1 # Copyright (C) The Arvados Authors. All rights reserved.
3 # SPDX-License-Identifier: AGPL-3.0
5 # Do not use this file for site configuration. Create
6 # /etc/arvados/config.yml instead.
8 # The order of precedence (highest to lowest):
9 # 1. Legacy component-specific config files (deprecated)
10 # 2. /etc/arvados/config.yml
11 # 3. config.default.yml
17 # Token to be included in all healthcheck requests. Disabled by default.
18 # Server expects request header of the format "Authorization: Bearer xxx"
23 # In each of the service sections below, the keys under
24 # InternalURLs are the endpoints where the service should be
25 # listening, and reachable from other hosts in the cluster.
28 "http://example.host:12345": {}
61 # Base URL for Workbench inline preview. If blank, use
62 # WebDAVDownload instead, and disable inline preview.
63 # If both are empty, downloading collections from workbench
66 # It is important to properly configure the download service
67 # to migitate cross-site-scripting (XSS) attacks. A HTML page
68 # can be stored in collection. If an attacker causes a victim
69 # to visit that page through Workbench, it will be rendered by
70 # the browser. If all collections are served at the same
71 # domain, the browser will consider collections as coming from
72 # the same origin and having access to the same browsing data,
73 # enabling malicious Javascript on that page to access Arvados
74 # on behalf of the victim.
76 # This is mitigating by having separate domains for each
77 # collection, or limiting preview to circumstances where the
78 # collection is not accessed with the user's regular
81 # Serve preview links using uuid or pdh in subdomain
82 # (requires wildcard DNS and TLS certificate)
83 # https://*.collections.uuid_prefix.arvadosapi.com
85 # Serve preview links using uuid or pdh in main domain
86 # (requires wildcard DNS and TLS certificate)
87 # https://*--collections.uuid_prefix.arvadosapi.com
89 # Serve preview links by setting uuid or pdh in the path.
90 # This configuration only allows previews of public data or
91 # collection-sharing links, because these use the anonymous
92 # user token or the token is already embedded in the URL.
93 # Other data must be handled as downloads via WebDAVDownload:
94 # https://collections.uuid_prefix.arvadosapi.com
100 # Base URL for download links. If blank, serve links to WebDAV
101 # with disposition=attachment query param. Unlike preview links,
102 # browsers do not render attachments, so there is no risk of XSS.
104 # If WebDAVDownload is blank, and WebDAV uses a
105 # single-origin form, then Workbench will show an error page
107 # Serve download links by setting uuid or pdh in the path:
108 # https://download.uuid_prefix.arvadosapi.com
120 # ShellInABox service endpoint URL for a given VM. If empty, do not
121 # offer web shell logins.
123 # E.g., using a path-based proxy server to forward connections to shell hosts:
124 # https://webshell.uuid_prefix.arvadosapi.com
126 # E.g., using a name-based proxy server to forward connections to shell hosts:
127 # https://*.webshell.uuid_prefix.arvadosapi.com
143 # max concurrent connections per arvados server daemon
146 # All parameters here are passed to the PG client library in a connection string;
147 # see https://www.postgresql.org/docs/current/static/libpq-connect.html#LIBPQ-PARAMKEYWORDS
155 # Maximum size (in bytes) allowed for a single API request. This
156 # limit is published in the discovery document for use by clients.
157 # Note: You must separately configure the upstream web server or
158 # proxy to actually enforce the desired maximum request size on the
160 MaxRequestSize: 134217728
162 # Limit the number of bytes read from the database during an index
163 # request (by retrieving and returning fewer rows than would
164 # normally be returned in a single response).
165 # Note 1: This setting never reduces the number of returned rows to
166 # zero, no matter how big the first data row is.
167 # Note 2: Currently, this is only checked against a specific set of
168 # columns that tend to get large (collections.manifest_text,
169 # containers.mounts, workflows.definition). Other fields (e.g.,
170 # "properties" hashes) are not counted against this limit.
171 MaxIndexDatabaseRead: 134217728
173 # Maximum number of items to return when responding to a APIs that
174 # can return partial result sets using limit and offset parameters
175 # (e.g., *.index, groups.contents). If a request specifies a "limit"
176 # parameter higher than this value, this value is used instead.
