devstack/tools/worlddump.py
Jens Harbott ce396d374b Fix worlddump log collection
All credit for figuring this out goes to frickler (and that was the hard
bit so thank you!). The worlddump files were not being collected because
they weren't in our log collection list. Add worlddump to this list
so that we collect these files.

One thing that makes this slightly complicated is the worlddump files
are named with a timestamp and we can't have globs in our collection
list. To address this we create a copy of the file with a -latest.txt
suffix. This gives us a deterministic file name for log collection
without using globs.

Note we do not use a symlink here because some jobs gzip their log files
(breaking symlinks) and others do not. This makes it painful to always
have a valid link. Not having a valid link can break log collection.

Hardlinks may be another option but simply making a copy is easier to
manage as you don't have to worry about links preexisting and the
dumpfiles are not that large.

Change-Id: I96ae5f5290546ad25ca434c1106c01354d2d053c
2019-09-12 10:21:17 +00:00

267 lines
7.6 KiB
Python
Executable File

#!/usr/bin/env python
#
# Copyright 2014 Hewlett-Packard Development Company, L.P.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
"""Dump the state of the world for post mortem."""
from __future__ import print_function
import argparse
import datetime
from distutils import spawn
import fnmatch
import os
import os.path
import shutil
import subprocess
import sys
GMR_PROCESSES = (
'nova-compute',
'neutron-dhcp-agent',
'neutron-l3-agent',
'neutron-linuxbridge-agent',
'neutron-metadata-agent',
'neutron-openvswitch-agent',
'cinder-volume',
)
def get_options():
parser = argparse.ArgumentParser(
description='Dump world state for debugging')
parser.add_argument('-d', '--dir',
default='.',
help='Output directory for worlddump')
parser.add_argument('-n', '--name',
default='',
help='Additional name to tag into file')
return parser.parse_args()
def filename(dirname, name=""):
now = datetime.datetime.utcnow()
fmt = "worlddump-%Y-%m-%d-%H%M%S"
if name:
fmt += "-" + name
fmt += ".txt"
return os.path.join(dirname, now.strftime(fmt))
def warn(msg):
print("WARN: %s" % msg)
def _dump_cmd(cmd):
print(cmd)
print("-" * len(cmd))
print()
try:
subprocess.check_call(cmd, shell=True)
print()
except subprocess.CalledProcessError as e:
print("*** Failed to run '%(cmd)s': %(err)s" % {'cmd': cmd, 'err': e})
def _find_cmd(cmd):
if not spawn.find_executable(cmd):
print("*** %s not found: skipping" % cmd)
return False
return True
def _header(name):
print()
print(name)
print("=" * len(name))
print()
def _bridge_list():
process = subprocess.Popen(['sudo', 'ovs-vsctl', 'list-br'],
stdout=subprocess.PIPE)
stdout, _ = process.communicate()
return stdout.split()
# This method gets a max openflow version supported by openvswitch.
# For example 'ovs-ofctl --version' displays the following:
#
# ovs-ofctl (Open vSwitch) 2.0.2
# Compiled Dec 9 2015 14:08:08
# OpenFlow versions 0x1:0x4
#
# The above shows that openvswitch supports from OpenFlow10 to OpenFlow13.
# This method gets max version searching 'OpenFlow versions 0x1:0x'.
# And return a version value converted to an integer type.
def _get_ofp_version():
process = subprocess.Popen(['ovs-ofctl', '--version'], stdout=subprocess.PIPE)
stdout, _ = process.communicate()
find_str = 'OpenFlow versions 0x1:0x'
offset = stdout.find(find_str)
return int(stdout[offset + len(find_str):-1]) - 1
def disk_space():
# the df output
_header("File System Summary")
dfraw = os.popen("df -Ph").read()
df = [s.split() for s in dfraw.splitlines()]
for fs in df:
try:
