Wrapping Retry for Py3.10
[osm/N2VC.git] / n2vc / k8s_helm_conn.py
1 ##
2 # Copyright 2019 Telefonica Investigacion y Desarrollo, S.A.U.
3 # This file is part of OSM
4 # All Rights Reserved.
5 #
6 # Licensed under the Apache License, Version 2.0 (the "License");
7 # you may not use this file except in compliance with the License.
8 # You may obtain a copy of the License at
9 #
10 # http://www.apache.org/licenses/LICENSE-2.0
11 #
12 # Unless required by applicable law or agreed to in writing, software
13 # distributed under the License is distributed on an "AS IS" BASIS,
14 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
15 # implied.
16 # See the License for the specific language governing permissions and
17 # limitations under the License.
18 #
19 # For those usages not covered by the Apache License, Version 2.0 please
20 # contact with: nfvlabs@tid.es
21 ##
22 import asyncio
23 from typing import Union
24 import os
25 import yaml
26
27 from n2vc.k8s_helm_base_conn import K8sHelmBaseConnector
28 from n2vc.exceptions import K8sException
29
30
31 class K8sHelmConnector(K8sHelmBaseConnector):
32
33 """
34 ####################################################################################
35 ################################### P U B L I C ####################################
36 ####################################################################################
37 """
38
39 def __init__(
40 self,
41 fs: object,
42 db: object,
43 kubectl_command: str = "/usr/bin/kubectl",
44 helm_command: str = "/usr/bin/helm",
45 log: object = None,
46 on_update_db=None,
47 ):
48 """
49 Initializes helm connector for helm v2
50
51 :param fs: file system for kubernetes and helm configuration
52 :param db: database object to write current operation status
53 :param kubectl_command: path to kubectl executable
54 :param helm_command: path to helm executable
55 :param log: logger
56 :param on_update_db: callback called when k8s connector updates database
57 """
58
59 # parent class
60 K8sHelmBaseConnector.__init__(
61 self,
62 db=db,
63 log=log,
64 fs=fs,
65 kubectl_command=kubectl_command,
66 helm_command=helm_command,
67 on_update_db=on_update_db,
68 )
69
70 self.log.info("Initializing K8S Helm2 connector")
71
72 # initialize helm client-only
73 self.log.debug("Initializing helm client-only...")
74 command = "{} init --client-only {} ".format(
75 self._helm_command,
76 "--stable-repo-url {}".format(self._stable_repo_url)
77 if self._stable_repo_url
78 else "--skip-repos",
79 )
80 try:
81 asyncio.create_task(
82 self._local_async_exec(command=command, raise_exception_on_error=False)
83 )
84 except Exception as e:
85 self.warning(
86 msg="helm init failed (it was already initialized): {}".format(e)
87 )
88
89 self.log.info("K8S Helm2 connector initialized")
90
91 async def install(
92 self,
93 cluster_uuid: str,
94 kdu_model: str,
95 kdu_instance: str,
96 atomic: bool = True,
97 timeout: float = 300,
98 params: dict = None,
99 db_dict: dict = None,
100 kdu_name: str = None,
101 namespace: str = None,
102 **kwargs,
103 ):
104 """
105 Deploys of a new KDU instance. It would implicitly rely on the `install` call
106 to deploy the Chart/Bundle properly parametrized (in practice, this call would
107 happen before any _initial-config-primitive_of the VNF is called).
108
109 :param cluster_uuid: UUID of a K8s cluster known by OSM
110 :param kdu_model: chart/reference (string), which can be either
111 of these options:
112 - a name of chart available via the repos known by OSM
113 (e.g. stable/openldap, stable/openldap:1.2.4)
114 - a path to a packaged chart (e.g. mychart.tgz)
115 - a path to an unpacked chart directory or a URL (e.g. mychart)
116 :param kdu_instance: Kdu instance name
117 :param atomic: If set, installation process purges chart/bundle on fail, also
118 will wait until all the K8s objects are active
119 :param timeout: Time in seconds to wait for the install of the chart/bundle
120 (defaults to Helm default timeout: 300s)
121 :param params: dictionary of key-value pairs for instantiation parameters
122 (overriding default values)
123 :param dict db_dict: where to write into database when the status changes.
