ceph/qa/tasks/rook.py
Patrick Donnelly 6739e11563
qa: time log compression
For debugging and ad-hoc analytics.

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
2023-03-27 14:55:16 -04:00

661 lines
21 KiB
Python

"""
Rook cluster task
"""
import argparse
import configobj
import contextlib
import json
import logging
import os
import yaml
from io import BytesIO
from tarfile import ReadError
from tasks.ceph_manager import CephManager
from teuthology import misc as teuthology
from teuthology.config import config as teuth_config
from teuthology.contextutil import safe_while
from teuthology.orchestra import run
from teuthology import contextutil
from tasks.ceph import healthy
from tasks.cephadm import update_archive_setting
log = logging.getLogger(__name__)
def path_to_examples(ctx, cluster_name : str) -> str:
for p in ['rook/deploy/examples/', 'rook/cluster/examples/kubernetes/ceph/']:
try:
ctx.rook[cluster_name].remote.get_file(p + 'operator.yaml')
return p
except:
pass
assert False, 'Path to examples not found'
def _kubectl(ctx, config, args, **kwargs):
cluster_name = config.get('cluster', 'ceph')
return ctx.rook[cluster_name].remote.run(
args=['kubectl'] + args,
**kwargs
)
def shell(ctx, config):
"""
Run command(s) inside the rook tools container.
tasks:
- kubeadm:
- rook:
- rook.shell:
- ceph -s
or
tasks:
- kubeadm:
- rook:
- rook.shell:
commands:
- ceph -s
"""
if isinstance(config, list):
config = {'commands': config}
for cmd in config.get('commands', []):
if isinstance(cmd, str):
_shell(ctx, config, cmd.split(' '))
else:
_shell(ctx, config, cmd)
def _shell(ctx, config, args, **kwargs):
cluster_name = config.get('cluster', 'ceph')
return _kubectl(
ctx, config,
[
'-n', 'rook-ceph',
'exec',
ctx.rook[cluster_name].toolbox, '--'
] + args,
**kwargs
)
@contextlib.contextmanager
def rook_operator(ctx, config):
cluster_name = config['cluster']
rook_branch = config.get('rook_branch', 'master')
rook_git_url = config.get('rook_git_url', 'https://github.com/rook/rook')
log.info(f'Cloning {rook_git_url} branch {rook_branch}')
ctx.rook[cluster_name].remote.run(
args=[
'rm', '-rf', 'rook',
run.Raw('&&'),
'git',
'clone',
'--single-branch',
'--branch', rook_branch,
rook_git_url,
'rook',
]
)
# operator.yaml
log.info(os.path.abspath(os.getcwd()))
object_methods = [method_name for method_name in dir(ctx.rook[cluster_name].remote)
if callable(getattr(ctx.rook[cluster_name].remote, method_name))]
log.info(object_methods)
operator_yaml = ctx.rook[cluster_name].remote.read_file(
(path_to_examples(ctx, cluster_name) + 'operator.yaml')
)
rook_image = config.get('rook_image')
if rook_image:
log.info(f'Patching operator to use image {rook_image}')
crs = list(yaml.load_all(operator_yaml, Loader=yaml.FullLoader))
assert len(crs) == 2
crs[1]['spec']['template']['spec']['containers'][0]['image'] = rook_image
operator_yaml = yaml.dump_all(crs)
ctx.rook[cluster_name].remote.write_file('operator.yaml', operator_yaml)
op_job = None
try:
log.info('Deploying operator')
_kubectl(ctx, config, [
'create',
'-f', (path_to_examples(ctx, cluster_name) + 'crds.yaml'),
'-f', (path_to_examples(ctx, cluster_name) + 'common.yaml'),
'-f', 'operator.yaml',
])
# on centos:
if teuthology.get_distro(ctx) == 'centos':
_kubectl(ctx, config, [
'-n', 'rook-ceph',
'set', 'env', 'deploy/rook-ceph-operator',
'ROOK_HOSTPATH_REQUIRES_PRIVILEGED=true'
])
# wait for operator
op_name = None
with safe_while(sleep=10, tries=90, action="wait for operator") as proceed:
while not op_name and proceed():
p = _kubectl(
ctx, config,
['-n', 'rook-ceph', 'get', 'pods', '-l', 'app=rook-ceph-operator'],
stdout=BytesIO(),
)
for line in p.stdout.getvalue().decode('utf-8').strip().splitlines():
name, ready, status, _ = line.split(None, 3)
if status == 'Running':
op_name = name
break
# log operator output
op_job = _kubectl(
ctx,
config,
['-n', 'rook-ceph', 'logs', '-f', op_name],
wait=False,
logger=log.getChild('operator'),
)
yield
except Exception as e:
log.exception(e)
raise
finally:
log.info('Cleaning up rook operator')
_kubectl(ctx, config, [
'delete',
'-f', 'operator.yaml',
])
if False:
# don't bother since we'll tear down k8s anyway (and this mysteriously
# fails sometimes when deleting some of the CRDs... not sure why!)
_kubectl(ctx, config, [
'delete',
'-f', (path_to_examples() + 'common.yaml'),
])
_kubectl(ctx, config, [
'delete',
'-f', (path_to_examples() + 'crds.yaml'),
])
ctx.rook[cluster_name].remote.run(args=['rm', '-rf', 'rook', 'operator.yaml'])
if op_job:
op_job.wait()
run.wait(
ctx.cluster.run(
args=[
'sudo', 'rm', '-rf', '/var/lib/rook'
]
)
)
@contextlib.contextmanager
def ceph_log(ctx, config):
cluster_name = config['cluster']
log_dir = '/var/lib/rook/rook-ceph/log'
update_archive_setting(ctx, 'log', log_dir)
try:
yield
except Exception:
# we need to know this below
ctx.summary['success'] = False
raise
finally:
log.info('Checking cluster log for badness...')
def first_in_ceph_log(pattern, excludes):
"""
Find the first occurrence of the pattern specified in the Ceph log,
Returns None if none found.
:param pattern: Pattern scanned for.
:param excludes: Patterns to ignore.
:return: First line of text (or None if not found)
"""
args = [
'sudo',
'egrep', pattern,
f'{log_dir}/ceph.log',
]
if excludes:
for exclude in excludes:
args.extend([run.Raw('|'), 'egrep', '-v', exclude])
args.extend([
run.Raw('|'), 'head', '-n', '1',
])
r = ctx.rook[cluster_name].remote.run(
stdout=BytesIO(),
args=args,
)
stdout = r.stdout.getvalue().decode()
if stdout:
return stdout
return None
if first_in_ceph_log('\[ERR\]|\[WRN\]|\[SEC\]',
config.get('log-ignorelist')) is not None:
log.warning('Found errors (ERR|WRN|SEC) in cluster log')
ctx.summary['success'] = False
# use the most severe problem as the failure reason
if 'failure_reason' not in ctx.summary:
for pattern in ['\[SEC\]', '\[ERR\]', '\[WRN\]']:
match = first_in_ceph_log(pattern, config['log-ignorelist'])
if match is not None:
ctx.summary['failure_reason'] = \
'"{match}" in cluster log'.format(
match=match.rstrip('\n'),
)
break
if ctx.archive is not None and \
not (ctx.config.get('archive-on-error') and ctx.summary['success']):
# and logs
log.info('Compressing logs...')
run.wait(
ctx.cluster.run(
args=[
'time',
'sudo',
'find',
log_dir,
'-name',
'*.log',
'-print0',
run.Raw('|'),
'sudo',
'xargs',
'--max-args=1',
'--max-procs=0',
'--verbose',
'-0',
'--no-run-if-empty',
'--',
'gzip',
'-5',
'--verbose',
'--',
],
wait=False,
),
)
log.info('Archiving logs...')
path = os.path.join(ctx.archive, 'remote')
try:
os.makedirs(path)
except OSError:
pass
for remote in ctx.cluster.remotes.keys():
sub = os.path.join(path, remote.name)
try:
os.makedirs(sub)
except OSError:
pass
try:
teuthology.pull_directory(remote, log_dir,
os.path.join(sub, 'log'))
except ReadError:
pass
def build_initial_config(ctx, config):
path = os.path.join(os.path.dirname(__file__), 'rook-ceph.conf')
conf = configobj.ConfigObj(path, file_error=True)
# overrides
for section, keys in config.get('conf',{}).items():
for key, value in keys.items():
log.info(" override: [%s] %s = %s" % (section, key, value))
if section not in conf:
conf[section] = {}
conf[section][key] = value
return conf
@contextlib.contextmanager
def rook_cluster(ctx, config):
cluster_name = config['cluster']
# count how many OSDs we'll create
num_devs = 0
num_hosts = 0
for remote in ctx.cluster.remotes.keys():
ls = remote.read_file('/scratch_devs').decode('utf-8').strip().splitlines()
num_devs += len(ls)
num_hosts += 1
ctx.rook[cluster_name].num_osds = num_devs
# config
ceph_conf = build_initial_config(ctx, config)
ceph_conf_fp = BytesIO()
ceph_conf.write(ceph_conf_fp)
log.info(f'Config:\n{ceph_conf_fp.getvalue()}')
_kubectl(ctx, ceph_conf, ['create', '-f', '-'], stdin=yaml.dump({
'apiVersion': 'v1',
'kind': 'ConfigMap',
'metadata': {
'name': 'rook-config-override',
'namespace': 'rook-ceph'},
'data': {
'config': ceph_conf_fp.getvalue()
}
}))
# cluster
cluster = {
'apiVersion': 'ceph.rook.io/v1',
'kind': 'CephCluster',
'metadata': {'name': 'rook-ceph', 'namespace': 'rook-ceph'},
'spec': {
'cephVersion': {
'image': ctx.rook[cluster_name].image,
'allowUnsupported': True,
},
'dataDirHostPath': '/var/lib/rook',
'skipUpgradeChecks': True,
