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"
67 # max concurrent connections per arvados server daemon
70 # All parameters here are passed to the PG client library in a connection string;
71 # see https://www.postgresql.org/docs/current/static/libpq-connect.html#LIBPQ-PARAMKEYWORDS
78 # Maximum size (in bytes) allowed for a single API request. This
79 # limit is published in the discovery document for use by clients.
80 # Note: You must separately configure the upstream web server or
81 # proxy to actually enforce the desired maximum request size on the
83 MaxRequestSize: 134217728
85 # Limit the number of bytes read from the database during an index
86 # request (by retrieving and returning fewer rows than would
87 # normally be returned in a single response).
88 # Note 1: This setting never reduces the number of returned rows to
89 # zero, no matter how big the first data row is.
90 # Note 2: Currently, this is only checked against a specific set of
91 # columns that tend to get large (collections.manifest_text,
92 # containers.mounts, workflows.definition). Other fields (e.g.,
93 # "properties" hashes) are not counted against this limit.
94 MaxIndexDatabaseRead: 134217728
96 # Maximum number of items to return when responding to a APIs that
97 # can return partial result sets using limit and offset parameters
98 # (e.g., *.index, groups.contents). If a request specifies a "limit"
99 # parameter higher than this value, this value is used instead.
100 MaxItemsPerResponse: 1000
102 # API methods to disable. Disabled methods are not listed in the
103 # discovery document, and respond 404 to all requests.
104 # Example: ["jobs.create", "pipeline_instances.create"]
107 # Interval (seconds) between asynchronous permission view updates. Any
108 # permission-updating API called with the 'async' parameter schedules a an
109 # update on the permission view in the future, if not already scheduled.
110 AsyncPermissionsUpdateInterval: 20
113 # Config parameters to automatically setup new users. If enabled,
114 # this users will be able to self-activate. Enable this if you want
115 # to run an open instance where anyone can create an account and use
116 # the system without requiring manual approval.
118 # The params auto_setup_new_users_with_* are meaningful only when auto_setup_new_users is turned on.
119 # auto_setup_name_blacklist is a list of usernames to be blacklisted for auto setup.
120 AutoSetupNewUsers: false
121 AutoSetupNewUsersWithVmUUID: ""
122 AutoSetupNewUsersWithRepository: false
123 AutoSetupUsernameBlacklist: [arvados, git, gitolite, gitolite-admin, root, syslog]
125 # When new_users_are_active is set to true, new users will be active
126 # immediately. This skips the "self-activate" step which enforces
127 # user agreements. Should only be enabled for development.
128 NewUsersAreActive: false
130 # The e-mail address of the user you would like to become marked as an admin
131 # user on their first login.
132 # In the default configuration, authentication happens through the Arvados SSO
133 # server, which uses OAuth2 against Google's servers, so in that case this
134 # should be an address associated with a Google account.
135 AutoAdminUserWithEmail: ""
137 # If auto_admin_first_user is set to true, the first user to log in when no
138 # other admin users exist will automatically become an admin user.
139 AutoAdminFirstUser: false
141 # Email address to notify whenever a user creates a profile for the
143 UserProfileNotificationAddress: ""
144 AdminNotifierEmailFrom: arvados@example.com
145 EmailSubjectPrefix: "[ARVADOS] "
146 UserNotifierEmailFrom: arvados@example.com
147 NewUserNotificationRecipients: []
148 NewInactiveUserNotificationRecipients: []
151 # Time to keep audit logs, in seconds. (An audit log is a row added
152 # to the "logs" table in the PostgreSQL database each time an
153 # Arvados object is created, modified, or deleted.)
155 # Currently, websocket event notifications rely on audit logs, so
156 # this should not be set lower than 600 (5 minutes).
159 # Maximum number of log rows to delete in a single SQL transaction.
161 # If max_audit_log_delete_batch is 0, log entries will never be
162 # deleted by Arvados. Cleanup can be done by an external process
163 # without affecting any Arvados system processes, as long as very
164 # recent (<5 minutes old) logs are not deleted.
166 # 100000 is a reasonable batch size for most sites.
169 # Attributes to suppress in events and audit logs. Notably,
170 # specifying ["manifest_text"] here typically makes the database
171 # smaller and faster.
173 # Warning: Using any non-empty value here can have undesirable side
174 # effects for any client or component that relies on event logs.
175 # Use at your own risk.