124 It contains a dict with {collection: <str>, filter: {},
125 path: <str>},
126 e.g. {collection: "nsrs", filter:
127 {_id: <nsd-id>, path: "_admin.deployed.K8S.3"}
128 :param kdu_name: Name of the KDU instance to be installed
129 :param namespace: K8s namespace to use for the KDU instance
130 :param kwargs: Additional parameters (None yet)
131 :return: True if successful
132 """
133 self.log.debug("installing {} in cluster {}".format(kdu_model, cluster_uuid))
134
135 # sync local dir
136 self.fs.sync(from_path=cluster_uuid)
137
138 # init env, paths
139 paths, env = self._init_paths_env(
140 cluster_name=cluster_uuid, create_if_not_exist=True
141 )
142
143 await self._install_impl(
144 cluster_uuid,
145 kdu_model,
146 paths,
147 env,
148 kdu_instance,
149 atomic=atomic,
150 timeout=timeout,
151 params=params,
152 db_dict=db_dict,
153 kdu_name=kdu_name,
154 namespace=namespace,
155 )
156
157 # sync fs
158 self.fs.reverse_sync(from_path=cluster_uuid)
159
160 self.log.debug("Returning kdu_instance {}".format(kdu_instance))
161 return True
162
163 async def inspect_kdu(self, kdu_model: str, repo_url: str = None) -> str:
164 self.log.debug(
165 "inspect kdu_model {} from (optional) repo: {}".format(kdu_model, repo_url)
166 )
167
168 return await self._exec_inspect_command(
169 inspect_command="", kdu_model=kdu_model, repo_url=repo_url
170 )
171
172 """
173 ####################################################################################
174 ################################### P R I V A T E ##################################
175 ####################################################################################
176 """
177
178 def _init_paths_env(self, cluster_name: str, create_if_not_exist: bool = True):
179 """
180 Creates and returns base cluster and kube dirs and returns them.
181 Also created helm3 dirs according to new directory specification, paths are
182 returned and also environment variables that must be provided to execute commands
183
184 Helm 2 directory specification uses helm_home dir:
185
186 The variables assigned for this paths are:
187 - Helm hone: $HELM_HOME
188 - helm kubeconfig: $KUBECONFIG
189
190 :param cluster_name: cluster_name
191 :return: Dictionary with config_paths and dictionary with helm environment variables
192 """
193 base = self.fs.path
194 if base.endswith("/") or base.endswith("\\"):
195 base = base[:-1]
196
197 # base dir for cluster
198 cluster_dir = base + "/" + cluster_name
199
200 # kube dir
201 kube_dir = cluster_dir + "/" + ".kube"
202 if create_if_not_exist and not os.path.exists(kube_dir):
203 self.log.debug("Creating dir {}".format(kube_dir))
204 os.makedirs(kube_dir)
205
206 # helm home dir
207 helm_dir = cluster_dir + "/" + ".helm"
208 if create_if_not_exist and not os.path.exists(helm_dir):
209 self.log.debug("Creating dir {}".format(helm_dir))
210 os.makedirs(helm_dir)
211
212 config_filename = kube_dir + "/config"
213
214 # 2 - Prepare dictionary with paths
215 paths = {
216 "kube_dir": kube_dir,
217 "kube_config": config_filename,
218 "cluster_dir": cluster_dir,
219 "helm_dir": helm_dir,
220 }
221
222 for file_name, file in paths.items():
223 if "dir" in file_name and not os.path.exists(file):
224 err_msg = "{} dir does not exist".format(file)
225 self.log.error(err_msg)
226 raise K8sException(err_msg)
227
228 # 3 - Prepare environment variables
229 env = {"HELM_HOME": helm_dir, "KUBECONFIG": config_filename}
230
231 return paths, env
232
233 async def _get_services(self, cluster_id, kdu_instance, namespace, kubeconfig):
234 # init config, env
235 paths, env = self._init_paths_env(
236 cluster_name=cluster_id, create_if_not_exist=True
237 )
238
239 command1 = "env KUBECONFIG={} {} get manifest {} ".format(
240 kubeconfig, self._helm_command, kdu_instance
241 )
242 command2 = "{} get --namespace={} -f -".format(self.kubectl_command, namespace)
243 output, _rc = await self._local_async_exec_pipe(
244 command1, command2, env=env, raise_exception_on_error=True
245 )
246 services = self._parse_services(output)
247
248 return services
249
250 async def _cluster_init(
251 self, cluster_id: str, namespace: str, paths: dict, env: dict
252 ):
253 """
254 Implements the helm version dependent cluster initialization:
255 For helm2 it initialized tiller environment if needed
256 """
257
258 # check if tiller pod is up in cluster
259 command = "{} --kubeconfig={} --namespace={} get deployments".format(
