LNST-developers June 2012

lnst-developers@lists.fedorahosted.org

4 participants
47 discussions

[PATCH] Exclude tarball and git directory from generated tarball
by Jan Tluka 06 Jun '12

06 Jun '12

Make tarball smaller excluding git metadata directory and possibly previously generated tarball. --- Common/SlaveUtils.py | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/Common/SlaveUtils.py b/Common/SlaveUtils.py index 3640f94..2c1f043 100644 --- a/Common/SlaveUtils.py +++ b/Common/SlaveUtils.py @@ -27,7 +27,7 @@ def prepare_client_session(host, port, login, passwd=None, command=None, if test_dir is None: test_dir = "lnst" - s = ShellProcess("tar -cjf lnst.tar.bz2 --exclude *.pyc --exclude 'Logs/*' -C '%s' ./" % sys.path[0]) + s = ShellProcess("tar -cjf lnst.tar.bz2 --exclude *.pyc --exclude 'Logs/*' --exclude '.git/*' --exclude 'lnst.tar.bz2' -C '%s' ./" % sys.path[0]) s.wait() scp_to_remote(host, port, login, passwd, "lnst.tar.bz2","/%s/" % (install_path)) -- 1.7.6.5

1 0

[PATCH] Make remote_exec working if executed outside of lnst root directory
by Jan Tluka 06 Jun '12

06 Jun '12

If controller is executed outside of LNST source directory the generated tarball contains content of current working directory instead of the LNST code. The fix is to change the current directory using -C tar option and use of ./ instead of * wildcard. --- Common/SlaveUtils.py | 7 ++++--- 1 files changed, 4 insertions(+), 3 deletions(-) diff --git a/Common/SlaveUtils.py b/Common/SlaveUtils.py index e7fdf3d..3640f94 100644 --- a/Common/SlaveUtils.py +++ b/Common/SlaveUtils.py @@ -8,7 +8,7 @@ jzupka(a)redhat.com (Jiri Zupka) from Common.SshUtils import scp_to_remote, wait_for_login from Common.ShellProcess import ShellProcess - +import sys def prepare_client_session(host, port, login, passwd=None, command=None, prompt=None, install_path=None, test_dir=None): @@ -26,7 +26,8 @@ def prepare_client_session(host, port, login, passwd=None, command=None, install_path = "/tmp" if test_dir is None: test_dir = "lnst" - s = ShellProcess("tar -cjf lnst.tar.bz2 --exclude *.pyc --exclude 'Logs/*' *") + + s = ShellProcess("tar -cjf lnst.tar.bz2 --exclude *.pyc --exclude 'Logs/*' -C '%s' ./" % sys.path[0]) s.wait() scp_to_remote(host, port, login, passwd, "lnst.tar.bz2","/%s/" % (install_path)) @@ -39,4 +40,4 @@ def prepare_client_session(host, port, login, passwd=None, command=None, prompt = "Started" command = "cd /%s/%s/ && ./%s" % (install_path, test_dir, command) return wait_for_login(host, port, login, passwd, prompt, - command=command, timeout=10) \ No newline at end of file + command=command, timeout=10) -- 1.7.6.5

1 0

[PATCH] Do not stop command sequence execution if command failed
by Jan Tluka 06 Jun '12

06 Jun '12

In case of a failed sequence command (even test that did not pass) the whole rest of the command sequence was skipped. This could result in hanging background process that would not be optionally killed at the end of command sequence using type="kill" command. The fix is to let the whole command sequence continue and save any failed command for the overall command sequence result. --- NetTest/NetTestController.py | 5 +++-- 1 files changed, 3 insertions(+), 2 deletions(-) diff --git a/NetTest/NetTestController.py b/NetTest/NetTestController.py index 51a401f..8f46bcb 100644 --- a/NetTest/NetTestController.py +++ b/NetTest/NetTestController.py @@ -176,6 +176,7 @@ class NetTestController: return cmd_res def _run_command_sequence(self, sequence): + seq_passed = True for command in sequence: logging.info("Executing command: [%s]" % str_command(command)) cmd_res = self._run_command(command) @@ -189,8 +190,8 @@ class NetTestController: logging.error("Command failed - command: [%s], " "Error message: \"%s\"" % (str_command(command), cmd_res["err_msg"])) - return False - return True + seq_passed = False + return seq_passed def dump_recipe(self): pprint(self._recipe) -- 1.7.6.5