177 MaxItemsPerResponse: 1000
179 # API methods to disable. Disabled methods are not listed in the
180 # discovery document, and respond 404 to all requests.
181 # Example: ["jobs.create", "pipeline_instances.create"]
184 # Interval (seconds) between asynchronous permission view updates. Any
185 # permission-updating API called with the 'async' parameter schedules a an
186 # update on the permission view in the future, if not already scheduled.
187 AsyncPermissionsUpdateInterval: 20s
189 # Maximum number of concurrent outgoing requests to make while
190 # serving a single incoming multi-cluster (federated) request.
191 MaxRequestAmplification: 4
193 # RailsSessionSecretToken is a string of alphanumeric characters
194 # used by Rails to sign session tokens. IMPORTANT: This is a
195 # site secret. It should be at least 50 characters.
196 RailsSessionSecretToken: ""
198 # Maximum wall clock time to spend handling an incoming request.
202 # Config parameters to automatically setup new users. If enabled,
203 # this users will be able to self-activate. Enable this if you want
204 # to run an open instance where anyone can create an account and use
205 # the system without requiring manual approval.
207 # The params auto_setup_new_users_with_* are meaningful only when auto_setup_new_users is turned on.
208 # auto_setup_name_blacklist is a list of usernames to be blacklisted for auto setup.
209 AutoSetupNewUsers: false
210 AutoSetupNewUsersWithVmUUID: ""
211 AutoSetupNewUsersWithRepository: false
212 AutoSetupUsernameBlacklist: [arvados, git, gitolite, gitolite-admin, root, syslog]
214 # When new_users_are_active is set to true, new users will be active
215 # immediately. This skips the "self-activate" step which enforces
216 # user agreements. Should only be enabled for development.
217 NewUsersAreActive: false
219 # The e-mail address of the user you would like to become marked as an admin
220 # user on their first login.
221 # In the default configuration, authentication happens through the Arvados SSO
222 # server, which uses OAuth2 against Google's servers, so in that case this
223 # should be an address associated with a Google account.
224 AutoAdminUserWithEmail: ""
226 # If auto_admin_first_user is set to true, the first user to log in when no
227 # other admin users exist will automatically become an admin user.
228 AutoAdminFirstUser: false
230 # Email address to notify whenever a user creates a profile for the
232 UserProfileNotificationAddress: ""
233 AdminNotifierEmailFrom: arvados@example.com
234 EmailSubjectPrefix: "[ARVADOS] "
235 UserNotifierEmailFrom: arvados@example.com
236 NewUserNotificationRecipients: []
237 NewInactiveUserNotificationRecipients: []
239 # Set anonymous_user_token to enable anonymous user access. You can get
240 # the token by running "bundle exec ./script/get_anonymous_user_token.rb"
241 # in the directory where your API server is running.
242 AnonymousUserToken: ""
245 # Time to keep audit logs, in seconds. (An audit log is a row added
246 # to the "logs" table in the PostgreSQL database each time an
247 # Arvados object is created, modified, or deleted.)
249 # Currently, websocket event notifications rely on audit logs, so
250 # this should not be set lower than 300 (5 minutes).
253 # Maximum number of log rows to delete in a single SQL transaction.
255 # If max_audit_log_delete_batch is 0, log entries will never be
256 # deleted by Arvados. Cleanup can be done by an external process
257 # without affecting any Arvados system processes, as long as very
258 # recent (<5 minutes old) logs are not deleted.
260 # 100000 is a reasonable batch size for most sites.
263 # Attributes to suppress in events and audit logs. Notably,
264 # specifying ["manifest_text"] here typically makes the database
265 # smaller and faster.
267 # Warning: Using any non-empty value here can have undesirable side
268 # effects for any client or component that relies on event logs.
269 # Use at your own risk.
270 UnloggedAttributes: []
274 # Logging threshold: panic, fatal, error, warn, info, debug, or
278 # Logging format: json or text
281 # Maximum characters of (JSON-encoded) query parameters to include
282 # in each request log entry. When params exceed this size, they will
283 # be JSON-encoded, truncated to this size, and logged as
285 MaxRequestLogParamsSize: 2000
288 # Allow clients to create collections by providing a manifest with
289 # unsigned data blob locators. IMPORTANT: This effectively disables
290 # access controls for data stored in Keep: a client who knows a hash
291 # can write a manifest that references the hash, pass it to
292 # collections.create (which will create a permission link), use
293 # collections.get to obtain a signature for that data locator, and
294 # use that signed locator to retrieve the data from Keep. Therefore,
295 # do not turn this on if your users expect to keep data private from
299 # BlobSigningKey is a string of alphanumeric characters used to
300 # generate permission signatures for Keep locators. It must be
301 # identical to the permission key given to Keep. IMPORTANT: This is
302 # a site secret. It should be at least 50 characters.