260 self.kubectl_command, paths["kube_config"], namespace
261 )
262 output, _rc = await self._local_async_exec(
263 command=command, raise_exception_on_error=True, env=env
264 )
265
266 output_table = self._output_to_table(output=output)
267
268 # find 'tiller' pod in all pods
269 already_initialized = False
270 try:
271 for row in output_table:
272 if row[0].startswith("tiller-deploy"):
273 already_initialized = True
274 break
275 except Exception:
276 pass
277
278 # helm init
279 n2vc_installed_sw = False
280 if not already_initialized:
281 self.log.info(
282 "Initializing helm in client and server: {}".format(cluster_id)
283 )
284 command = "{} --kubeconfig={} --namespace kube-system create serviceaccount {}".format(
285 self.kubectl_command, paths["kube_config"], self.service_account
286 )
287 _, _rc = await self._local_async_exec(
288 command=command, raise_exception_on_error=False, env=env
289 )
290
291 command = (
292 "{} --kubeconfig={} create clusterrolebinding osm-tiller-cluster-rule "
293 "--clusterrole=cluster-admin --serviceaccount=kube-system:{}"
294 ).format(self.kubectl_command, paths["kube_config"], self.service_account)
295 _, _rc = await self._local_async_exec(
296 command=command, raise_exception_on_error=False, env=env
297 )
298
299 command = (
300 "{} --kubeconfig={} --tiller-namespace={} --home={} --service-account {} "
301 " {} init"
302 ).format(
303 self._helm_command,
304 paths["kube_config"],
305 namespace,
306 paths["helm_dir"],
307 self.service_account,
308 "--stable-repo-url {}".format(self._stable_repo_url)
309 if self._stable_repo_url
310 else "--skip-repos",
311 )
312 _, _rc = await self._local_async_exec(
313 command=command, raise_exception_on_error=True, env=env
314 )
315 n2vc_installed_sw = True
316 else:
317 # check client helm installation
318 check_file = paths["helm_dir"] + "/repository/repositories.yaml"
319 if not self._check_file_exists(
320 filename=check_file, exception_if_not_exists=False
321 ):
322 self.log.info("Initializing helm in client: {}".format(cluster_id))
323 command = (
324 "{} --kubeconfig={} --tiller-namespace={} "
325 "--home={} init --client-only {} "
326 ).format(
327 self._helm_command,
328 paths["kube_config"],
329 namespace,
330 paths["helm_dir"],
331 "--stable-repo-url {}".format(self._stable_repo_url)
332 if self._stable_repo_url
333 else "--skip-repos",
334 )
335 output, _rc = await self._local_async_exec(
336 command=command, raise_exception_on_error=True, env=env
337 )
338 else:
339 self.log.info("Helm client already initialized")
340
341 repo_list = await self.repo_list(cluster_id)
342 for repo in repo_list:
343 if repo["name"] == "stable" and repo["url"] != self._stable_repo_url:
344 self.log.debug("Add new stable repo url: {}")
345 await self.repo_remove(cluster_id, "stable")
346 if self._stable_repo_url:
347 await self.repo_add(cluster_id, "stable", self._stable_repo_url)
348 break
349
350 return n2vc_installed_sw
351
352 async def _uninstall_sw(self, cluster_id: str, namespace: str):
353 # uninstall Tiller if necessary
354
355 self.log.debug("Uninstalling tiller from cluster {}".format(cluster_id))
356
357 # init paths, env
358 paths, env = self._init_paths_env(
359 cluster_name=cluster_id, create_if_not_exist=True
360 )
361
362 if not namespace:
363 # find namespace for tiller pod
364 command = "{} --kubeconfig={} get deployments --all-namespaces".format(
365 self.kubectl_command, paths["kube_config"]
366 )
367 output, _rc = await self._local_async_exec(
368 command=command, raise_exception_on_error=False, env=env
369 )
370 output_table = self._output_to_table(output=output)
371 namespace = None
372 for r in output_table:
373 try:
374 if "tiller-deploy" in r[1]:
375 namespace = r[0]
376 break
377 except Exception:
378 pass
379 else:
380 msg = "Tiller deployment not found in cluster {}".format(cluster_id)
381 self.log.error(msg)
382
383 self.log.debug("namespace for tiller: {}".format(namespace))
384
385 if namespace:
386 # uninstall tiller from cluster
387 self.log.debug("Uninstalling tiller from cluster {}".format(cluster_id))
388 command = "{} --kubeconfig={} --home={} reset".format(
389 self._helm_command, paths["kube_config"], paths["helm_dir"]
390 )
391 self.log.debug("resetting: {}".format(command))
392 output, _rc = await self._local_async_exec(
393 command=command, raise_exception_on_error=True, env=env
394 )
395 # Delete clusterrolebinding and serviceaccount.
