summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--.tito/packages/openshift-ansible2
-rw-r--r--openshift-ansible.spec8
-rw-r--r--playbooks/byo/openshift-cluster/config.yml13
-rw-r--r--playbooks/common/openshift-checks/health.yml11
-rw-r--r--playbooks/common/openshift-checks/pre-install.yml11
-rw-r--r--playbooks/common/openshift-cluster/validate_hostnames.yml4
-rw-r--r--roles/lib_openshift/library/oc_process.py2
-rw-r--r--roles/lib_openshift/src/class/oc_process.py2
-rw-r--r--roles/openshift_health_checker/action_plugins/openshift_health_check.py26
-rw-r--r--roles/openshift_health_checker/callback_plugins/zz_failure_summary.py87
-rw-r--r--roles/openshift_health_checker/openshift_checks/disk_availability.py6
-rw-r--r--roles/openshift_health_checker/openshift_checks/memory_availability.py6
-rw-r--r--roles/openshift_health_checker/test/action_plugin_test.py16
-rw-r--r--roles/openshift_health_checker/test/disk_availability_test.py36
-rw-r--r--roles/openshift_health_checker/test/memory_availability_test.py43
-rw-r--r--roles/openshift_logging/filter_plugins/openshift_logging.py2
-rw-r--r--roles/openshift_storage_glusterfs/filter_plugins/openshift_storage_glusterfs.py2
17 files changed, 213 insertions, 64 deletions
diff --git a/.tito/packages/openshift-ansible b/.tito/packages/openshift-ansible
index 0de85441a..8e2a52845 100644
--- a/.tito/packages/openshift-ansible
+++ b/.tito/packages/openshift-ansible
@@ -1 +1 @@
-3.6.82-1 ./
+3.6.84-1 ./
diff --git a/openshift-ansible.spec b/openshift-ansible.spec
index b039bf33e..06245f55c 100644
--- a/openshift-ansible.spec
+++ b/openshift-ansible.spec
@@ -9,7 +9,7 @@
%global __requires_exclude ^/usr/bin/ansible-playbook$
Name: openshift-ansible
-Version: 3.6.82
+Version: 3.6.84
Release: 1%{?dist}
Summary: Openshift and Atomic Enterprise Ansible
License: ASL 2.0
@@ -280,6 +280,12 @@ Atomic OpenShift Utilities includes
%changelog
+* Wed May 24 2017 Jenkins CD Merge Bot <tdawson@redhat.com> 3.6.84-1
+- oc_process: Better error output on failed template() call (jarrpa@redhat.com)
+
+* Wed May 24 2017 Jenkins CD Merge Bot <tdawson@redhat.com> 3.6.83-1
+- Allow a hostname to resolve to 127.0.0.1 during validation (dms@redhat.com)
+
* Wed May 24 2017 Jenkins CD Merge Bot <tdawson@redhat.com> 3.6.82-1
- Fixing tux warnings and some final clean up (ewolinet@redhat.com)
- Appease travis (sdodson@redhat.com)
diff --git a/playbooks/byo/openshift-cluster/config.yml b/playbooks/byo/openshift-cluster/config.yml
index acf5469bf..fd4a9eb26 100644
--- a/playbooks/byo/openshift-cluster/config.yml
+++ b/playbooks/byo/openshift-cluster/config.yml
@@ -3,6 +3,19 @@
tags:
- always
+- name: Verify Requirements
+ hosts: OSEv3
+ roles:
+ - openshift_health_checker
+ vars:
+ - r_openshift_health_checker_playbook_context: "install"
+ post_tasks:
+ - action: openshift_health_check
+ args:
+ checks:
+ - disk_availability
+ - memory_availability
+
- include: ../../common/openshift-cluster/std_include.yml
tags:
- always
diff --git a/playbooks/common/openshift-checks/health.yml b/playbooks/common/openshift-checks/health.yml
index fc0f523d5..1bee460e8 100644
--- a/playbooks/common/openshift-checks/health.yml
+++ b/playbooks/common/openshift-checks/health.yml
@@ -2,9 +2,10 @@
- name: Run OpenShift health checks
hosts: OSEv3
roles:
- - openshift_health_checker
+ - openshift_health_checker
+ vars:
+ - r_openshift_health_checker_playbook_context: "health"
post_tasks:
- - action: openshift_health_check # https://github.com/ansible/ansible/issues/20513
- args:
- checks:
- - '@health'
+ - action: openshift_health_check # https://github.com/ansible/ansible/issues/20513
+ args:
+ checks: ['@health']
diff --git a/playbooks/common/openshift-checks/pre-install.yml b/playbooks/common/openshift-checks/pre-install.yml
index c8ffc3d91..e01c6f38d 100644
--- a/playbooks/common/openshift-checks/pre-install.yml
+++ b/playbooks/common/openshift-checks/pre-install.yml
@@ -2,9 +2,10 @@
- hosts: OSEv3
name: run OpenShift pre-install checks
roles:
- - openshift_health_checker
+ - openshift_health_checker
+ vars:
+ - r_openshift_health_checker_playbook_context: "pre-install"
post_tasks:
- - action: openshift_health_check # https://github.com/ansible/ansible/issues/20513
- args:
- checks:
- - '@preflight'
+ - action: openshift_health_check # https://github.com/ansible/ansible/issues/20513
+ args:
+ checks: ['@preflight']
diff --git a/playbooks/common/openshift-cluster/validate_hostnames.yml b/playbooks/common/openshift-cluster/validate_hostnames.yml
index 48cc03b19..33fc5630f 100644
--- a/playbooks/common/openshift-cluster/validate_hostnames.yml
+++ b/playbooks/common/openshift-cluster/validate_hostnames.yml
@@ -13,4 +13,6 @@
pause:
prompt: "The hostname \"{{ openshift.common.hostname }}\" for \"{{ ansible_nodename }}\" doesn't resolve to an ip address owned by this host. Please set openshift_hostname variable to a hostname that when resolved on the host in question resolves to an IP address matching an interface on this host. This host will fail liveness checks for pods utilizing hostPorts, press ENTER to continue or CTRL-C to abort."
seconds: "{{ 10 if openshift_override_hostname_check | default(false) | bool else omit }}"
- when: lookupip.stdout not in ansible_all_ipv4_addresses
+ when:
+ - lookupip.stdout != '127.0.0.1'
+ - lookupip.stdout not in ansible_all_ipv4_addresses
diff --git a/roles/lib_openshift/library/oc_process.py b/roles/lib_openshift/library/oc_process.py
index 330de07eb..15e9c606d 100644
--- a/roles/lib_openshift/library/oc_process.py
+++ b/roles/lib_openshift/library/oc_process.py
@@ -1450,7 +1450,7 @@ class OCProcess(OpenShiftCLI):
if self._template is None:
results = self._process(self.name, False, self.params, self.data)
if results['returncode'] != 0:
- raise OpenShiftCLIError('Error processing template [%s].' % self.name)
+ raise OpenShiftCLIError('Error processing template [%s]: %s' %(self.name, results))
self._template = results['results']['items']
return self._template
diff --git a/roles/lib_openshift/src/class/oc_process.py b/roles/lib_openshift/src/class/oc_process.py
index eba9a43cd..62a6bd571 100644
--- a/roles/lib_openshift/src/class/oc_process.py
+++ b/roles/lib_openshift/src/class/oc_process.py
@@ -30,7 +30,7 @@ class OCProcess(OpenShiftCLI):
if self._template is None:
results = self._process(self.name, False, self.params, self.data)
if results['returncode'] != 0:
- raise OpenShiftCLIError('Error processing template [%s].' % self.name)
+ raise OpenShiftCLIError('Error processing template [%s]: %s' %(self.name, results))
self._template = results['results']['items']
return self._template
diff --git a/roles/openshift_health_checker/action_plugins/openshift_health_check.py b/roles/openshift_health_checker/action_plugins/openshift_health_check.py
index 03c40b78b..a62e4331e 100644
--- a/roles/openshift_health_checker/action_plugins/openshift_health_check.py
+++ b/roles/openshift_health_checker/action_plugins/openshift_health_check.py
@@ -25,9 +25,11 @@ class ActionModule(ActionBase):
def run(self, tmp=None, task_vars=None):
result = super(ActionModule, self).run(tmp, task_vars)
+ task_vars = task_vars or {}
- if task_vars is None:
- task_vars = {}
+ # vars are not supportably available in the callback plugin,
+ # so record any it will need in the result.
