1 # Copyright (C) The Arvados Authors. All rights reserved.
3 # SPDX-License-Identifier: AGPL-3.0
5 # Google Compute Engine configuration for Arvados Node Manager.
6 # All times are in seconds unless specified otherwise.
9 # The management server responds to http://addr:port/status.json with
10 # a snapshot of internal state.
12 # Management server listening address (default 127.0.0.1)
15 # Management server port number (default -1, server is disabled)
19 # Node Manager will ensure that there are at least this many nodes running at
20 # all times. If node manager needs to start new idle nodes for the purpose of
21 # satisfying min_nodes, it will use the cheapest node type. However, depending
22 # on usage patterns, it may also satisfy min_nodes by keeping alive some
23 # more-expensive nodes
26 # Node Manager will not start any compute nodes when at least this
27 # running at all times. By default, these will be the cheapest node size.
30 # Poll compute nodes and Arvados for new information every N seconds.
33 # Upper limit on rate of spending (in $/hr), will not boot additional nodes
34 # if total price of already running nodes meets or exceeds this threshold.
35 # default 0 means no limit.
38 # Polls have exponential backoff when services fail to respond.
39 # This is the longest time to wait between polls.
42 # If Node Manager can't succesfully poll a service for this long,
43 # it will never start or stop compute nodes, on the assumption that its
44 # information is too outdated.
45 poll_stale_after = 600
47 # "Node stale time" affects two related behaviors.
48 # 1. If a compute node has been running for at least this long, but it
49 # isn't paired with an Arvados node, do not shut it down, but leave it alone.
50 # This prevents the node manager from shutting down a node that might
51 # actually be doing work, but is having temporary trouble contacting the
53 # 2. When the Node Manager starts a new compute node, it will try to reuse
54 # an Arvados node that hasn't been updated for this long.
55 node_stale_after = 14400
57 # Number of consecutive times a node must report as "idle" before it
58 # will be considered eligible for shutdown. Node status is checked
59 # each poll period, and node can go idle at any point during a poll
60 # period (meaning a node could be reported as idle that has only been
61 # idle for 1 second). With a 60 second poll period, three consecutive
62 # status updates of "idle" suggests the node has been idle at least
64 consecutive_idle_count = 3
66 # Scaling factor to be applied to nodes' available RAM size. Usually there's a
67 # variable discrepancy between the advertised RAM value on cloud nodes and the
68 # actual amount available.
69 # If not set, this value will be set to 0.95
70 node_mem_scaling = 0.95
72 # File path for Certificate Authorities
73 certs_file = /etc/ssl/certs/ca-certificates.crt
77 file = /var/log/arvados/node-manager.log
79 # Log level for most Node Manager messages.
80 # Choose one of DEBUG, INFO, WARNING, ERROR, or CRITICAL.
81 # WARNING lets you know when polling a service fails.
82 # INFO additionally lets you know when a compute node is started or stopped.
85 # You can also set different log levels for specific libraries.
86 # Pykka is the Node Manager's actor library.
87 # Setting this to DEBUG will display tracebacks for uncaught
88 # exceptions in the actors, but it's also very chatty.
91 # Setting apiclient to INFO will log the URL of every Arvados API request.
95 host = zyxwv.arvadosapi.com
98 jobs_queue = yes # Get work request from Arvados jobs queue (jobs API)
99 slurm_queue = yes # Get work request from squeue (containers API)
101 # Accept an untrusted SSL certificate from the API server?
107 # Shutdown windows define periods of time when a node may and may not
108 # be shut down. These are windows in full minutes, separated by
109 # commas. Counting from the time the node is booted, the node WILL
110 # NOT shut down for N1 minutes; then it MAY shut down for N2 minutes;
111 # then it WILL NOT shut down for N3 minutes; and so on. For example,
112 # "54, 5, 1" means the node may shut down from the 54th to the 59th
113 # minute of each hour of uptime.
114 # GCE bills by the minute, and does not provide information about when
115 # a node booted. Node Manager will store this information in metadata
116 # when it boots a node; if that information is not available, it will
117 # assume the node booted at the epoch. These shutdown settings are
118 # very aggressive. You may want to adjust this if you want more
119 # continuity of service from a single node.
120 shutdown_windows = 20, 999999
123 user_id = client_email_address@developer.gserviceaccount.com
124 key = path_to_certificate.pem
125 project = project-id-from-google-cloud-dashboard
128 # Valid location (zone) names: https://cloud.google.com/compute/docs/zones
129 datacenter = us-central1-a
131 # Optional settings. For full documentation see
132 # http://libcloud.readthedocs.org/en/latest/compute/drivers/gce.html#libcloud.compute.drivers.gce.GCENodeDriver
134 # auth_type = SA # SA, IA or GCE
135 # scopes = https://www.googleapis.com/auth/compute
139 # A comma-separated list of tags that must be applied to a node for it to
140 # be considered a compute node.
141 # The driver will automatically apply these tags to nodes it creates.
142 tags = zyxwv, compute
145 # New compute nodes will send pings to Arvados at this host.
146 # You may specify a port, and use brackets to disambiguate IPv6 addresses.
147 ping_host = hostname:port
149 # A file path for an SSH key that can log in to the compute node.
152 # The GCE image name and network zone name to use when creating new nodes.
154 # network = your_network_name
156 # JSON string of service account authorizations for this cluster.
157 # See http://libcloud.readthedocs.org/en/latest/compute/drivers/gce.html#specifying-service-account-scopes
158 # service_accounts = [{'email':'account@example.com', 'scopes':['storage-ro']}]
161 # You can define any number of Size sections to list node sizes you're
162 # willing to use. The Node Manager should boot the cheapest size(s) that
163 # can run jobs in the queue.
165 # The Size fields are interpreted the same way as with a libcloud NodeSize:
166 # http://libcloud.readthedocs.org/en/latest/compute/api.html#libcloud.compute.base.NodeSize
168 # See https://cloud.google.com/compute/docs/machine-types for a list
169 # of known machine types that may be used as a Size parameter.
171 # Each size section MUST define the number of cores are available in this
172 # size class (since libcloud does not provide any consistent API for exposing
174 # You may also want to define the amount of scratch space (expressed
175 # in GB) for Crunch jobs.
176 # You can also override Google's provided data fields (such as price per hour)
177 # by setting them here.