396 # Ignore if errors for backward compatibility
397 command = (
398 "{} --kubeconfig={} delete clusterrolebinding.rbac.authorization.k8s."
399 "io/osm-tiller-cluster-rule"
400 ).format(self.kubectl_command, paths["kube_config"])
401 output, _rc = await self._local_async_exec(
402 command=command, raise_exception_on_error=False, env=env
403 )
404 command = (
405 "{} --kubeconfig={} --namespace {} delete serviceaccount/{}".format(
406 self.kubectl_command,
407 paths["kube_config"],
408 namespace,
409 self.service_account,
410 )
411 )
412 output, _rc = await self._local_async_exec(
413 command=command, raise_exception_on_error=False, env=env
414 )
415
416 else:
417 self.log.debug("namespace not found")
418
419 async def _instances_list(self, cluster_id):
420 # init paths, env
421 paths, env = self._init_paths_env(
422 cluster_name=cluster_id, create_if_not_exist=True
423 )
424
425 command = "{} list --output yaml".format(self._helm_command)
426
427 output, _rc = await self._local_async_exec(
428 command=command, raise_exception_on_error=True, env=env
429 )
430
431 if output and len(output) > 0:
432 # parse yaml and update keys to lower case to unify with helm3
433 instances = yaml.load(output, Loader=yaml.SafeLoader).get("Releases")
434 new_instances = []
435 for instance in instances:
436 new_instance = dict((k.lower(), v) for k, v in instance.items())
437 new_instances.append(new_instance)
438 return new_instances
439 else:
440 return []
441
442 def _get_inspect_command(
443 self, show_command: str, kdu_model: str, repo_str: str, version: str
444 ):
445 inspect_command = "{} inspect {} {}{} {}".format(
446 self._helm_command, show_command, kdu_model, repo_str, version
447 )
448 return inspect_command
449
450 def _get_get_command(
451 self, get_command: str, kdu_instance: str, namespace: str, kubeconfig: str
452 ):
453 get_command = "env KUBECONFIG={} {} get {} {} --output yaml".format(
454 kubeconfig, self._helm_command, get_command, kdu_instance
455 )
456 return get_command
457
458 async def _status_kdu(
459 self,
460 cluster_id: str,
461 kdu_instance: str,
462 namespace: str = None,
463 yaml_format: bool = False,
464 show_error_log: bool = False,
465 ) -> Union[str, dict]:
466 self.log.debug(
467 "status of kdu_instance: {}, namespace: {} ".format(kdu_instance, namespace)
468 )
469
470 # init config, env
471 paths, env = self._init_paths_env(
472 cluster_name=cluster_id, create_if_not_exist=True
473 )
474 command = ("env KUBECONFIG={} {} status {} --output yaml").format(
475 paths["kube_config"], self._helm_command, kdu_instance
476 )
477 output, rc = await self._local_async_exec(
478 command=command,
479 raise_exception_on_error=True,
480 show_error_log=show_error_log,
481 env=env,
482 )
483
484 if yaml_format:
485 return str(output)
486
487 if rc != 0:
488 return None
489
490 data = yaml.load(output, Loader=yaml.SafeLoader)
491
492 # remove field 'notes'
493 try:
494 del data.get("info").get("status")["notes"]
495 except KeyError:
496 pass
497
498 # parse the manifest to a list of dictionaries
499 if "manifest" in data:
500 manifest_str = data.get("manifest")
501 manifest_docs = yaml.load_all(manifest_str, Loader=yaml.SafeLoader)
502
503 data["manifest"] = []
504 for doc in manifest_docs:
505 data["manifest"].append(doc)
506
507 # parse field 'resources'
508 try:
509 resources = str(data.get("info").get("status").get("resources"))
510 resource_table = self._output_to_table(resources)
511 data.get("info").get("status")["resources"] = resource_table
512 except Exception:
513 pass
514
515 # set description to lowercase (unify with helm3)
516 try:
517 data.get("info")["description"] = data.get("info").pop("Description")