+ result['playbook_context'] = task_vars.get('r_openshift_health_checker_playbook_context')
if "openshift" not in task_vars:
result["failed"] = True
@@ -46,19 +48,27 @@ class ActionModule(ActionBase):
result["checks"] = check_results = {}
+ user_disabled_checks = [
+ check.strip()
+ for check in task_vars.get("openshift_disable_check", "").split(",")
+ ]
+
for check_name in resolved_checks:
display.banner("CHECK [{} : {}]".format(check_name, task_vars["ansible_host"]))
check = known_checks[check_name]
- if check.is_active(task_vars):
+ if not check.is_active(task_vars):
+ r = dict(skipped=True, skipped_reason="Not active for this host")
+ elif check_name in user_disabled_checks:
+ r = dict(skipped=True, skipped_reason="Disabled by user request")
+ else:
try:
r = check.run(tmp, task_vars)
except OpenShiftCheckException as e:
- r = {}
- r["failed"] = True
- r["msg"] = str(e)
- else:
- r = {"skipped": True}
+ r = dict(
+ failed=True,
+ msg=str(e),
+ )
check_results[check_name] = r
diff --git a/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py b/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py
index 7bce7f107..64c29a8d9 100644
--- a/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py
+++ b/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py
@@ -2,6 +2,12 @@
Ansible callback plugin.
'''
+# Reason: In several locations below we disable pylint protected-access
+# for Ansible objects that do not give us any public way
+# to access the full details we need to report check failures.
+# Status: disabled permanently or until Ansible object has a public API.
+# This does leave the code more likely to be broken by future Ansible changes.
+
from pprint import pformat
from ansible.plugins.callback import CallbackBase
@@ -20,38 +26,37 @@ class CallbackModule(CallbackBase):
CALLBACK_TYPE = 'aggregate'
CALLBACK_NAME = 'failure_summary'
CALLBACK_NEEDS_WHITELIST = False
+ _playbook_file = None
def __init__(self):
super(CallbackModule, self).__init__()
self.__failures = []
+ def v2_playbook_on_start(self, playbook):
+ super(CallbackModule, self).v2_playbook_on_start(playbook)
+ # re: playbook attrs see top comment # pylint: disable=protected-access
+ self._playbook_file = playbook._file_name
+
def v2_runner_on_failed(self, result, ignore_errors=False):
super(CallbackModule, self).v2_runner_on_failed(result, ignore_errors)
self.__failures.append(dict(result=result, ignore_errors=ignore_errors))
def v2_playbook_on_stats(self, stats):
super(CallbackModule, self).v2_playbook_on_stats(stats)
- # TODO: update condition to consider a host var or env var to
- # enable/disable the summary, so that we can control the output from a
- # play.
if self.__failures:
- self._print_failure_summary()
+ self._print_failure_details(self.__failures)
- def _print_failure_summary(self):
- '''Print a summary of failed tasks (including ignored failures).'''
+ def _print_failure_details(self, failures):
+ '''Print a summary of failed tasks or checks.'''
self._display.display(u'\nFailure summary:\n')
- # TODO: group failures by host or by task. If grouped by host, it is
- # easy to see all problems of a given host. If grouped by task, it is
- # easy to see what hosts needs the same fix.