304 # Modifying blob_signing_key will invalidate all existing
305 # signatures, which can cause programs to fail (e.g., arv-put,
306 # arv-get, and Crunch jobs). To avoid errors, rotate keys only when
307 # no such processes are running.
310 # Default replication level for collections. This is used when a
311 # collection's replication_desired attribute is nil.
312 DefaultReplication: 2
314 # Lifetime (in seconds) of blob permission signatures generated by
315 # the API server. This determines how long a client can take (after
316 # retrieving a collection record) to retrieve the collection data
317 # from Keep. If the client needs more time than that (assuming the
318 # collection still has the same content and the relevant user/token
319 # still has permission) the client can retrieve the collection again
320 # to get fresh signatures.
322 # This must be exactly equal to the -blob-signature-ttl flag used by
323 # keepstore servers. Otherwise, reading data blocks and saving
324 # collections will fail with HTTP 403 permission errors.
326 # Modifying blob_signature_ttl invalidates existing signatures; see
327 # blob_signing_key note above.
329 # The default is 2 weeks.
332 # Default lifetime for ephemeral collections: 2 weeks. This must not
333 # be less than blob_signature_ttl.
334 DefaultTrashLifetime: 336h
336 # Interval (seconds) between trash sweeps. During a trash sweep,
337 # collections are marked as trash if their trash_at time has
338 # arrived, and deleted if their delete_at time has arrived.
339 TrashSweepInterval: 60s
341 # If true, enable collection versioning.
342 # When a collection's preserve_version field is true or the current version
343 # is older than the amount of seconds defined on preserve_version_if_idle,
344 # a snapshot of the collection's previous state is created and linked to
345 # the current collection.
346 CollectionVersioning: false
348 # 0s = auto-create a new version on every update.
349 # -1s = never auto-create new versions.
350 # > 0s = auto-create a new version when older than the specified number of seconds.
351 PreserveVersionIfIdle: -1s
353 # Managed collection properties. At creation time, if the client didn't
354 # provide the listed keys, they will be automatically populated following
355 # one of the following behaviors:
357 # * UUID of the user who owns the containing project.
358 # responsible_person_uuid: {Function: original_owner, Protected: true}
360 # * Default concrete value.
361 # foo_bar: {Value: baz, Protected: false}
363 # If Protected is true, only an admin user can modify its value.
365 SAMPLE: {Function: original_owner, Protected: true}
367 # In "trust all content" mode, Workbench will redirect download
368 # requests to WebDAV preview link, even in the cases when
369 # WebDAV would have to expose XSS vulnerabilities in order to
370 # handle the redirect (see discussion on Services.WebDAV).
372 # This setting has no effect in the recommended configuration,
373 # where the WebDAV is configured to have a separate domain for
374 # every collection; in this case XSS protection is provided by
375 # browsers' same-origin policy.
377 # The default setting (false) is appropriate for a multi-user site.
378 TrustAllContent: false
381 # These settings are provided by your OAuth2 provider (e.g.,
383 ProviderAppSecret: ""
387 # Git repositories must be readable by api server, or you won't be
388 # able to submit crunch jobs. To pass the test suites, put a clone
389 # of the arvados tree in {git_repositories_dir}/arvados.git or
390 # {git_repositories_dir}/arvados/.git
391 Repositories: /var/lib/arvados/git/repositories
399 # List of supported Docker Registry image formats that compute nodes
400 # are able to use. `arv keep docker` will error out if a user tries
401 # to store an image with an unsupported format. Use an empty array
402 # to skip the compatibility check (and display a warning message to
405 # Example for sites running docker < 1.10: ["v1"]
406 # Example for sites running docker >= 1.10: ["v2"]
407 # Example for disabling check: []
408 SupportedDockerImageFormats: ["v2"]
410 # Include details about job reuse decisions in the server log. This
411 # causes additional database queries to run, so it should not be
412 # enabled unless you expect to examine the resulting logs for
413 # troubleshooting purposes.
