diff options
Diffstat (limited to 'roles/openshift_certificate_expiry/library/openshift_cert_expiry.py')
| -rw-r--r-- | roles/openshift_certificate_expiry/library/openshift_cert_expiry.py | 635 |
1 files changed, 635 insertions, 0 deletions
diff --git a/roles/openshift_certificate_expiry/library/openshift_cert_expiry.py b/roles/openshift_certificate_expiry/library/openshift_cert_expiry.py new file mode 100644 index 000000000..f18ab75d0 --- /dev/null +++ b/roles/openshift_certificate_expiry/library/openshift_cert_expiry.py @@ -0,0 +1,635 @@ +#!/usr/bin/python +# -*- coding: utf-8 -*- +# pylint: disable=line-too-long,invalid-name + +"""For details on this module see DOCUMENTATION (below)""" + +# router/registry cert grabbing +import subprocess +# etcd config file +import ConfigParser +# Expiration parsing +import datetime +# File path stuff +import os +# Config file parsing +import yaml +# Certificate loading +import OpenSSL.crypto + +DOCUMENTATION = ''' +--- +module: openshift_cert_expiry +short_description: Check OpenShift Container Platform (OCP) and Kube certificate expirations on a cluster +description: + - The M(openshift_cert_expiry) module has two basic functions: to flag certificates which will expire in a set window of time from now, and to notify you about certificates which have already expired. + - When the module finishes, a summary of the examination is returned. Each certificate in the summary has a C(health) key with a value of one of the following: + - C(ok) - not expired, and outside of the expiration C(warning_days) window. + - C(warning) - not expired, but will expire between now and the C(warning_days) window. + - C(expired) - an expired certificate. + - Certificate flagging follow this logic: + - If the expiration date is before now then the certificate is classified as C(expired). + - The certificates time to live (expiration date - now) is calculated, if that time window is less than C(warning_days) the certificate is classified as C(warning). + - All other conditions are classified as C(ok). + - The following keys are ALSO present in the certificate summary: + - C(cert_cn) - The common name of the certificate (additional CNs present in SAN extensions are omitted) + - C(days_remaining) - The number of days until the certificate expires. + - C(expiry) - The date the certificate expires on. + - C(path) - The full path to the certificate on the examined host. +version_added: "0.0" +options: + config_base: + description: + - Base path to OCP system settings. + required: false + default: /etc/origin + warning_days: + description: + - Flag certificates which will expire in C(warning_days) days from now. + required: false + default: 30 + show_all: + description: + - Enable this option to show analysis of ALL certificates examined by this module. + - By default only certificates which have expired, or will expire within the C(warning_days) window will be reported. + required: false + default: false + +author: "Tim Bielawa (@tbielawa) <tbielawa@redhat.com>" +''' + +EXAMPLES = ''' +# Default invocation, only notify about expired certificates or certificates which will expire within 30 days from now +- openshift_cert_expiry: + +# Expand the warning window to show certificates expiring within a year from now +- openshift_cert_expiry: warning_days=365 + +# Show expired, soon to expire (now + 30 days), and all other certificates examined +- openshift_cert_expiry: show_all=true +''' + + +# We only need this for one thing, we don't care if it doesn't have +# that many public methods +# +# pylint: disable=too-few-public-methods +class FakeSecHead(object): + """etcd does not begin their config file with an opening [section] as +required by the Python ConfigParser module. We hack around it by +slipping one in ourselves prior to parsing. + +Source: Alex Martelli - http://stackoverflow.com/a/2819788/6490583 + """ + def __init__(self, fp): + self.fp = fp + self.sechead = '[ETCD]\n' + + def readline(self): + """Make this look like a file-type object""" + if self.sechead: + try: + return self.sechead + finally: + self.sechead = None + else: + return self.fp.readline() + + +###################################################################### + +def filter_paths(path_list): + """`path_list` - A list of file paths to check. Only files which exist +will be returned + """ + return [p for p in path_list if os.path.exists(os.path.realpath(p))] + + +def load_and_handle_cert(cert_string, now, base64decode=False): + """Load a certificate, split off the good parts, and return some +useful data + +Params: + +- `cert_string` (string) - a certificate loaded into a