-
- width = len(str(len(self.__failures)))
+ width = len(str(len(failures)))
initial_indent_format = u' {{:>{width}}}. '.format(width=width)
initial_indent_len = len(initial_indent_format.format(0))
subsequent_indent = u' ' * initial_indent_len
subsequent_extra_indent = u' ' * (initial_indent_len + 10)
- for i, failure in enumerate(self.__failures, 1):
+ for i, failure in enumerate(failures, 1):
entries = _format_failure(failure)
self._display.display(u'\n{}{}'.format(initial_indent_format.format(i), entries[0]))
for entry in entries[1:]:
@@ -59,11 +64,52 @@ class CallbackModule(CallbackBase):
indented = u'{}{}'.format(subsequent_indent, entry)
self._display.display(indented)
-
-# Reason: disable pylint protected-access because we need to access _*
-# attributes of a task result to implement this method.
-# Status: permanently disabled unless Ansible's API changes.
-# pylint: disable=protected-access
+ failed_checks = set()
+ playbook_context = None
+ # re: result attrs see top comment # pylint: disable=protected-access
+ for failure in failures:
+ # get context from check task result since callback plugins cannot access task vars
+ playbook_context = playbook_context or failure['result']._result.get('playbook_context')
+ failed_checks.update(
+ name
+ for name, result in failure['result']._result.get('checks', {}).items()
+ if result.get('failed')
+ )
+ if failed_checks:
+ self._print_check_failure_summary(failed_checks, playbook_context)
+
+ def _print_check_failure_summary(self, failed_checks, context):
+ checks = ','.join(sorted(failed_checks))
+ # NOTE: context is not set if all failures occurred prior to checks task
+ summary = (
+ '\n'
+ 'The execution of "{playbook}"\n'
+ 'includes checks designed to fail early if the requirements\n'
+ 'of the playbook are not met. One or more of these checks\n'
+ 'failed. To disregard these results, you may choose to\n'
+ 'disable failing checks by setting an Ansible variable:\n\n'
+ ' openshift_disable_check={checks}\n\n'
+ 'Failing check names are shown in the failure details above.\n'
+ 'Some checks may be configurable by variables if your requirements\n'
+ 'are different from the defaults; consult check documentation.\n'
+ 'Variables can be set in the inventory or passed on the\n'
+ 'command line using the -e flag to ansible-playbook.\n'
+ ).format(playbook=self._playbook_file, checks=checks)
+ if context in ['pre-install', 'health']:
+ summary = (
+ '\n'
+ 'You may choose to configure or disable failing checks by\n'
+ 'setting Ansible variables. To disable those above:\n\n'
+ ' openshift_disable_check={checks}\n\n'
+ 'Consult check documentation for configurable variables.\n'
+ 'Variables can be set in the inventory or passed on the\n'
+ 'command line using the -e flag to ansible-playbook.\n'
+ ).format(checks=checks)
+ # other expected contexts: install, upgrade
+ self._display.display(summary)
+
+
+# re: result attrs see top comment # pylint: disable=protected-access
def _format_failure(failure):
'''Return a list of pretty-formatted text entries describing a failure, including
relevant information about it. Expect that the list of text entries will be joined
@@ -100,11 +146,8 @@ def _format_failed_checks(checks):
return stringc(pformat(checks), C.COLOR_ERROR)
-# Reason: disable pylint protected-access because we need to access _*
-# attributes of obj to implement this function.
-# This is inspired by ansible.playbook.base.Base.dump_me.
-# Status: permanently disabled unless Ansible's API changes.
-# pylint: disable=protected-access
+# This is inspired by ansible.playbook.base.Base.dump_me.
+# re: play/task/block attrs see top comment # pylint: disable=protected-access
def _get_play(obj):
'''Given a task or block, recursively tries to find its parent play.'''