414 LogReuseDecisions: false
416 # Default value for keep_cache_ram of a container's runtime_constraints.
417 DefaultKeepCacheRAM: 268435456
419 # Number of times a container can be unlocked before being
420 # automatically cancelled.
421 MaxDispatchAttempts: 5
423 # Default value for container_count_max for container requests. This is the
424 # number of times Arvados will create a new container to satisfy a container
425 # request. If a container is cancelled it will retry a new container if
426 # container_count < container_count_max on any container requests associated
427 # with the cancelled container.
430 # The maximum number of compute nodes that can be in use simultaneously
431 # If this limit is reduced, any existing nodes with slot number >= new limit
432 # will not be counted against the new limit. In other words, the new limit
433 # won't be strictly enforced until those nodes with higher slot numbers
437 # Preemptible instance support (e.g. AWS Spot Instances)
438 # When true, child containers will get created with the preemptible
439 # scheduling parameter parameter set.
440 UsePreemptibleInstances: false
442 # PEM encoded SSH key (RSA, DSA, or ECDSA) used by the
443 # (experimental) cloud dispatcher for executing containers on
444 # worker VMs. Begins with "-----BEGIN RSA PRIVATE KEY-----\n"
445 # and ends with "\n-----END RSA PRIVATE KEY-----\n".
446 DispatchPrivateKey: none
448 # Maximum time to wait for workers to come up before abandoning
449 # stale locks from a previous dispatch process.
453 # When you run the db:delete_old_container_logs task, it will find
454 # containers that have been finished for at least this many seconds,
455 # and delete their stdout, stderr, arv-mount, crunch-run, and
456 # crunchstat logs from the logs table.
459 # These two settings control how frequently log events are flushed to the
460 # database. Log lines are buffered until either crunch_log_bytes_per_event
461 # has been reached or crunch_log_seconds_between_events has elapsed since
463 LogBytesPerEvent: 4096
464 LogSecondsBetweenEvents: 1
466 # The sample period for throttling logs.
467 LogThrottlePeriod: 60s
469 # Maximum number of bytes that job can log over crunch_log_throttle_period
470 # before being silenced until the end of the period.
471 LogThrottleBytes: 65536
473 # Maximum number of lines that job can log over crunch_log_throttle_period
474 # before being silenced until the end of the period.
475 LogThrottleLines: 1024
477 # Maximum bytes that may be logged by a single job. Log bytes that are
478 # silenced by throttling are not counted against this total.
479 LimitLogBytesPerJob: 67108864
481 LogPartialLineThrottlePeriod: 5s
483 # Container logs are written to Keep and saved in a
484 # collection, which is updated periodically while the
485 # container runs. This value sets the interval between
486 # collection updates.
489 # The log collection is also updated when the specified amount of
490 # log data (given in bytes) is produced in less than one update
496 # Path to dns server configuration directory
497 # (e.g. /etc/unbound.d/conf.d). If false, do not write any config
498 # files or touch restart.txt (see below).
501 # Template file for the dns server host snippets. See
502 # unbound.template in this directory for an example. If false, do
503 # not write any config files.
504 DNSServerConfTemplate: ""
506 # String to write to {dns_server_conf_dir}/restart.txt (with a
507 # trailing newline) after updating local data. If false, do not
508 # open or write the restart.txt file.
509 DNSServerReloadCommand: ""
511 # Command to run after each DNS update. Template variables will be
512 # substituted; see the "unbound" example below. If false, do not run
514 DNSServerUpdateCommand: ""
516 ComputeNodeDomain: ""
517 ComputeNodeNameservers:
520 # Hostname to assign to a compute node when it sends a "ping" and the
521 # hostname in its Node record is nil.
522 # During bootstrapping, the "ping" script is expected to notice the
523 # hostname given in the ping response, and update its unix hostname
525 # If false, leave the hostname alone (this is appropriate if your compute
526 # nodes' hostnames are already assigned by some other mechanism).
528 # One way or another, the hostnames of your node records should agree
529 # with your DNS records and your /etc/slurm-llnl/slurm.conf files.