518 except KeyError:
519 pass
520
521 return data
522
523 def _get_helm_chart_repos_ids(self, cluster_uuid) -> list:
524 repo_ids = []
525 cluster_filter = {"_admin.helm-chart.id": cluster_uuid}
526 cluster = self.db.get_one("k8sclusters", cluster_filter)
527 if cluster:
528 repo_ids = cluster.get("_admin").get("helm_chart_repos") or []
529 return repo_ids
530 else:
531 raise K8sException(
532 "k8cluster with helm-id : {} not found".format(cluster_uuid)
533 )
534
535 async def _is_install_completed(self, cluster_id: str, kdu_instance: str) -> bool:
536 # init config, env
537 paths, env = self._init_paths_env(
538 cluster_name=cluster_id, create_if_not_exist=True
539 )
540
541 status = await self._status_kdu(
542 cluster_id=cluster_id, kdu_instance=kdu_instance, yaml_format=False
543 )
544
545 # extract info.status.resources-> str
546 # format:
547 # ==> v1/Deployment
548 # NAME READY UP-TO-DATE AVAILABLE AGE
549 # halting-horse-mongodb 0/1 1 0 0s
550 # halting-petit-mongodb 1/1 1 0 0s
551 # blank line
552 resources = K8sHelmBaseConnector._get_deep(
553 status, ("info", "status", "resources")
554 )
555
556 # convert to table
557 resources = K8sHelmBaseConnector._output_to_table(resources)
558
559 num_lines = len(resources)
560 index = 0
561 ready = True
562 while index < num_lines:
563 try:
564 line1 = resources[index]
565 index += 1
566 # find '==>' in column 0
567 if line1[0] == "==>":
568 line2 = resources[index]
569 index += 1
570 # find READY in column 1
571 if line2[1] == "READY":
572 # read next lines
573 line3 = resources[index]
574 index += 1
575 while len(line3) > 1 and index < num_lines:
576 ready_value = line3[1]
577 parts = ready_value.split(sep="/")
578 current = int(parts[0])
579 total = int(parts[1])
580 if current < total:
581 self.log.debug("NOT READY:\n {}".format(line3))
582 ready = False
583 line3 = resources[index]
584 index += 1
585
586 except Exception:
587 pass
588
589 return ready
590
591 def _get_install_command(
592 self,
593 kdu_model,
594 kdu_instance,
595 namespace,
596 params_str,
597 version,
598 atomic,
599 timeout,
600 kubeconfig,
601 ) -> str:
602 timeout_str = ""
603 if timeout:
604 timeout_str = "--timeout {}".format(timeout)
605
606 # atomic
607 atomic_str = ""
608 if atomic:
609 atomic_str = "--atomic"
610 # namespace
611 namespace_str = ""
612 if namespace:
613 namespace_str = "--namespace {}".format(namespace)
614
615 # version
616 version_str = ""
617 if version:
618 version_str = "--version {}".format(version)
619
620 command = (
621 "env KUBECONFIG={kubeconfig} {helm} install {atomic} --output yaml "
622 "{params} {timeout} --name={name} {ns} {model} {ver}".format(
623 kubeconfig=kubeconfig,
624 helm=self._helm_command,
625 atomic=atomic_str,
626 params=params_str,
627 timeout=timeout_str,
628 name=kdu_instance,
629 ns=namespace_str,
630 model=kdu_model,
631 ver=version_str,
632 )
633 )
634 return command
635
636 def _get_upgrade_scale_command(
637 self,
638 kdu_model: str,
639 kdu_instance: str,
640 namespace: str,
641 scale: int,
642 version: str,
643 atomic: bool,
644 replica_str: str,
645 timeout: float,
646 resource_name: str,
647 kubeconfig: str,
648 ) -> str:
649 """Generates the command to scale a Helm Chart release
650
651 Args:
652 kdu_model (str): Kdu model name, corresponding to the Helm local location or repository
653 kdu_instance (str): KDU instance, corresponding to the Helm Chart release in question
654 namespace (str): Namespace where this KDU instance is deployed
655 scale (int): Scale count
656 version (str): Constraint with specific version of the Chart to use
657 atomic (bool): If set, upgrade process rolls back changes made in case of failed upgrade.