176 UnloggedAttributes: []
179 # Maximum characters of (JSON-encoded) query parameters to include
180 # in each request log entry. When params exceed this size, they will
181 # be JSON-encoded, truncated to this size, and logged as
183 MaxRequestLogParamsSize: 2000
186 # Allow clients to create collections by providing a manifest with
187 # unsigned data blob locators. IMPORTANT: This effectively disables
188 # access controls for data stored in Keep: a client who knows a hash
189 # can write a manifest that references the hash, pass it to
190 # collections.create (which will create a permission link), use
191 # collections.get to obtain a signature for that data locator, and
192 # use that signed locator to retrieve the data from Keep. Therefore,
193 # do not turn this on if your users expect to keep data private from
197 # blob_signing_key is a string of alphanumeric characters used to
198 # generate permission signatures for Keep locators. It must be
199 # identical to the permission key given to Keep. IMPORTANT: This is
200 # a site secret. It should be at least 50 characters.
202 # Modifying blob_signing_key will invalidate all existing
203 # signatures, which can cause programs to fail (e.g., arv-put,
204 # arv-get, and Crunch jobs). To avoid errors, rotate keys only when
205 # no such processes are running.
208 # Default replication level for collections. This is used when a
209 # collection's replication_desired attribute is nil.
210 DefaultReplication: 2
212 # Lifetime (in seconds) of blob permission signatures generated by
213 # the API server. This determines how long a client can take (after
214 # retrieving a collection record) to retrieve the collection data
215 # from Keep. If the client needs more time than that (assuming the
216 # collection still has the same content and the relevant user/token
217 # still has permission) the client can retrieve the collection again
218 # to get fresh signatures.
220 # This must be exactly equal to the -blob-signature-ttl flag used by
221 # keepstore servers. Otherwise, reading data blocks and saving
222 # collections will fail with HTTP 403 permission errors.
224 # Modifying blob_signature_ttl invalidates existing signatures; see
225 # blob_signing_key note above.
227 # The default is 2 weeks.
228 BlobSigningTTL: 1209600
230 # Default lifetime for ephemeral collections: 2 weeks. This must not
231 # be less than blob_signature_ttl.
232 DefaultTrashLifetime: 1209600
234 # Interval (seconds) between trash sweeps. During a trash sweep,
235 # collections are marked as trash if their trash_at time has
236 # arrived, and deleted if their delete_at time has arrived.
237 TrashSweepInterval: 60
239 # If true, enable collection versioning.
240 # When a collection's preserve_version field is true or the current version
241 # is older than the amount of seconds defined on preserve_version_if_idle,
242 # a snapshot of the collection's previous state is created and linked to
243 # the current collection.
244 CollectionVersioning: false
246 # 0 = auto-create a new version on every update.
247 # -1 = never auto-create new versions.
248 # > 0 = auto-create a new version when older than the specified number of seconds.
249 PreserveVersionIfIdle: -1
252 # These settings are provided by your OAuth2 provider (e.g.,
254 ProviderAppSecret: ""
258 # Git repositories must be readable by api server, or you won't be
259 # able to submit crunch jobs. To pass the test suites, put a clone
260 # of the arvados tree in {git_repositories_dir}/arvados.git or
261 # {git_repositories_dir}/arvados/.git
262 Repositories: /var/lib/arvados/git/repositories
268 # List of supported Docker Registry image formats that compute nodes
269 # are able to use. `arv keep docker` will error out if a user tries
270 # to store an image with an unsupported format. Use an empty array
271 # to skip the compatibility check (and display a warning message to
274 # Example for sites running docker < 1.10: ["v1"]
275 # Example for sites running docker >= 1.10: ["v2"]
276 # Example for disabling check: []
277 SupportedDockerImageFormats: ["v2"]
279 # Include details about job reuse decisions in the server log. This
280 # causes additional database queries to run, so it should not be
281 # enabled unless you expect to examine the resulting logs for
282 # troubleshooting purposes.
283 LogReuseDecisions: false
285 # Default value for keep_cache_ram of a container's runtime_constraints.
286 DefaultKeepCacheRAM: 268435456
288 # Number of times a container can be unlocked before being
289 # automatically cancelled.
290 MaxDispatchAttempts: 5
292 # Default value for container_count_max for container requests. This is the
293 # number of times Arvados will create a new container to satisfy a container
294 # request. If a container is cancelled it will retry a new container if
295 # container_count < container_count_max on any container requests associated
296 # with the cancelled container.
299 # The maximum number of compute nodes that can be in use simultaneously
300 # If this limit is reduced, any existing nodes with slot number >= new limit
301 # will not be counted against the new limit. In other words, the new limit
302 # won't be strictly enforced until those nodes with higher slot numbers
306 # Preemptible instance support (e.g. AWS Spot Instances)
307 # When true, child containers will get created with the preemptible
308 # scheduling parameter parameter set.
309 UsePreemptibleInstances: false
311 # Include details about job reuse decisions in the server log. This
312 # causes additional database queries to run, so it should not be
313 # enabled unless you expect to examine the resulting logs for
314 # troubleshooting purposes.