531 # Example for compute0000, compute0001, ....:
532 # assign_node_hostname: compute%<slot_number>04d
533 # (See http://ruby-doc.org/core-2.2.2/Kernel.html#method-i-format for more.)
534 AssignNodeHostname: "compute%<slot_number>d"
537 # Enable the legacy Jobs API. This value must be a string.
538 # 'auto' -- (default) enable the Jobs API only if it has been used before
539 # (i.e., there are job records in the database)
540 # 'true' -- enable the Jobs API despite lack of existing records.
541 # 'false' -- disable the Jobs API despite presence of existing records.
544 # Git repositories must be readable by api server, or you won't be
545 # able to submit crunch jobs. To pass the test suites, put a clone
546 # of the arvados tree in {git_repositories_dir}/arvados.git or
547 # {git_repositories_dir}/arvados/.git
548 GitInternalDir: /var/lib/arvados/internal.git
550 # Docker image to be used when none found in runtime_constraints of a job
551 DefaultDockerImage: ""
553 # none or slurm_immediate
554 CrunchJobWrapper: none
556 # username, or false = do not set uid when running jobs.
557 CrunchJobUser: crunch
559 # The web service must be able to create/write this file, and
560 # crunch-job must be able to stat() it.
561 CrunchRefreshTrigger: /tmp/crunch_refresh_trigger
563 # Control job reuse behavior when two completed jobs match the
564 # search criteria and have different outputs.
566 # If true, in case of a conflict, reuse the earliest job (this is
567 # similar to container reuse behavior).
569 # If false, in case of a conflict, do not reuse any completed job,
570 # but do reuse an already-running job if available (this is the
571 # original job reuse behavior, and is still the default).
572 ReuseJobIfOutputsDiffer: false
575 # Enable the cloud scheduler (experimental).
578 # Name/number of port where workers' SSH services listen.
581 # Interval between queue polls.
584 # Shell command to execute on each worker to determine whether
585 # the worker is booted and ready to run containers. It should
586 # exit zero if the worker is ready.
587 BootProbeCommand: "docker ps -q"
589 # Minimum interval between consecutive probes to a single
593 # Maximum probes per second, across all workers in a pool.
594 MaxProbesPerSecond: 10
596 # Time before repeating SIGTERM when killing a container.
599 # Time to give up on SIGTERM and write off the worker.
602 # Maximum create/destroy-instance operations per second (0 =
604 MaxCloudOpsPerSecond: 0
606 # Interval between cloud provider syncs/updates ("list all
610 # Time to leave an idle worker running (in case new containers
611 # appear in the queue that it can run) before shutting it
615 # Time to wait for a new worker to boot (i.e., pass
616 # BootProbeCommand) before giving up and shutting it down.
619 # Maximum time a worker can stay alive with no successful
620 # probes before being automatically shut down.
623 # Time after shutting down a worker to retry the
624 # shutdown/destroy operation.
627 # Worker VM image ID.
630 # Tags to add on all resources (VMs, NICs, disks) created by
631 # the container dispatcher. (Arvados's own tags --
632 # InstanceType, IdleBehavior, and InstanceSecret -- will also
637 # Prefix for predefined tags used by Arvados (InstanceSetID,
638 # InstanceType, InstanceSecret, IdleBehavior). With the
639 # default value "Arvados", tags are "ArvadosInstanceSetID",
640 # "ArvadosInstanceSecret", etc.
642 # This should only be changed while no cloud resources are in
643 # use and the cloud dispatcher is not running. Otherwise,
644 # VMs/resources that were added using the old tag prefix will
645 # need to be detected and cleaned up manually.
646 TagKeyPrefix: Arvados
648 # Cloud driver: "azure" (Microsoft Azure) or "ec2" (Amazon AWS).
651 # Cloud-specific driver parameters.
658 # (ec2) Instance configuration.
664 AdminUsername: debian
666 # (azure) Credentials.
672 # (azure) Instance configuration.
673 CloudEnvironment: AzurePublicCloud
680 DeleteDanglingResourcesAfter: 20s
681 AdminUsername: arvados
685 # Use the instance type name as the key (in place of "SAMPLE" in
686 # this sample entry).
688 # Cloud provider's instance type. Defaults to the configured type name.
