst.Common.ConnectionHandler import send_data from lnst.Common.ConnectionHandler import ConnectionHandler from lnst.Common.Config import lnst_config @@ -359,6 +360,9 @@ class SlaveMethods: return True def start_packet_capture(self, filt): + if not is_installed("tcpdump"): + raise Exception("Can't start packet capture, tcpdump not available") + files = {} for if_id, dev in self._if_manager.get_mapped_devices().iteritems(): if dev.get_netns() != None: -- 2.9.0

From: Ondrej Lichtner <olichtne(a)redhat.com&gt; When there were no options specified for an ovs internal port, the slave configuration would fail, with an exception, trying to access a nonexistant dictionary key. This could be fixe both on the controller side by sending an empty list of options from the XML parser or on the slave by checking if the key is defined. However, since we're soon going to move to pure python recipes I decided to fix it just on the slave since it's sufficient and the problem won't reappear after the Python recipes switch. Fixes #175. Signed-off-by: Ondrej Lichtner <olichtne(a)redhat.com&gt; --- lnst/Slave/NetConfigDevice.py | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/lnst/Slave/NetConfigDevice.py b/lnst/Slave/NetConfigDevice.py index a0fa0e1..89d471e 100644 --- a/lnst/Slave/NetConfigDevice.py +++ b/lnst/Slave/NetConfigDevice.py @@ -518,11 +518,12 @@ class NetConfigDeviceOvsBridge(NetConfigDeviceGeneric): i["name"] = self._if_manager.assign_name_generic(prefix="int") options = "" - for opt in i["options"]: - options += " %s=%s" % (opt["name"], opt["value"]) + if "options" in i: + for opt in i["options"]: + options += " %s=%s" % (opt["name"], opt["value"]) - if opt["name"] == "name": - i["name"] = opt["value"] + if opt["name"] == "name": + i["name"] = opt["value"] exec_cmd("ovs-vsctl add-port %s %s -- set Interface %s "\ "type=internal %s" % (br_name, i["name"], -- 2.9.0