2 # Copyright (C) The Arvados Authors. All rights reserved.
4 # SPDX-License-Identifier: Apache-2.0
7 # Migration tool for merging user accounts belonging to the same user
8 # but on separate clusters to use a single user account managed by a
11 # If you're working on this, see
12 # arvados/sdk/python/tests/fed-migrate/README for information about
13 # the testing infrastructure.
25 from arvados._version import __version__
32 def connect_clusters(args):
37 print("Reading %s" % args.tokens)
38 with open(args.tokens, "rt") as f:
39 for r in csv.reader(f):
44 print("Contacting %s" % (host))
45 arv = arvados.api(host=host, token=token, cache=False)
46 clusters[arv._rootDesc["uuidPrefix"]] = arv
48 arv = arvados.api(cache=False)
49 rh = arv._rootDesc["remoteHosts"]
50 tok = arv.api_client_authorizations().current().execute()
51 token = "v2/%s/%s" % (tok["uuid"], tok["api_token"])
53 for k,v in rh.items():
54 arv = arvados.api(host=v, token=token, cache=False, insecure=os.environ.get("ARVADOS_API_HOST_INSECURE"))
57 for _, arv in clusters.items():
58 config = arv.configs().get().execute()
59 if config["Login"]["LoginCluster"] != "" and loginCluster is None:
60 loginCluster = config["Login"]["LoginCluster"]
62 print("Checking that the federation is well connected")
63 for arv in clusters.values():
64 config = arv.configs().get().execute()
65 if loginCluster and config["Login"]["LoginCluster"] != loginCluster and config["ClusterID"] != loginCluster:
66 errors.append("Inconsistent login cluster configuration, expected '%s' on %s but was '%s'" % (loginCluster, config["ClusterID"], config["Login"]["LoginCluster"]))
69 if arv._rootDesc["revision"] < "20190926":
70 errors.append("Arvados API server revision on cluster '%s' is too old, must be updated to at least Arvados 1.5 before running migration." % config["ClusterID"])
74 cur = arv.users().current().execute()
75 except arvados.errors.ApiError as e:
76 errors.append("checking token for %s %s" % (arv._rootDesc["rootUrl"], e))
79 if not cur["is_admin"]:
80 errors.append("User %s is not admin on %s" % (cur["uuid"], arv._rootDesc["uuidPrefix"]))
84 if r != arv._rootDesc["uuidPrefix"] and r not in arv._rootDesc["remoteHosts"]:
85 errors.append("%s is missing from remoteHosts of %s" % (r, arv._rootDesc["uuidPrefix"]))
86 for r in arv._rootDesc["remoteHosts"]:
87 if r != "*" and r not in clusters:
88 print("WARNING: %s is federated with %s but %s is missing from the tokens file or the token is invalid" % (arv._rootDesc["uuidPrefix"], r, r))
90 return clusters, errors, loginCluster
93 def fetch_users(clusters, loginCluster):
99 for c, arv in clusters.items():
100 print("Getting user list from %s" % c)
101 ul = arvados.util.list_all(arv.users().list)
103 if l["uuid"].startswith(c):
106 # Users list is sorted by email
107 # Go through users and collect users with same email
108 # when we see a different email (or get to the end)
109 # call add_accum_rows() to generate the report rows with
110 # the "home cluster" set, and also fill in the by_email table.