658 The --wait flag will be set automatically if --atomic is used
659 replica_str (str): The key under resource_name key where the scale count is stored
660 timeout (float): The time, in seconds, to wait
661 resource_name (str): The KDU's resource to scale
662 kubeconfig (str): Kubeconfig file path
663
664 Returns:
665 str: command to scale a Helm Chart release
666 """
667
668 # scale
669 if resource_name:
670 scale_dict = {"{}.{}".format(resource_name, replica_str): scale}
671 else:
672 scale_dict = {replica_str: scale}
673
674 scale_str = self._params_to_set_option(scale_dict)
675
676 return self._get_upgrade_command(
677 kdu_model=kdu_model,
678 kdu_instance=kdu_instance,
679 namespace=namespace,
680 params_str=scale_str,
681 version=version,
682 atomic=atomic,
683 timeout=timeout,
684 kubeconfig=kubeconfig,
685 )
686
687 def _get_upgrade_command(
688 self,
689 kdu_model,
690 kdu_instance,
691 namespace,
692 params_str,
693 version,
694 atomic,
695 timeout,
696 kubeconfig,
697 force: bool = False,
698 ) -> str:
699 """Generates the command to upgrade a Helm Chart release
700
701 Args:
702 kdu_model (str): Kdu model name, corresponding to the Helm local location or repository
703 kdu_instance (str): KDU instance, corresponding to the Helm Chart release in question
704 namespace (str): Namespace where this KDU instance is deployed
705 params_str (str): Params used to upgrade the Helm Chart release
706 version (str): Constraint with specific version of the Chart to use
707 atomic (bool): If set, upgrade process rolls back changes made in case of failed upgrade.
708 The --wait flag will be set automatically if --atomic is used
709 timeout (float): The time, in seconds, to wait
710 kubeconfig (str): Kubeconfig file path
711 force (bool): If set, helm forces resource updates through a replacement strategy. This may recreate pods.
712 Returns:
713 str: command to upgrade a Helm Chart release
714 """
715
716 timeout_str = ""
717 if timeout:
718 timeout_str = "--timeout {}".format(timeout)
719
720 # atomic
721 atomic_str = ""
722 if atomic:
723 atomic_str = "--atomic"
724
725 # force
726 force_str = ""
727 if force:
728 force_str = "--force "
729
730 # version
731 version_str = ""
732 if version:
733 version_str = "--version {}".format(version)
734
735 # namespace
736 namespace_str = ""
737 if namespace:
738 namespace_str = "--namespace {}".format(namespace)
739
740 command = (
741 "env KUBECONFIG={kubeconfig} {helm} upgrade {namespace} {atomic} --output yaml {params} {timeout} {force}"
742 "--reuse-values {name} {model} {ver}"
743 ).format(
744 kubeconfig=kubeconfig,
745 helm=self._helm_command,
746 namespace=namespace_str,
747 atomic=atomic_str,
748 force=force_str,
749 params=params_str,
750 timeout=timeout_str,
751 name=kdu_instance,
752 model=kdu_model,
753 ver=version_str,
754 )
755 return command
756
757 def _get_rollback_command(
758 self, kdu_instance, namespace, revision, kubeconfig
759 ) -> str:
760 return "env KUBECONFIG={} {} rollback {} {} --wait".format(
761 kubeconfig, self._helm_command, kdu_instance, revision
762 )
763
764 def _get_uninstall_command(
765 self, kdu_instance: str, namespace: str, kubeconfig: str
766 ) -> str:
767 return "env KUBECONFIG={} {} delete --purge {}".format(
768 kubeconfig, self._helm_command, kdu_instance
769 )