if hasattr(obj, '_play'):
diff --git a/roles/openshift_health_checker/openshift_checks/disk_availability.py b/roles/openshift_health_checker/openshift_checks/disk_availability.py
index c2792a0fe..962148cb8 100644
--- a/roles/openshift_health_checker/openshift_checks/disk_availability.py
+++ b/roles/openshift_health_checker/openshift_checks/disk_availability.py
@@ -27,10 +27,12 @@ class DiskAvailability(NotContainerizedMixin, OpenShiftCheck):
def run(self, tmp, task_vars):
group_names = get_var(task_vars, "group_names")
ansible_mounts = get_var(task_vars, "ansible_mounts")
-
- min_free_bytes = max(self.recommended_disk_space_bytes.get(name, 0) for name in group_names)
free_bytes = self.openshift_available_disk(ansible_mounts)
+ recommended_min = max(self.recommended_disk_space_bytes.get(name, 0) for name in group_names)
+ configured_min = int(get_var(task_vars, "openshift_check_min_host_disk_gb", default=0)) * 10**9
+ min_free_bytes = configured_min or recommended_min
+
if free_bytes < min_free_bytes:
return {
'failed': True,
diff --git a/roles/openshift_health_checker/openshift_checks/memory_availability.py b/roles/openshift_health_checker/openshift_checks/memory_availability.py
index 28805dc37..8b1a58ef4 100644
--- a/roles/openshift_health_checker/openshift_checks/memory_availability.py
+++ b/roles/openshift_health_checker/openshift_checks/memory_availability.py
@@ -13,7 +13,7 @@ class MemoryAvailability(OpenShiftCheck):
recommended_memory_bytes = {
"masters": 16 * 10**9,
"nodes": 8 * 10**9,
- "etcd": 20 * 10**9,
+ "etcd": 8 * 10**9,
}
@classmethod
@@ -27,7 +27,9 @@ class MemoryAvailability(OpenShiftCheck):
group_names = get_var(task_vars, "group_names")
total_memory_bytes = get_var(task_vars, "ansible_memtotal_mb") * 10**6
- min_memory_bytes = max(self.recommended_memory_bytes.get(name, 0) for name in group_names)
+ recommended_min = max(self.recommended_memory_bytes.get(name, 0) for name in group_names)
+ configured_min = int(get_var(task_vars, "openshift_check_min_host_memory_gb", default=0)) * 10**9
+ min_memory_bytes = configured_min or recommended_min
if total_memory_bytes < min_memory_bytes:
return {
diff --git a/roles/openshift_health_checker/test/action_plugin_test.py b/roles/openshift_health_checker/test/action_plugin_test.py
index 2693ae37b..6ebf0ebb2 100644
--- a/roles/openshift_health_checker/test/action_plugin_test.py
+++ b/roles/openshift_health_checker/test/action_plugin_test.py
@@ -67,6 +67,7 @@ def changed(result):
return result.get('changed', False)
+# tests whether task is skipped, not individual checks
def skipped(result):
return result.get('skipped', False)
@@ -101,7 +102,20 @@ def test_action_plugin_skip_non_active_checks(plugin, task_vars, monkeypatch):
result = plugin.run(tmp=None, task_vars=task_vars)
- assert result['checks']['fake_check'] == {'skipped': True}
+ assert result['checks']['fake_check'] == dict(skipped=True, skipped_reason="Not active for this host")
+ assert not failed(result)
+ assert not changed(result)
+ assert not skipped(result)
+
+
+def test_action_plugin_skip_disabled_checks(plugin, task_vars, monkeypatch):
+ checks = [fake_check('fake_check', is_active=True)]
+ monkeypatch.setattr('openshift_checks.OpenShiftCheck.subclasses', classmethod(lambda cls: checks))
+
+ task_vars['openshift_disable_check'] = 'fake_check'
+ result = plugin.run(tmp=None, task_vars=task_vars)
+
+ assert result['checks']['fake_check'] == dict(skipped=True, skipped_reason="Disabled by user request")
assert not failed(result)
assert not changed(result)
assert not skipped(result)
diff --git a/roles/openshift_health_checker/test/disk_availability_test.py b/roles/openshift_health_checker/test/disk_availability_test.py
index 970b474d7..b353fa610 100644
--- a/roles/openshift_health_checker/test/disk_availability_test.py
+++ b/roles/openshift_health_checker/test/disk_availability_test.py
@@ -42,9 +42,10 @@ def test_cannot_determine_available_disk(ansible_mounts, extra_words):
assert word in str(excinfo.value)
-@pytest.mark.parametrize('group_names,ansible_mounts', [
+@pytest.mark.parametrize('group_names,configured_min,ansible_mounts', [
(
['masters'],
+ 0,
[{
'mount': '/',
'size_available': 40 * 10**9 + 1,
@@ -52,6 +53,7 @@ def test_cannot_determine_available_disk(ansible_mounts, extra_words):
),
(
['nodes'],
+ 0,
[{
'mount': '/',
'size_available': 15 * 10**9 + 1,
@@ -59,6 +61,7 @@ def test_cannot_determine_available_disk(ansible_mounts, extra_words):
),
(
['etcd'],
+ 0,
[{
'mount': '/',
'size_available': 20 * 10**9 + 1,
@@ -66,6 +69,15 @@ def test_cannot_determine_available_disk(ansible_mounts, extra_words):