315 LogReuseDecisions: false
318 # When you run the db:delete_old_container_logs task, it will find
319 # containers that have been finished for at least this many seconds,
320 # and delete their stdout, stderr, arv-mount, crunch-run, and
321 # crunchstat logs from the logs table.
324 # These two settings control how frequently log events are flushed to the
325 # database. Log lines are buffered until either crunch_log_bytes_per_event
326 # has been reached or crunch_log_seconds_between_events has elapsed since
328 LogBytesPerEvent: 4096
329 LogSecondsBetweenEvents: 1
331 # The sample period for throttling logs, in seconds.
332 LogThrottlePeriod: 60
334 # Maximum number of bytes that job can log over crunch_log_throttle_period
335 # before being silenced until the end of the period.
336 LogThrottleBytes: 65536
338 # Maximum number of lines that job can log over crunch_log_throttle_period
339 # before being silenced until the end of the period.
340 LogThrottleLines: 1024
342 # Maximum bytes that may be logged by a single job. Log bytes that are
343 # silenced by throttling are not counted against this total.
344 LimitLogBytesPerJob: 67108864
346 LogPartialLineThrottlePeriod: 5
348 # Container logs are written to Keep and saved in a collection,
349 # which is updated periodically while the container runs. This
350 # value sets the interval (given in seconds) between collection
352 LogUpdatePeriod: 1800
354 # The log collection is also updated when the specified amount of
355 # log data (given in bytes) is produced in less than one update
357 LogUpdateSize: 33554432
361 # Path to dns server configuration directory
362 # (e.g. /etc/unbound.d/conf.d). If false, do not write any config
363 # files or touch restart.txt (see below).
366 # Template file for the dns server host snippets. See
367 # unbound.template in this directory for an example. If false, do
368 # not write any config files.
369 DNSServerConfTemplate: ""
371 # String to write to {dns_server_conf_dir}/restart.txt (with a
372 # trailing newline) after updating local data. If false, do not
373 # open or write the restart.txt file.
374 DNSServerReloadCommand: ""
376 # Command to run after each DNS update. Template variables will be
377 # substituted; see the "unbound" example below. If false, do not run
379 DNSServerUpdateCommand: ""
381 ComputeNodeDomain: ""
382 ComputeNodeNameservers:
385 # Hostname to assign to a compute node when it sends a "ping" and the
386 # hostname in its Node record is nil.
387 # During bootstrapping, the "ping" script is expected to notice the
388 # hostname given in the ping response, and update its unix hostname
390 # If false, leave the hostname alone (this is appropriate if your compute
391 # nodes' hostnames are already assigned by some other mechanism).
393 # One way or another, the hostnames of your node records should agree
394 # with your DNS records and your /etc/slurm-llnl/slurm.conf files.
396 # Example for compute0000, compute0001, ....:
397 # assign_node_hostname: compute%<slot_number>04d
398 # (See http://ruby-doc.org/core-2.2.2/Kernel.html#method-i-format for more.)
399 AssignNodeHostname: "compute%<slot_number>d"
402 # Enable the legacy Jobs API. This value must be a string.
403 # 'auto' -- (default) enable the Jobs API only if it has been used before
404 # (i.e., there are job records in the database)
405 # 'true' -- enable the Jobs API despite lack of existing records.
406 # 'false' -- disable the Jobs API despite presence of existing records.
409 # Git repositories must be readable by api server, or you won't be
410 # able to submit crunch jobs. To pass the test suites, put a clone
411 # of the arvados tree in {git_repositories_dir}/arvados.git or
412 # {git_repositories_dir}/arvados/.git
413 GitInternalDir: /var/lib/arvados/internal.git
415 # Docker image to be used when none found in runtime_constraints of a job
416 DefaultDockerImage: ""
418 # none or slurm_immediate
419 CrunchJobWrapper: none
421 # username, or false = do not set uid when running jobs.
422 CrunchJobUser: crunch
424 # The web service must be able to create/write this file, and
425 # crunch-job must be able to stat() it.
426 CrunchRefreshTrigger: /tmp/crunch_refresh_trigger
428 # Control job reuse behavior when two completed jobs match the
429 # search criteria and have different outputs.
431 # If true, in case of a conflict, reuse the earliest job (this is
432 # similar to container reuse behavior).
434 # If false, in case of a conflict, do not reuse any completed job,
435 # but do reuse an already-running job if available (this is the
436 # original job reuse behavior, and is still the default).
437 ReuseJobIfOutputsDiffer: false
442 SendUserSetupNotificationEmail: ""
443 IssueReporterEmailFrom: ""
444 IssueReporterEmailTo: ""
445 SupportEmailAddress: ""