'mgr': {
'count': 1,
'modules': [
{ 'name': 'rook', 'enabled': True },
],
},
'mon': {
'count': num_hosts,
'allowMultiplePerNode': True,
},
'storage': {
'storageClassDeviceSets': [
{
'name': 'scratch',
'count': num_devs,
'portable': False,
'volumeClaimTemplates': [
{
'metadata': {'name': 'data'},
'spec': {
'resources': {
'requests': {
'storage': '10Gi' # <= (lte) the actual PV size
}
},
'storageClassName': 'scratch',
'volumeMode': 'Block',
'accessModes': ['ReadWriteOnce'],
},
},
],
}
],
},
}
}
teuthology.deep_merge(cluster['spec'], config.get('spec', {}))
cluster_yaml = yaml.dump(cluster)
log.info(f'Cluster:\n{cluster_yaml}')
try:
ctx.rook[cluster_name].remote.write_file('cluster.yaml', cluster_yaml)
_kubectl(ctx, config, ['create', '-f', 'cluster.yaml'])
yield
except Exception as e:
log.exception(e)
raise
finally:
_kubectl(ctx, config, ['delete', '-f', 'cluster.yaml'], check_status=False)
# wait for cluster to shut down
log.info('Waiting for cluster to stop')
running = True
with safe_while(sleep=5, tries=100, action="wait for teardown") as proceed:
while running and proceed():
p = _kubectl(
ctx, config,
['-n', 'rook-ceph', 'get', 'pods'],
stdout=BytesIO(),
)
running = False
for line in p.stdout.getvalue().decode('utf-8').strip().splitlines():
name, ready, status, _ = line.split(None, 3)
if (
name != 'NAME'
and not name.startswith('csi-')
and not name.startswith('rook-ceph-operator-')
and not name.startswith('rook-ceph-tools-')
):
running = True
break
_kubectl(
ctx, config,
['-n', 'rook-ceph', 'delete', 'configmap', 'rook-config-override'],
check_status=False,
)
ctx.rook[cluster_name].remote.run(args=['rm', '-f', 'cluster.yaml'])
@contextlib.contextmanager
def rook_toolbox(ctx, config):
cluster_name = config['cluster']
try:
_kubectl(ctx, config, [
'create',
'-f', (path_to_examples(ctx, cluster_name) + 'toolbox.yaml'),
])
log.info('Waiting for tools container to start')
toolbox = None
with safe_while(sleep=5, tries=100, action="wait for toolbox") as proceed:
while not toolbox and proceed():
p = _kubectl(
ctx, config,
['-n', 'rook-ceph', 'get', 'pods', '-l', 'app=rook-ceph-tools'],
stdout=BytesIO(),
)
_kubectl(
ctx, config,
['-n', 'rook-ceph', 'get', 'pods'],
stdout=BytesIO(),
)
for line in p.stdout.getvalue().decode('utf-8').strip().splitlines():
name, ready, status, _ = line.split(None, 3)
if status == 'Running':
toolbox = name
break
ctx.rook[cluster_name].toolbox = toolbox
yield
except Exception as e:
log.exception(e)
raise
finally:
_kubectl(ctx, config, [
'delete',
'-f', (path_to_examples(ctx, cluster_name) + 'toolbox.yaml'),
], check_status=False)
@contextlib.contextmanager
def wait_for_osds(ctx, config):
cluster_name = config.get('cluster', 'ceph')
want = ctx.rook[cluster_name].num_osds
log.info(f'Waiting for {want} OSDs')
with safe_while(sleep=10, tries=90, action="check osd count") as proceed:
while proceed():
p = _shell(ctx, config, ['ceph', 'osd', 'stat', '-f', 'json'],
stdout=BytesIO(),
check_status=False)
if p.exitstatus == 0:
r = json.loads(p.stdout.getvalue().decode('utf-8'))
have = r.get('num_up_osds', 0)
if have == want:
break
log.info(f' have {have}/{want} OSDs')
yield
@contextlib.contextmanager
def ceph_config_keyring(ctx, config):
# get config and push to hosts
log.info('Distributing ceph config and client.admin keyring')
p = _shell(ctx, config, ['cat', '/etc/ceph/ceph.conf'], stdout=BytesIO())
conf = p.stdout.getvalue()
p = _shell(ctx, config, ['cat', '/etc/ceph/keyring'], stdout=BytesIO())
keyring = p.stdout.getvalue()
ctx.cluster.run(args=['sudo', 'mkdir', '-p', '/etc/ceph'])
for remote in ctx.cluster.remotes.keys():
remote.write_file(
'/etc/ceph/ceph.conf',
conf,
sudo=True,
)
remote.write_file(
'/etc/ceph/keyring',
keyring,
sudo=True,
)
try:
yield
except Exception as e:
log.exception(e)
raise
finally:
log.info('Cleaning up config and client.admin keyring')
ctx.cluster.run(args=[
'sudo', 'rm', '-f',
'/etc/ceph/ceph.conf',
'/etc/ceph/ceph.client.admin.keyring'
])
@contextlib.contextmanager
def ceph_clients(ctx, config):
cluster_name = config['cluster']
log.info('Setting up client nodes...')