1 0

[PATCH] Fix system_config bug for non-remote_exec case
by Jan Tluka 06 Jun '12

06 Jun '12

The system_config implementation worked only for remote_exec runs. In case of non-remote test execution the system_config key was missing and lead to unhandled exception. The fix is to move the system_config initialization to netconfig initialization code. --- NetConfig/NetConfigParse.py | 1 + NetTest/NetTestController.py | 1 - 2 files changed, 1 insertions(+), 1 deletions(-) diff --git a/NetConfig/NetConfigParse.py b/NetConfig/NetConfigParse.py index f94a73d..dbd7a6d 100644 --- a/NetConfig/NetConfigParse.py +++ b/NetConfig/NetConfigParse.py @@ -42,6 +42,7 @@ class NetConfigParse: info["rootpass"] = rootpass if rpcport: info["rpcport"] = int(rpcport) + info["system_config"] = {} dom_netdevs = dom_netmachine.getElementsByTagName("netdevice") netdevs = {} diff --git a/NetTest/NetTestController.py b/NetTest/NetTestController.py index 12a871d..51a401f 100644 --- a/NetTest/NetTestController.py +++ b/NetTest/NetTestController.py @@ -58,7 +58,6 @@ class NetTestController: "nettestslave.py") session.add_kill_handler(self._session_die) info["session"] = session - info["system_config"] = {} def _cleanup_slaves(self): for machine_id in self._recipe["machines"]: -- 1.7.6.5

1 0

[PATCH] Iperf test option threshold added
by Jan Tluka 06 Jun '12

06 Jun '12

The new 'threshold' option for TestIperf lets user configure an expected transfer rate that has to be met to make the test pass. Works for single threaded unidirectional test only. The format of the option value string is anything similar to the following: <option name="threshold" value="1000 bits/sec" /> <option name="threshold" value="5.5 kbits/sec" /> <option name="threshold" value="600 Mbits/sec" /> <option name="threshold" value="6.5 Gbits/sec" /> --- Tests/TestIperf.py | 68 +++++++++++++++++++++++++++++++++++++++++++++++++-- 1 files changed, 65 insertions(+), 3 deletions(-) diff --git a/Tests/TestIperf.py b/Tests/TestIperf.py index e4fabbb..cfea973 100644 --- a/Tests/TestIperf.py +++ b/Tests/TestIperf.py @@ -12,6 +12,7 @@ from Common.ExecCmd import exec_cmd from Common.ShellProcess import ShellProcess import time import errno +import re class TestIperf(TestGeneric): def _install_iperf(self): @@ -36,6 +37,54 @@ class TestIperf(TestGeneric): return cmd + def _rate_over_threshold(self, rate): + # convert rate to the same unit as threshold unit + pattern = "(\d*(\.\d*){0,1})\s*([ kMGT])bits\/sec" + + # parse threshold value + r1 = re.match(pattern, self.threshold) + thr_units = r1.group(3) + + # parse measured rate value + r2 = re.match(pattern, rate) + rate_units = r2.group(3) + + if r1.group(2) != None: + thr_val = float(r1.group(1) + r1.group(2)) + else: + thr_val = float(r1.group(1)) + + if r2.group(2) != None: + rate_val = float(r2.group(1) + r2.group(2)) + else: + rate_val = float(r2.group(1)) + + # do the conversion of rate units + if thr_units != rate_units: + # remove any k,M,G,T from measured rate + if rate_units == 'k': + rate_val *= 1000 + elif rate_units == 'M': + rate_val *= 1000*1000 + elif rate_val == 'G': + rate_val *= 1000*1000*1000 + + # divide by k or M or G if present + if thr_units == 'k': + rate_val /= 1000 + elif thr_units == 'M': + rate_val /= 1000*1000 + elif thr_units == 'G': + rate_val /= 1000*1000*1000 + + if rate_val < thr_val: + logging.info("measured rate is below threshold! " \ + "(measured: %s < threshold: %s)" % \ + (rate, self.threshold)) + return False + + return True + def run_client(self, cmd): client = ShellProcess(cmd) try: @@ -44,8 +93,17 @@ class TestIperf(TestGeneric): # we got interrupted, let's gather data if e.errno == errno.EINTR: client.kill() - - client.read_nonblocking() + + output = client.read_nonblocking() + if self.threshold is not None: + # check if expected threshold is reached + m = re.search("\[[^0-9]*[0-9]*\]\s*0.0-\d*\.\d sec\s*\d*(\.\d*){0,1}\s*[ kGMT]Bytes\s*(\d*(\.\d*){0,1}\s*[ kGMT]bits\/sec)", output) + if m is None: + logging.info("Could not get performance throughput!") + return False + + rate = m.group(2) + return self._rate_over_threshold(rate) def run_server(self, cmd): server = ShellProcess(cmd) @@ -83,6 +141,8 @@ class TestIperf(TestGeneric): else: self._temp_dir = installed + self.threshold = self.get_opt("threshold") + role = self.get_mopt("role") cmd = self._compose_iperf_cmd(role) logging.debug("compiled command: %s" % cmd) @@ -92,6 +152,8 @@ class TestIperf(TestGeneric): self.run_server(cmd) elif role == "client": logging.debug("running as client ...") - self.run_client(cmd) + rv = self.run_client(cmd) + if rv == False: + return self.set_fail("iperf test failed, measured rate is below expected threshold") return self.set_pass() -- 1.7.6.5

