Project

General

Profile

Migrating from arvados-node-manager to arvados-dispatch-cloud » History » Version 1

Tom Clegg, 02/07/2019 08:07 PM

1 1 Tom Clegg
h1. Migrating from arvados-node-manager to crunch-dispatch-cloud
2
3
{{toc}}
4
5
h2. Choose a node
6
7
The dispatch service can run on any host that can connect to the Arvados API service, the cloud provider's API, and the SSH service on cloud VMs. In the following example it runs on the same node as the API server and controller.
8
9
h2. Update cluster configuration file
10
11
In @/etc/arvados/config.yml@, add configuration items for the dispatch service.
12
13
<pre><code class="yaml">
14
Clusters:
15
  uuid_prefix:
16
    CloudVMs:
17
      BootProbeCommand: "mount | grep /mnt/scratch"
18
      SSHPort: "2222"
19
      SyncInterval: 1m
20
      TimeoutIdle: 2m
21
      TimeoutBooting: 10m
22
      TimeoutProbe: 5m
23
      TimeoutShutdown: 30s
24
      ImageID: "image-12345678"
25
      Driver: Azure
26
      DriverParameters:
27
        ...
28
    Dispatch:
29
      PrivateKey: "..."
30
      StaleLockTimeout: 1m
31
      PollInterval: 10s
32
      ProbeInterval: 10s
33
      MaxProbesPerSecond: 10
34
    InstanceTypes:
35
      x1lg:
36
        ProviderType: x1.large
37
        VCPUs: 16
38
        RAM: 128G
39
        Scratch: 128G
40
        Price: 1.23
41
    ManagementToken: "example-secret-management-token"
42
    NodeProfiles:
43
      apiserver:                       # references ARVADOS_NODE_PROFILE in environment file (see below).
44
        arvados-dispatch-cloud:
45
          Listen: ":9005"
46
</code></pre>
47
48
Create the host configuration file @/etc/arvados/environment@.
49
50
<pre>
51
ARVADOS_NODE_PROFILE=apiserver
52
</pre>
53
54
h2. Stop crunch-dispatch-slurm
55
56
Stop and disable the crunch-dispatch-slurm service, and uninstall the package to make sure it doesn't start after the next reboot/upgrade.
57
58
<pre>
59
# systemctl stop crunch-dispatch-slurm
60
# systemctl disable crunch-dispatch-slurm
61
# apt-get remove crunch-dispatch-slurm
62
</pre>
63
64
Containers that have already been locked and submitted to SLURM will make their way through the SLURM queue, but newly queued containers will be left for crunch-dispatch-cloud to run.
65
66
h2. Install crunch-dispatch-cloud
67
68
<pre>
69
# apt-get install crunch-dispatch-cloud
70
</pre>
71
72
h2. Verify the service is running
73
74
<pre>
75
$ token="example-secret-management-token"
76
$ curl -H "Authorization: Bearer $token" http://localhost:9005/metrics
77
</pre>
78
79
h2. Verify the service is functional