112 users = sorted(users, key=lambda u: u["email"]+"::"+(u["username"] or "")+"::"+u["uuid"])
117 def add_accum_rows():
120 uuids = set(a["uuid"] for a in accum)
121 homeuuid = ((len(uuids) == 1) and uuids.pop()) or ""
123 r = (a["email"], a["username"], a["uuid"], loginCluster or homeuuid[0:5])
124 by_email.setdefault(a["email"], {})
125 by_email[a["email"]][a["uuid"]] = r
126 homeuuid_and_username = "%s::%s" % (r[HOMECLUSTER], a["username"])
127 if homeuuid_and_username not in by_username:
128 by_username[homeuuid_and_username] = a["email"]
129 elif by_username[homeuuid_and_username] != a["email"]:
130 print("ERROR: the username '%s' is listed for both '%s' and '%s' on cluster '%s'" % (r[USERNAME], r[EMAIL], by_username[homeuuid_and_username], r[HOMECLUSTER]))
135 if u["uuid"].endswith("-anonymouspublic") or u["uuid"].endswith("-000000000000000"):
137 if lastemail == None:
138 lastemail = u["email"]
139 if u["email"] == lastemail:
143 lastemail = u["email"]
148 return rows, by_email, by_username
151 def read_migrations(args, by_email, by_username):
153 with open(args.migrate or args.dry_run, "rt") as f:
154 for r in csv.reader(f):
155 if r[EMAIL] == "email":
157 by_email.setdefault(r[EMAIL], {})
158 by_email[r[EMAIL]][r[UUID]] = r
160 homeuuid_and_username = "%s::%s" % (r[HOMECLUSTER], r[USERNAME])
161 if homeuuid_and_username not in by_username:
162 by_username[homeuuid_and_username] = r[EMAIL]
163 elif by_username[homeuuid_and_username] != r[EMAIL]:
164 print("ERROR: the username '%s' is listed for both '%s' and '%s' on cluster '%s'" % (r[USERNAME], r[EMAIL], by_username[homeuuid_and_username], r[HOMECLUSTER]))
170 def update_username(args, email, user_uuid, username, migratecluster, migratearv):
171 print("(%s) Updating username of %s to '%s' on %s" % (email, user_uuid, username, migratecluster))
174 conflicts = migratearv.users().list(filters=[["username", "=", username]]).execute()
175 if conflicts["items"]:
176 migratearv.users().update(uuid=conflicts["items"][0]["uuid"], body={"user": {"username": username+"migrate"}}).execute()
177 migratearv.users().update(uuid=user_uuid, body={"user": {"username": username}}).execute()
178 except arvados.errors.ApiError as e:
179 print("(%s) Error updating username of %s to '%s' on %s: %s" % (email, user_uuid, username, migratecluster, e))
182 def choose_new_user(args, by_email, email, userhome, username, old_user_uuid, clusters):
185 for b in by_email[email].values():
186 if b[2].startswith(userhome):
188 if b[1] != username and b[3] == userhome:
189 print("(%s) Cannot migrate %s, conflicting usernames %s and %s" % (email, old_user_uuid, b[1], username))
194 if len(candidates) == 0:
195 if len(userhome) == 5 and userhome not in clusters:
196 print("(%s) Cannot migrate %s, unknown home cluster %s (typo?)" % (email, old_user_uuid, userhome))
198 print("(%s) No user listed with same email to migrate %s to %s, will create new user with username '%s'" % (email, old_user_uuid, userhome, username))
200 oldhomecluster = old_user_uuid[0:5]
201 oldhomearv = clusters[oldhomecluster]
202 newhomecluster = userhome[0:5]
203 homearv = clusters[userhome]
206 olduser = oldhomearv.users().get(uuid=old_user_uuid).execute()
207 conflicts = homearv.users().list(filters=[["username", "=", username]]).execute()
208 if conflicts["items"]:
209 homearv.users().update(uuid=conflicts["items"][0]["uuid"], body={"user": {"username": username+"migrate"}}).execute()
210 user = homearv.users().create(body={"user": {"email": email, "username": username, "is_active": olduser["is_active"]}}).execute()
211 except arvados.errors.ApiError as e:
212 print("(%s) Could not create user: %s" % (email, str(e)))
215 tup = (email, username, user["uuid"], userhome)
218 tup = (email, username, "%s-tpzed-xfakexfakexfake" % (userhome[0:5]), userhome)
219 by_email[email][tup[2]] = tup
220 candidates.append(tup)
221 if len(candidates) > 1:
222 print("(%s) Multiple users listed to migrate %s to %s, use full uuid" % (email, old_user_uuid, userhome))
224 return candidates[0][2]
227 def activate_remote_user(args, email, homearv, migratearv, old_user_uuid, new_user_uuid):
228 # create a token for the new user and salt it for the
229 # migration cluster, then use it to access the migration
230 # cluster as the new user once before merging to ensure
231 # the new user is known on that cluster.