string object +- `now` (datetime) - a datetime object of the time to calculate the certificate 'time_remaining' against +- `base64decode` (bool) - run .decode('base64') on the input? + +Returns: +A 3-tuple of the form: (certificate_common_name, certificate_expiry_date, certificate_time_remaining) + + """ + if base64decode: + _cert_string = cert_string.decode('base-64') + else: + _cert_string = cert_string + + cert_loaded = OpenSSL.crypto.load_certificate( + OpenSSL.crypto.FILETYPE_PEM, _cert_string) + + ###################################################################### + # Read just the first name from the cert - DISABLED while testing + # out the 'get all possible names' function (below) + # + # Strip the subject down to just the value of the first name + # cert_subject = cert_loaded.get_subject().get_components()[0][1] + + ###################################################################### + # Read all possible names from the cert + cert_subjects = [] + for name, value in cert_loaded.get_subject().get_components(): + cert_subjects.append('{}:{}'.format(name, value)) + + # To read SANs from a cert we must read the subjectAltName + # extension from the X509 Object. What makes this more difficult + # is that pyOpenSSL does not give extensions as a list, nor does + # it provide a count of all loaded extensions. + # + # Rather, extensions are REQUESTED by index. We must iterate over + # all extensions until we find the one called 'subjectAltName'. If + # we don't find that extension we'll eventually request an + # extension at an index where no extension exists (IndexError is + # raised). When that happens we know that the cert has no SANs so + # we break out of the loop. + i = 0 + checked_all_extensions = False + while not checked_all_extensions: + try: + # Read the extension at index 'i' + ext = cert_loaded.get_extension(i) + except IndexError: + # We tried to read an extension but it isn't there, that + # means we ran out of extensions to check. Abort + san = None + checked_all_extensions = True + else: + # We were able to load the extension at index 'i' + if ext.get_short_name() == 'subjectAltName': + san = ext + checked_all_extensions = True + else: + # Try reading the next extension + i += 1 + + if san is not None: + # The X509Extension object for subjectAltName prints as a + # string with the alt names separated by a comma and a + # space. Split the string by ', ' and then add our new names + # to the list of existing names + cert_subjects.extend(str(san).split(', ')) + + cert_subject = ', '.join(cert_subjects) + ###################################################################### + + # Grab the expiration date + cert_expiry = cert_loaded.get_notAfter() + cert_expiry_date = datetime.datetime.strptime( + cert_expiry, + # example get_notAfter() => 20180922170439Z + '%Y%m%d%H%M%SZ') + + time_remaining = cert_expiry_date - now + + return (cert_subject, cert_expiry_date, time_remaining) + + +def classify_cert(cert_meta, now, time_remaining, expire_window, cert_list): + """Given metadata about a certificate under examination, classify it + into one of three categories, 'ok', 'warning', and 'expired'. + +Params: + +- `cert_meta` dict - A dict with certificate metadata. Required fields + include: 'cert_cn', 'path', 'expiry', 'days_remaining', 'health'. +- `now` (datetime) - a datetime object of the time to calculate the certificate 'time_remaining' against +- `time_remaining` (datetime.timedelta) - a timedelta for how long until the cert expires +- `expire_window` (datetime.timedelta) - a timedelta for how long the warning window is +- `cert_list` list - A list to shove the classified cert into + +Return: +- `cert_list` - The updated list of classified certificates + """ + expiry_str = str(cert_meta['expiry']) + # Categorization + if cert_meta['expiry'] < now: + # This already expired, must NOTIFY + cert_meta['health'] = 'expired' + elif time_remaining < expire_window: + # WARN about this upcoming expirations + cert_meta['health'] = 'warning' + else: + # Not expired or about to expire + cert_meta['health'] = 'ok' + + cert_meta['expiry'] = expiry_str + cert_list.append(cert_meta) + return cert_list + + +def tabulate_summary(certificates, kubeconfigs, etcd_certs, router_certs, registry_certs): + """Calculate the summary text for when the module finishes +running. This includes counds of each classification and what have +you. + +Params: + +- `certificates` (list of dicts) - Processed `expire_check_result` + dicts with filled in `health` keys for system certificates. +- `kubeconfigs` - as above for kubeconfigs +- `etcd_certs` - as above for etcd certs +Return: + +- `summary_results` (dict) - Counts of each cert type classification + and total items examined. + """ + items = certificates + kubeconfigs + etcd_certs + router_certs + registry_certs + + summary_results = { + 'system_certificates': len(certificates), + 'kubeconfig_certificates': len(kubeconfigs), + 'etcd_certificates': len(etcd_certs), + 'router_certs': len(router_certs), + 'registry_certs': len(registry_certs), + 'total': len(items), + 'ok': 0, + 'warning': 0, + 'expired': 0 + } + + summary_results['expired'] = len([c for c in items if c['health'] == 'expired']) + summary_results['warning'] = len([c for c in items if c['health'] == 'warning']) + summary_results['ok'] = len([c for c in items if c['health'] == 'ok']) + + return summary_results + + +###################################################################### +# This is our module MAIN function after all, so there's bound to be a +# lot of code bundled up into one block +# +# pylint: disable=too-many-locals,too-many-locals,too-many-statements,too-many-branches +def main(): + """This module examines certificates (in various forms) which compose +an OpenShift Container Platform cluster + """ + + module = AnsibleModule( + argument_spec=dict( + config_base=dict( + required=False, + default="/etc/origin", + type='str'), + warning_days=dict( + required=False, + default=int(30), + type='int'), + show_all=dict( + required=False, + default="False", + type='bool') + ), + supports_check_mode=True, + ) + + # Basic scaffolding for OpenShift spcific certs + openshift_base_config_path = module.params['config_base'] + openshift_master_config_path = os.path.normpath( + os.path.join(openshift_base_config_path, "master/master-config.yaml") + ) + openshift_node_config_path = os.path.normpath( + os.path.join(openshift_base_config_path, "node/node-config.yaml") + ) + openshift_cert_check_paths = [ + openshift_master_config_path, + openshift_node_config_path, + ] + + # Paths for Kubeconfigs. Additional kubeconfigs are conditionally + # checked later in the code + master_kube_configs = ['admin', 'openshift-master', + 'openshift-node', 'openshift-router', + 'openshift-registry'] + + kubeconfig_paths = [] + for m_kube_config in master_kube_configs: + kubeconfig_paths.append( + os.path.normpath( + os.path.join(openshift_base_config_path, "master/%s.kubeconfig" % m_kube_config) + ) + ) + + # etcd, where do you hide your certs? Used when parsing etcd.conf + etcd_cert_params = [ + "ETCD_CA_FILE", + "ETCD_CERT_FILE", + "ETCD_PEER_CA_FILE", + "ETCD_PEER_CERT_FILE", + ] + + # Expiry checking stuff + now = datetime.datetime.now() + # todo, catch exception for invalid input and return a fail_json + warning_days = int(module.params['warning_days']) + expire_window = datetime.timedelta(days=warning_days) + + # Module stuff + # + # The results of our cert checking to return from the task call + check_results = {} + check_results['meta'] = {} + check_results['meta']['warning_days'] = warning_days + check_results['meta']['checked_at_time'] = str(now) + check_results['meta']['warn_after_date'] = str(now + expire_window) + check_results['meta']['show_all'] = str(module.params['show_all']) + # All the analyzed certs accumulate here + ocp_certs = [] + + ###################################################################### + # Sure, why not? Let's enable check mode. + if module.check_mode: + check_results['ocp_certs'] = [] + module.exit_json( + check_results=check_results, + msg="Checked 0 total certificates. Expired/Warning/OK: 0/0/0. Warning window: %s days" % module.params['warning_days'], + rc=0, + changed=False + ) + + ###################################################################### + # Check for OpenShift Container Platform specific certs + ###################################################################### + for os_cert in filter_paths(openshift_cert_check_paths): + # Open up that config file and locate the cert and CA + with open(os_cert, 'r') as fp: + cert_meta = {} + cfg = yaml.load(fp) + # cert files are specified in parsed `fp` as relative to the path + # of the original config file. 