1 0

[PATCH] Let count of packets in PktgenTx test be configurable
by Jan Tluka 01 Jun '12

01 Jun '12

This patch lets user configure the count of packets to sent by pktgen facility. The default behaviour has not been changed. --- Tests/TestPktgenTx.py | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-) diff --git a/Tests/TestPktgenTx.py b/Tests/TestPktgenTx.py index 5ec500f..792a991 100644 --- a/Tests/TestPktgenTx.py +++ b/Tests/TestPktgenTx.py @@ -58,6 +58,7 @@ class TestPktgenTx(TestGeneric): hwaddr = self.get_mopt("hwaddr") vlan_tci = self.get_opt("vlan_tci", default=0) skb_clone = self.get_opt("skb_clone", default=100000) + count = self.get_opt("count", default=10000000) exec_cmd("modprobe pktgen") @@ -74,7 +75,7 @@ class TestPktgenTx(TestGeneric): pg.set("dst_mac %s" % hwaddr) if vlan_tci: pg.set("vlan_id %d" % vlan_tci) - pg.set("count 10000000") + pg.set("count %d" % count) pgctl.set("start") except ExecCmdFail: return self.set_fail("pktgen failed") -- 1.7.6.5

1 0

Obtaining controller interface IP
by Radek Pazdera 01 Jun '12

01 Jun '12

Hi, I looked further into the problem of obtaining IP address of the controler interface from the guest VM. And I'm not sure what's the best solution. There are the following options: 1) ARP table lookup I tried this manually, I booted up a VM which was connected to a bridge together with a physical interface of my laptop connected to RedHat network. The machine requested an IP, got it, but a record didn't show up in the ARP table at all. Am I doing something wrong? There's also a script that is a part of `arping' which does this reverse IP resolution using ARP [1]. But this solution seems pretty bad to me. They try to ping every possible address on the subnet and then see if the MAC's match. [1] http://pastebin.test.redhat.com/91334 2) Sniffing DHCP replies This is (if I understand it correctly) the way autotest does it. The code is here [2] (thanks to jzupka for helping me out!). I think this might not be suitable for LNST, because the tcpdump needs to be running during the guest startup. LNST is designed to work with machines that are already running. [2] https://github.com/autotest/autotest/blob/master/client/virt/virt_env_proce… 3) Setup through serial console I dug into the beaker task that handles these virtual installs [6] and they have script for obtaining a hostname of the guest [3]. It's a `expect' script that uses serial console via `virsh console' command. Works ok, but it assumes that the guest systems' language is English (which might be true in beaker, but outside of that it's not very nice). Also libvirt's python bindings can open a serial console connection directly in python [4], which could be useful. Here's example of it's usage from libvirt [5]. [3] http://pastebin.test.redhat.com/91348 [4] http://libvirt.org/html/libvirt-libvirt.html#virDomainOpenConsole [5] http://pastebin.test.redhat.com/91337 [6] https://beaker.engineering.redhat.com/tasks/5436 What do you think is the best way to do this? Maybe there's another way I'm not seeing ... Thanks, Radek

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

LNST-developers June 2012