if int(fs[4][:-1]) > 95:
warn("Device %s (%s) is %s full, might be an issue" % (
fs[0], fs[5], fs[4]))
except ValueError:
# if it doesn't look like an int, that's fine
pass
print(dfraw)
def ebtables_dump():
tables = ['filter', 'nat', 'broute']
_header("EB Tables Dump")
if not _find_cmd('ebtables'):
return
for table in tables:
_dump_cmd("sudo ebtables -t %s -L" % table)
def iptables_dump():
tables = ['filter', 'nat', 'mangle']
_header("IP Tables Dump")
for table in tables:
_dump_cmd("sudo iptables --line-numbers -L -nv -t %s" % table)
def _netns_list():
process = subprocess.Popen(['ip', 'netns'], stdout=subprocess.PIPE)
stdout, _ = process.communicate()
# NOTE(jlvillal): Sometimes 'ip netns list' can return output like:
# qrouter-0805fd7d-c493-4fa6-82ca-1c6c9b23cd9e (id: 1)
# qdhcp-bb2cc6ae-2ae8-474f-adda-a94059b872b5 (id: 0)
output = [x.split()[0] for x in stdout.splitlines()]
return output
def network_dump():
_header("Network Dump")
_dump_cmd("bridge link")
_dump_cmd("brctl show")
_dump_cmd("ip link show type bridge")
ip_cmds = ["neigh", "addr", "link", "route"]
for cmd in ip_cmds + ['netns']:
_dump_cmd("ip %s" % cmd)
for netns_ in _netns_list():
for cmd in ip_cmds:
args = {'netns': netns_, 'cmd': cmd}
_dump_cmd('sudo ip netns exec %(netns)s ip %(cmd)s' % args)
def ovs_dump():
_header("Open vSwitch Dump")
# NOTE(cdent): If we're not using neutron + ovs these commands
# will not be present so
if not _find_cmd('ovs-vsctl'):
return
bridges = _bridge_list()
ofctl_cmds = ('show', 'dump-ports-desc', 'dump-ports', 'dump-flows')
ofp_max = _get_ofp_version()
vers = 'OpenFlow10'
for i in range(1, ofp_max + 1):
vers += ',OpenFlow1' + str(i)
_dump_cmd("sudo ovs-vsctl show")
for ofctl_cmd in ofctl_cmds:
for bridge in bridges:
args = {'vers': vers, 'cmd': ofctl_cmd, 'bridge': bridge}
_dump_cmd("sudo ovs-ofctl --protocols=%(vers)s %(cmd)s %(bridge)s" % args)
def process_list():
_header("Process Listing")
_dump_cmd("ps axo "
"user,ppid,pid,pcpu,pmem,vsz,rss,tty,stat,start,time,args")
def compute_consoles():
_header("Compute consoles")
for root, dirnames, filenames in os.walk('/opt/stack'):
for filename in fnmatch.filter(filenames, 'console.log'):
fullpath = os.path.join(root, filename)
_dump_cmd("sudo cat %s" % fullpath)
def guru_meditation_reports():
for service in GMR_PROCESSES:
_header("%s Guru Meditation Report" % service)
try:
subprocess.check_call(['pgrep', '-f', service])
except subprocess.CalledProcessError:
print("Skipping as %s does not appear to be running" % service)
continue
_dump_cmd("killall -e -USR2 %s" % service)
print("guru meditation report in %s log" % service)
def var_core():
if os.path.exists('/var/core'):
_header("/var/core dumps")
# NOTE(ianw) : see DEBUG_LIBVIRT_COREDUMPS. We could think
# about getting backtraces out of these. There are other
# tools out there that can do that sort of thing though.
_dump_cmd("ls -ltrah /var/core")
def main():
opts = get_options()
fname = filename(opts.dir, opts.name)
print("World dumping... see %s for details" % fname)
sys.stdout = os.fdopen(sys.stdout.fileno(), 'w', 0)
with open(fname, 'w') as f:
os.dup2(f.fileno(), sys.stdout.fileno())
disk_space()
process_list()
network_dump()
ovs_dump()
iptables_dump()
ebtables_dump()
compute_consoles()
guru_meditation_reports()
var_core()
# Singular name for ease of log retrieval
copyname = os.path.join(opts.dir, 'worlddump')
if opts.name:
copyname += '-' + opts.name
copyname += '-latest.txt'
# We make a full copy to deal with jobs that may or may not
# gzip logs breaking symlinks.
shutil.copyfile(fname, copyname)
if __name__ == '__main__':
try:
sys.exit(main())
except KeyboardInterrupt:
sys.exit(1)