'master-config.yaml' with certFile + # = 'foo.crt' implies that 'foo.crt' is in the same + # directory. certFile = '../foo.crt' is in the parent directory. + cfg_path = os.path.dirname(fp.name) + cert_meta['certFile'] = os.path.join(cfg_path, cfg['servingInfo']['certFile']) + cert_meta['clientCA'] = os.path.join(cfg_path, cfg['servingInfo']['clientCA']) + + ###################################################################### + # Load the certificate and the CA, parse their expiration dates into + # datetime objects so we can manipulate them later + for _, v in cert_meta.iteritems(): + with open(v, 'r') as fp: + cert = fp.read() + cert_subject, cert_expiry_date, time_remaining = load_and_handle_cert(cert, now) + + expire_check_result = { + 'cert_cn': cert_subject, + 'path': fp.name, + 'expiry': cert_expiry_date, + 'days_remaining': time_remaining.days, + 'health': None, + } + + classify_cert(expire_check_result, now, time_remaining, expire_window, ocp_certs) + + ###################################################################### + # /Check for OpenShift Container Platform specific certs + ###################################################################### + + ###################################################################### + # Check service Kubeconfigs + ###################################################################### + kubeconfigs = [] + + # There may be additional kubeconfigs to check, but their naming + # is less predictable than the ones we've already assembled. + + try: + # Try to read the standard 'node-config.yaml' file to check if + # this host is a node. + with open(openshift_node_config_path, 'r') as fp: + cfg = yaml.load(fp) + + # OK, the config file exists, therefore this is a + # node. Nodes have their own kubeconfig files to + # communicate with the master API. Let's read the relative + # path to that file from the node config. + node_masterKubeConfig = cfg['masterKubeConfig'] + # As before, the path to the 'masterKubeConfig' file is + # relative to `fp` + cfg_path = os.path.dirname(fp.name) + node_kubeconfig = os.path.join(cfg_path, node_masterKubeConfig) + + with open(node_kubeconfig, 'r') as fp: + # Read in the nodes kubeconfig file and grab the good stuff + cfg = yaml.load(fp) + + c = cfg['users'][0]['user']['client-certificate-data'] + (cert_subject, + cert_expiry_date, + time_remaining) = load_and_handle_cert(c, now, base64decode=True) + + expire_check_result = { + 'cert_cn': cert_subject, + 'path': fp.name, + 'expiry': cert_expiry_date, + 'days_remaining': time_remaining.days, + 'health': None, + } + + classify_cert(expire_check_result, now, time_remaining, expire_window, kubeconfigs) + except IOError: + # This is not a node + pass + + for kube in filter_paths(kubeconfig_paths): + with open(kube, 'r') as fp: + # TODO: Maybe consider catching exceptions here? + cfg = yaml.load(fp) + + # Per conversation, "the kubeconfigs you care about: + # admin, router, registry should all be single + # value". Following that advice we only grab the data for + # the user at index 0 in the 'users' list. There should + # not be more than one user. + c = cfg['users'][0]['user']['client-certificate-data'] + (cert_subject, + cert_expiry_date, + time_remaining) = load_and_handle_cert(c, now, base64decode=True) + + expire_check_result = { + 'cert_cn': cert_subject, + 'path': fp.name, + 'expiry': cert_expiry_date, + 'days_remaining': time_remaining.days, + 'health': None, + } + + classify_cert(expire_check_result, now, time_remaining, expire_window, kubeconfigs) + + ###################################################################### + # /Check service Kubeconfigs + ###################################################################### + + ###################################################################### + # Check etcd certs + ###################################################################### + # Some values may be duplicated, make this a set for now so we + # unique them all + etcd_certs_to_check = set([]) + etcd_certs = [] + etcd_cert_params.append('dne') + try: + with open('/etc/etcd/etcd.conf', 'r') as fp: + etcd_config = ConfigParser.ConfigParser() + etcd_config.readfp(FakeSecHead(fp)) + + for param in etcd_cert_params: + try: + etcd_certs_to_check.add(etcd_config.get('ETCD', param)) + except ConfigParser.NoOptionError: + # That parameter does not exist, oh well... + pass + except IOError: + # No etcd to see here, move along + pass + + for etcd_cert in filter_paths(etcd_certs_to_check): + with open(etcd_cert, 'r') as fp: + c = fp.read() + (cert_subject, + cert_expiry_date, + time_remaining) = load_and_handle_cert(c, now) + + expire_check_result = { + 'cert_cn': cert_subject, + 'path': fp.name, + 'expiry': cert_expiry_date, + 'days_remaining': time_remaining.days, + 'health': None, + } + + classify_cert(expire_check_result, now, time_remaining, expire_window, etcd_certs) + + ###################################################################### + # /Check etcd certs + ###################################################################### + + ###################################################################### + # Check router/registry certs + # + # These are saved as secrets in etcd. That means that we can not + # simply read a file to grab the data. Instead we're going to + # subprocess out to the 'oc get' command. On non-masters this + # command will fail, that is expected so we catch that exception. + ###################################################################### + router_certs = [] + registry_certs = [] + + ###################################################################### + # First the router certs + try: + router_secrets_raw = subprocess.Popen('oc get secret router-certs -o yaml'.split(), + stdout=subprocess.PIPE) + router_ds = yaml.load(router_secrets_raw.communicate()[0]) + router_c = router_ds['data']['tls.crt'] + router_path = router_ds['metadata']['selfLink'] + except TypeError: + # YAML couldn't load the result, this is not a master + pass + else: + (cert_subject, + cert_expiry_date, + time_remaining) = load_and_handle_cert(router_c, now, base64decode=True) + + expire_check_result = { + 'cert_cn': cert_subject, + 'path': router_path, + 'expiry': cert_expiry_date, + 'days_remaining': time_remaining.days, + 'health': None, + } + + classify_cert(expire_check_result, now, time_remaining, expire_window, router_certs) + + check_results['router'] = router_certs + + ###################################################################### + # Now for registry + # registry_secrets = subprocess.call('oc get secret registry-certificates -o yaml'.split()) + # out = subprocess.PIPE + try: + registry_secrets_raw = subprocess.Popen('oc get secret registry-certificates -o yaml'.split(), + stdout=subprocess.PIPE) + registry_ds = yaml.load(registry_secrets_raw.communicate()[0]) + registry_c = registry_ds['data']['registry.crt'] + registry_path = registry_ds['metadata']['selfLink'] + except TypeError: + # YAML couldn't load the result, this is not a master + pass + else: + (cert_subject, + cert_expiry_date, + time_remaining) = load_and_handle_cert(registry_c, now, base64decode=True) + + expire_check_result = { + 'cert_cn': cert_subject, + 'path': registry_path, + 'expiry': cert_expiry_date, + 'days_remaining': time_remaining.days, + 'health': None, + } + + classify_cert(expire_check_result, now, time_remaining, expire_window, registry_certs) + + check_results['registry'] = registry_certs + + ###################################################################### + # /Check router/registry certs + ###################################################################### + + res = tabulate_summary(ocp_certs, kubeconfigs, etcd_certs, router_certs, registry_certs) + + msg = "Checked {count} total certificates. Expired/Warning/OK: {exp}/{warn}/{ok}. Warning window: {window} days".format( + count=res['total'], + exp=res['expired'], + warn=res['warning'], + ok=res['ok'], + window=int(module.params['warning_days']), + ) + + # By default we only return detailed information about expired or + # warning certificates. If show_all is true then we will print all + # the certificates examined. + if not module.params['show_all']: + check_results['ocp_certs'] = [crt for crt in ocp_certs if crt['health'] in ['expired', 'warning']] + check_results['kubeconfigs'] = [crt for crt in kubeconfigs if crt['health'] in ['expired', 'warning']] + check_results['etcd'] = [crt for crt in etcd_certs if crt['health'] in ['expired', 'warning']] + else: + check_results['ocp_certs'] = ocp_certs + check_results['kubeconfigs'] = kubeconfigs + check_results['etcd'] = etcd_certs + + # Sort the final results to report in order of ascending safety + # time. That is to say, the certificates which will expire sooner + # will be at the front of the list and certificates which will + # expire later are at the end. + check_results['ocp_certs'] = sorted(check_results['ocp_certs'], cmp=lambda x, y: cmp(x['days_remaining'], y['days_remaining'])) + check_results['kubeconfigs'] = sorted(check_results['kubeconfigs'], cmp=lambda x, y: cmp(x['days_remaining'], y['days_remaining'])) + check_results['etcd'] = sorted(check_results['etcd'], cmp=lambda x, y: cmp(x['days_remaining'], y['days_remaining'])) + + # This module will never change anything, but we might want to + # change the return code parameter if there is some catastrophic + # error we noticed earlier + module.exit_json( + check_results=check_results, + summary=res, + msg=msg, + rc=0, + changed=False + ) + +###################################################################### +# It's just the way we do things in Ansible. So disable this warning +# +# pylint: disable=wrong-import-position,import-error +from ansible.module_utils.basic import AnsibleModule +if __name__ == '__main__': + main() |