mirror of
https://github.com/galaxyproject/galaxy.git
synced 2026-10-16 15:45:04 +08:00
Nate Coraor
08f66a924b
1. Create a new API endpoint and client spin to wait for GIE containers to become ready. IEs are not required to use it, but usage does require a slight change to their `load_notebook()` functions. 2. Add a Docker swarm manager daemon that runs automatically to add nodes when services are waiting, clean up exited services, and remove idle nodes. Shuts down automatically when there's nothing to do. The format of the GIE proxy session database has changed, if GIEs are running when you restart Galaxy, they will be interrupted. If you run the GIE proxy separate from Galaxy, you should restart it after upgrading.
85 lines
3.5 KiB
Plaintext
85 lines
3.5 KiB
Plaintext
---
|
|
# Galaxy docker swarm manager configuration file
|
|
#
|
|
# To configure the location of this file, use the `swarm_manager_config_file`
|
|
# setting in galaxy.ini
|
|
|
|
# When the swarm manager daemonizes, it writes a pid file so that only one
|
|
# manager will run at a time. This is the path to that pid file.
|
|
# {xdg_data_home} will be templated automatically and defaults to
|
|
# ~/.local/share as per the XDG specification
|
|
#pid_file: '{xdg_data_home}/galaxy_swarm_manager.pid'
|
|
|
|
# Program output will be written to the log
|
|
#log_file: '{xdg_data_home}/galaxy_swarm_manager.log'
|
|
|
|
# As with GIE plugins, you can modify the base docker command ({docker_args}
|
|
# must be present and will be filled in with the docker subcommand and
|
|
# arguments)
|
|
#command: 'docker {docker_args}'
|
|
|
|
# Managed services should be started with this string at the beginning of their
|
|
# name. It should match the value of CONTAINER_NAME_PREFIX in
|
|
# lib/galaxy/web/base/interactive_environments.py, so you should not change
|
|
# this unless you change both.
|
|
#service_prefix: galaxy_gie_
|
|
|
|
# Limits:
|
|
#
|
|
# - max_waiting_services: number of services that should be waiting of each
|
|
# "CPU class" (number of CPUs requested e.g. with --reserve-cpu) before
|
|
# attempting to spawn a node
|
|
# - max_wait_time: number of seconds a service should be waiting before
|
|
# attempting to spawn a node
|
|
# - max_node_idle_time: number of seconds a node should be idle before
|
|
# terminating it
|
|
# - max_node_counts: a dictionary controlling the maximum number of nodes of
|
|
# each CPU class that the swarm manager will attempt to spawn, e.g.:
|
|
# max_node_counts:
|
|
# 1: 10 # spawn up to 10 x 1-CPU nodes
|
|
# 2: 3 # spawn up to 3 x 2-CPU nodes
|
|
# 4: 1 # spawn up to 1 x 4-CPU nodes
|
|
#max_waiting_services: 0
|
|
#max_wait_time: 5
|
|
#max_node_idle_time: 120
|
|
#max_node_counts: {}
|
|
|
|
# If set, only manage nodes whose swarm hostnames begin with this prefix.
|
|
# Otherwise, attempt to manage all nodes
|
|
#node_prefix: null
|
|
|
|
# Amount of time to wait for a spawning node to appear in `docker node ls`
|
|
# before considering it failed
|
|
#spawn_wait_time: 30
|
|
|
|
# Command to run to spawn new nodes. This command should join the node to the
|
|
# swarm. Can include template variables:
|
|
# - {cpu_class}: CPU class as explained above (the value of --reserve-cpu)
|
|
# - {cpus_needed}: Total number of CPUs of the given class needed to run the
|
|
# waiting services
|
|
# If this command does not block until the node is joined to the swarm, make
|
|
# sure it at least completes that step in `spawn_wait_time` once it returns
|
|
# control. This command should return a space-separated list of nodes. If the
|
|
# nodes have a different number of CPUs than the class that they were started
|
|
# for, you can include that number after a colon (e.g. `node1:4`).
|
|
#spawn_command: /bin/true
|
|
|
|
# Command to run to destroy idle nodes. Can include template variables:
|
|
# - {nodes}: Space-separated list of node names to destroy
|
|
# This command should block until at least the point at which any nodes being
|
|
# deallocated no longer appear in `docker node ls`.
|
|
#destroy_command: /bin/true
|
|
|
|
# Command to run if either of the above commands failed (e.g. to notify an
|
|
# administrator). Can include template variables:
|
|
# - {failed_command}: Command line of the command that failed
|
|
#command_failure_command: /bin/true
|
|
|
|
# Number of times to retry spawn/destroy commands before considering them to
|
|
# have failed, and seconds to wait between retries
|
|
#command_retries: 0
|
|
#command_retry_wait: 10
|
|
|
|
# Stop the swarm manager daemon when there are no services or nodes to manage
|
|
#terminate_when_idle: True
|