),
(
['etcd'],
+ 1, # configure lower threshold
+ [{
+ 'mount': '/',
+ 'size_available': 1 * 10**9 + 1, # way smaller than recommended
+ }],
+ ),
+ (
+ ['etcd'],
+ 0,
[{
# not enough space on / ...
'mount': '/',
@@ -77,9 +89,10 @@ def test_cannot_determine_available_disk(ansible_mounts, extra_words):
}],
),
])
-def test_succeeds_with_recommended_disk_space(group_names, ansible_mounts):
+def test_succeeds_with_recommended_disk_space(group_names, configured_min, ansible_mounts):
task_vars = dict(
group_names=group_names,
+ openshift_check_min_host_disk_gb=configured_min,
ansible_mounts=ansible_mounts,
)
@@ -89,9 +102,10 @@ def test_succeeds_with_recommended_disk_space(group_names, ansible_mounts):
assert not result.get('failed', False)
-@pytest.mark.parametrize('group_names,ansible_mounts,extra_words', [
+@pytest.mark.parametrize('group_names,configured_min,ansible_mounts,extra_words', [
(
['masters'],
+ 0,
[{
'mount': '/',
'size_available': 1,
@@ -99,7 +113,17 @@ def test_succeeds_with_recommended_disk_space(group_names, ansible_mounts):
['0.0 GB'],
),
(
+ ['masters'],
+ 100, # set a higher threshold
+ [{
+ 'mount': '/',
+ 'size_available': 50 * 10**9, # would normally be enough...
+ }],
+ ['100.0 GB'],
+ ),
+ (
['nodes'],
+ 0,
[{
'mount': '/',
'size_available': 1 * 10**9,
@@ -108,6 +132,7 @@ def test_succeeds_with_recommended_disk_space(group_names, ansible_mounts):
),
(
['etcd'],
+ 0,
[{
'mount': '/',
'size_available': 1,
@@ -116,6 +141,7 @@ def test_succeeds_with_recommended_disk_space(group_names, ansible_mounts):
),
(
['nodes', 'masters'],
+ 0,
[{
'mount': '/',
# enough space for a node, not enough for a master
@@ -125,6 +151,7 @@ def test_succeeds_with_recommended_disk_space(group_names, ansible_mounts):