692 IncludedScratch: 16GB
700 SendUserSetupNotificationEmail: true
702 # Bug/issue report notification to and from addresses
703 IssueReporterEmailFrom: "arvados@example.com"
704 IssueReporterEmailTo: "arvados@example.com"
705 SupportEmailAddress: "arvados@example.com"
707 # Generic issue email from
708 EmailFrom: "arvados@example.com"
717 # API endpoint host or host:port; default is {id}.arvadosapi.com
718 Host: sample.arvadosapi.com
720 # Perform a proxy request when a local client requests an
721 # object belonging to this remote.
724 # Default "https". Can be set to "http" for testing.
727 # Disable TLS verify. Can be set to true for testing.
730 # When users present tokens issued by this remote cluster, and
731 # their accounts are active on the remote cluster, activate
732 # them on this cluster too.
738 ActivationContactLink: mailto:info@arvados.org
739 ArvadosDocsite: https://doc.arvados.org
740 ArvadosPublicDataDocURL: https://playground.arvados.org/projects/public
741 ShowUserAgreementInline: false
744 # Scratch directory used by the remote repository browsing
745 # feature. If it doesn't exist, it (and any missing parents) will be
746 # created using mkdir_p.
747 RepositoryCache: /var/www/arvados-workbench/current/tmp/git
749 # Below is a sample setting of user_profile_form_fields config parameter.
750 # This configuration parameter should be set to either false (to disable) or
751 # to a map as shown below.
752 # Configure the map of input fields to be displayed in the profile page
753 # using the attribute "key" for each of the input fields.
754 # This sample shows configuration with one required and one optional form fields.
755 # For each of these input fields:
756 # You can specify "Type" as "text" or "select".
757 # List the "Options" to be displayed for each of the "select" menu.
758 # Set "Required" as "true" for any of these fields to make them required.
759 # If any of the required fields are missing in the user's profile, the user will be
760 # redirected to the profile page before they can access any Workbench features.
761 UserProfileFormFields:
764 FormFieldTitle: Best color
765 FormFieldDescription: your favorite color
774 # exampleTextValue: # key that will be set in properties
777 # FormFieldDescription: ""
780 # exampleOptionsValue:
783 # FormFieldDescription: ""
791 # Use "UserProfileFormMessage to configure the message you want
792 # to display on the profile page.
793 UserProfileFormMessage: 'Welcome to Arvados. All <span style="color:red">required fields</span> must be completed before you can proceed.'
795 # Mimetypes of applications for which the view icon
796 # would be enabled in a collection's show page.
797 # It is sufficient to list only applications here.
798 # No need to list text and image types.
799 ApplicationMimetypesWithViewIcon:
817 # The maximum number of bytes to load in the log viewer
818 LogViewerMaxBytes: 1M
820 # When anonymous_user_token is configured, show public projects page
821 EnablePublicProjectsPage: true
823 # By default, disable the "Getting Started" popup which is specific to Arvados playground
824 EnableGettingStartedPopup: false
826 # Ask Arvados API server to compress its response payloads.
827 APIResponseCompression: true
829 # Timeouts for API requests.
830 APIClientConnectTimeout: 2m
831 APIClientReceiveTimeout: 5m
833 # Maximum number of historic log records of a running job to fetch
834 # and display in the Log tab, while subscribing to web sockets.
835 RunningJobLogRecordsToFetch: 2000
837 # In systems with many shared projects, loading of dashboard and topnav
838 # cab be slow due to collections indexing; use the following parameters
839 # to suppress these properties
840 ShowRecentCollectionsOnDashboard: true
841 ShowUserNotifications: true
843 # Enable/disable "multi-site search" in top nav ("true"/"false"), or
844 # a link to the multi-site search page on a "home" Workbench site.
847 # https://workbench.qr1hi.arvadosapi.com/collections/multisite
850 # Should workbench allow management of local git repositories? Set to false if
851 # the jobs api is disabled and there are no local git repositories.
854 SiteName: Arvados Workbench
855 ProfilingEnabled: false
857 # This is related to obsolete Google OpenID 1.0 login
858 # but some workbench stuff still expects it to be set.
859 DefaultOpenIdPrefix: "https://www.google.com/accounts/o8/id"
863 FileViewersConfigURL: ""
865 # Use experimental controller code (see https://dev.arvados.org/issues/14287)
866 EnableBetaController14287: false