clients = ctx.cluster.only(teuthology.is_type('client', cluster_name))
for remote, roles_for_host in clients.remotes.items():
for role in teuthology.cluster_roles_of_type(roles_for_host, 'client',
cluster_name):
name = teuthology.ceph_role(role)
client_keyring = '/etc/ceph/{0}.{1}.keyring'.format(cluster_name,
name)
r = _shell(ctx, config,
args=[
'ceph', 'auth',
'get-or-create', name,
'mon', 'allow *',
'osd', 'allow *',
'mds', 'allow *',
'mgr', 'allow *',
],
stdout=BytesIO(),
)
keyring = r.stdout.getvalue()
remote.write_file(client_keyring, keyring, sudo=True, mode='0644')
yield
@contextlib.contextmanager
def task(ctx, config):
"""
Deploy rook-ceph cluster
tasks:
- kubeadm:
- rook:
branch: wip-foo
spec:
mon:
count: 1
The spec item is deep-merged against the cluster.yaml. The branch, sha1, or
image items are used to determine the Ceph container image.
"""
if not config:
config = {}
assert isinstance(config, dict), \
"task only supports a dictionary for configuration"
log.info('Rook start')
overrides = ctx.config.get('overrides', {})
teuthology.deep_merge(config, overrides.get('ceph', {}))
teuthology.deep_merge(config, overrides.get('rook', {}))
log.info('Config: ' + str(config))
# set up cluster context
if not hasattr(ctx, 'rook'):
ctx.rook = {}
if 'cluster' not in config:
config['cluster'] = 'ceph'
cluster_name = config['cluster']
if cluster_name not in ctx.rook:
ctx.rook[cluster_name] = argparse.Namespace()
ctx.rook[cluster_name].remote = list(ctx.cluster.remotes.keys())[0]
# image
teuth_defaults = teuth_config.get('defaults', {})
cephadm_defaults = teuth_defaults.get('cephadm', {})
containers_defaults = cephadm_defaults.get('containers', {})
container_image_name = containers_defaults.get('image', None)
if 'image' in config:
ctx.rook[cluster_name].image = config.get('image')
else:
sha1 = config.get('sha1')
flavor = config.get('flavor', 'default')
if sha1:
if flavor == "crimson":
ctx.rook[cluster_name].image = container_image_name + ':' + sha1 + '-' + flavor
else:
ctx.rook[cluster_name].image = container_image_name + ':' + sha1
else:
# hmm, fall back to branch?
branch = config.get('branch', 'master')
ctx.rook[cluster_name].image = container_image_name + ':' + branch
log.info('Ceph image is %s' % ctx.rook[cluster_name].image)
with contextutil.nested(
lambda: rook_operator(ctx, config),
lambda: ceph_log(ctx, config),
lambda: rook_cluster(ctx, config),
lambda: rook_toolbox(ctx, config),
lambda: wait_for_osds(ctx, config),
lambda: ceph_config_keyring(ctx, config),
lambda: ceph_clients(ctx, config),
):
if not hasattr(ctx, 'managers'):
ctx.managers = {}
ctx.managers[cluster_name] = CephManager(
ctx.rook[cluster_name].remote,
ctx=ctx,
logger=log.getChild('ceph_manager.' + cluster_name),
cluster=cluster_name,
rook=True,
)
try:
if config.get('wait-for-healthy', True):
healthy(ctx=ctx, config=config)
log.info('Rook complete, yielding')
yield
finally:
log.info('Tearing down rook')