),
(
['etcd'],
+ 0,
[{
# enough space on / ...
'mount': '/',
@@ -137,9 +164,10 @@ def test_succeeds_with_recommended_disk_space(group_names, ansible_mounts):
['0.0 GB'],
),
])
-def test_fails_with_insufficient_disk_space(group_names, ansible_mounts, extra_words):
+def test_fails_with_insufficient_disk_space(group_names, configured_min, ansible_mounts, extra_words):
task_vars = dict(
group_names=group_names,
+ openshift_check_min_host_disk_gb=configured_min,
ansible_mounts=ansible_mounts,
)
diff --git a/roles/openshift_health_checker/test/memory_availability_test.py b/roles/openshift_health_checker/test/memory_availability_test.py
index e161a5b9e..1db203854 100644
--- a/roles/openshift_health_checker/test/memory_availability_test.py
+++ b/roles/openshift_health_checker/test/memory_availability_test.py
@@ -20,27 +20,37 @@ def test_is_active(group_names, is_active):
assert MemoryAvailability.is_active(task_vars=task_vars) == is_active
-@pytest.mark.parametrize('group_names,ansible_memtotal_mb', [
+@pytest.mark.parametrize('group_names,configured_min,ansible_memtotal_mb', [
(
['masters'],
+ 0,
17200,
),
(
['nodes'],
+ 0,
8200,
),
(
+ ['nodes'],
+ 1, # configure lower threshold
+ 2000, # too low for recommended but not for configured
+ ),
+ (
['etcd'],
- 22200,
+ 0,
+ 8200,
),
(
['masters', 'nodes'],
+ 0,
17000,
),
])
-def test_succeeds_with_recommended_memory(group_names, ansible_memtotal_mb):
+def test_succeeds_with_recommended_memory(group_names, configured_min, ansible_memtotal_mb):
task_vars = dict(
group_names=group_names,
+ openshift_check_min_host_memory_gb=configured_min,
ansible_memtotal_mb=ansible_memtotal_mb,
)
@@ -50,39 +60,56 @@ def test_succeeds_with_recommended_memory(group_names, ansible_memtotal_mb):
assert not result.get('failed', False)
-@pytest.mark.parametrize('group_names,ansible_memtotal_mb,extra_words', [
+@pytest.mark.parametrize('group_names,configured_min,ansible_memtotal_mb,extra_words', [
(
['masters'],
0,
+ 0,
['0.0 GB'],
),
(
['nodes'],
+ 0,
100,
['0.1 GB'],
),
(
+ ['nodes'],
+ 24, # configure higher threshold
+ 20000, # enough to meet recommended but not configured
+ ['20.0 GB'],
+ ),
+ (
['etcd'],
- -1,
- ['0.0 GB'],
+ 0,
+ 7000,
+ ['7.0 GB'],
+ ),
+ (
+ ['etcd', 'masters'],
+ 0,
+ 9000, # enough memory for etcd, not enough for a master
+ ['9.0 GB'],
),
(
['nodes', 'masters'],
+ 0,
# enough memory for a node, not enough for a master
11000,
['11.0 GB'],
),
])
-def test_fails_with_insufficient_memory(group_names, ansible_memtotal_mb, extra_words):
+def test_fails_with_insufficient_memory(group_names, configured_min, ansible_memtotal_mb, extra_words):
task_vars = dict(
group_names=group_names,
+ openshift_check_min_host_memory_gb=configured_min,
ansible_memtotal_mb=ansible_memtotal_mb,
)
check = MemoryAvailability(execute_module=fake_execute_module)
result = check.run(tmp=None, task_vars=task_vars)
- assert result['failed']
+ assert result.get('failed', False)
for word in 'below recommended'.split() + extra_words:
assert word in result['msg']
diff --git a/roles/openshift_logging/filter_plugins/openshift_logging.py b/roles/openshift_logging/filter_plugins/openshift_logging.py
index 44b0b2d48..eac086e81 100644
--- a/roles/openshift_logging/filter_plugins/openshift_logging.py
+++ b/roles/openshift_logging/filter_plugins/openshift_logging.py
@@ -42,7 +42,7 @@ def map_from_pairs(source, delim="="):
if source == '':
return dict()
- return dict(source.split(delim) for item in source.split(","))
+ return dict(item.split(delim) for item in source.split(","))
# pylint: disable=too-few-public-methods
diff --git a/roles/openshift_storage_glusterfs/filter_plugins/openshift_storage_glusterfs.py b/roles/openshift_storage_glusterfs/filter_plugins/openshift_storage_glusterfs.py
index 88801e487..a86c96df7 100644
--- a/roles/openshift_storage_glusterfs/filter_plugins/openshift_storage_glusterfs.py
+++ b/roles/openshift_storage_glusterfs/filter_plugins/openshift_storage_glusterfs.py
@@ -8,7 +8,7 @@ def map_from_pairs(source, delim="="):
if source == '':
return dict()
- return dict(source.split(delim) for item in source.split(","))
+ return dict(item.split(delim) for item in source.split(","))
# pylint: disable=too-few-public-methods