232 migratecluster = migratearv._rootDesc["uuidPrefix"]
235 newtok = homearv.api_client_authorizations().create(body={
236 "api_client_authorization": {'owner_uuid': new_user_uuid}}).execute()
238 newtok = {"uuid": "dry-run", "api_token": "12345"}
239 except arvados.errors.ApiError as e:
240 print("(%s) Could not create API token for %s: %s" % (email, new_user_uuid, e))
244 olduser = migratearv.users().get(uuid=old_user_uuid).execute()
245 except arvados.errors.ApiError as e:
246 if e.resp.status != 404:
247 print("(%s) Could not retrieve user %s from %s, user may have already been migrated: %s" % (email, old_user_uuid, migratecluster, e))
250 salted = 'v2/' + newtok["uuid"] + '/' + hmac.new(newtok["api_token"].encode(),
251 msg=migratecluster.encode(),
252 digestmod=hashlib.sha1).hexdigest()
254 ru = urllib.parse.urlparse(migratearv._rootDesc["rootUrl"])
256 newuser = arvados.api(host=ru.netloc, token=salted, insecure=os.environ.get("ARVADOS_API_HOST_INSECURE")).users().current().execute()
258 newuser = {"is_active": True, "username": username}
259 except arvados.errors.ApiError as e:
260 print("(%s) Error getting user info for %s from %s: %s" % (email, new_user_uuid, migratecluster, e))
263 if not newuser["is_active"] and olduser["is_active"]:
264 print("(%s) Activating user %s on %s" % (email, new_user_uuid, migratecluster))
267 migratearv.users().update(uuid=new_user_uuid, body={"is_active": True}).execute()
268 except arvados.errors.ApiError as e:
269 print("(%s) Could not activate user %s on %s: %s" % (email, new_user_uuid, migratecluster, e))
272 if olduser["is_admin"] and not newuser["is_admin"]:
273 print("(%s) Not migrating %s because user is admin but target user %s is not admin on %s" % (email, old_user_uuid, new_user_uuid, migratecluster))
278 def migrate_user(args, migratearv, email, new_user_uuid, old_user_uuid):
281 grp = migratearv.groups().create(body={
282 "owner_uuid": new_user_uuid,
283 "name": "Migrated from %s (%s)" % (email, old_user_uuid),
284 "group_class": "project"
285 }, ensure_unique_name=True).execute()
286 migratearv.users().merge(old_user_uuid=old_user_uuid,
287 new_user_uuid=new_user_uuid,
288 new_owner_uuid=grp["uuid"],
289 redirect_to_new_user=True).execute()
290 except arvados.errors.ApiError as e:
291 print("(%s) Error migrating user: %s" % (email, e))
296 parser = argparse.ArgumentParser(description='Migrate users to federated identity, see https://doc.arvados.org/admin/merge-remote-account.html')
298 '--version', action='version', version="%s %s" % (sys.argv[0], __version__),
299 help='Print version and exit.')
300 parser.add_argument('--tokens', type=str, required=False)
301 group = parser.add_mutually_exclusive_group(required=True)
302 group.add_argument('--report', type=str, help="Generate report .csv file listing users by email address and their associated Arvados accounts")
303 group.add_argument('--migrate', type=str, help="Consume report .csv and migrate users to designated Arvados accounts")
304 group.add_argument('--dry-run', type=str, help="Consume report .csv and report how user would be migrated to designated Arvados accounts")
305 group.add_argument('--check', action="store_true", help="Check that tokens are usable and the federation is well connected")
306 args = parser.parse_args()
308 clusters, errors, loginCluster = connect_clusters(args)
312 print("ERROR: "+str(e))
316 print("Tokens file passed checks")
319 rows, by_email, by_username = fetch_users(clusters, loginCluster)
322 out = csv.writer(open(args.report, "wt"))
323 out.writerow(("email", "username", "user uuid", "home cluster"))
326 print("Wrote %s" % args.report)
329 if args.migrate or args.dry_run:
331 print("Performing dry run")
333 rows = read_migrations(args, by_email, by_username)
337 username = r[USERNAME]
338 old_user_uuid = r[UUID]
339 userhome = r[HOMECLUSTER]
342 print("(%s) Skipping %s, no home cluster specified" % (email, old_user_uuid))
343 if old_user_uuid.startswith(userhome):
344 migratecluster = old_user_uuid[0:5]
345 migratearv = clusters[migratecluster]
346 if migratearv.users().get(uuid=old_user_uuid).execute()["username"] != username:
347 update_username(args, email, old_user_uuid, username, migratecluster, migratearv)
350 new_user_uuid = choose_new_user(args, by_email, email, userhome, username, old_user_uuid, clusters)
351 if new_user_uuid is None:
354 # cluster where the migration is happening
355 for migratecluster in clusters:
356 migratearv = clusters[migratecluster]
358 # the user's new home cluster
359 newhomecluster = userhome[0:5]
360 homearv = clusters[newhomecluster]
362 newuser = activate_remote_user(args, email, homearv, migratearv, old_user_uuid, new_user_uuid)
366 print("(%s) Migrating %s to %s on %s" % (email, old_user_uuid, new_user_uuid, migratecluster))
368 migrate_user(args, migratearv, email, new_user_uuid, old_user_uuid)
370 if newuser['username'] != username:
371 update_username(args, email, new_user_uuid, username, migratecluster, migratearv)
373 if __name__ == "__main__":