Disk reservation

From Grid5000
Revision as of 13:56, 7 May 2017 by Pneyron (talk | contribs) (rephrase a bit of intro)
Jump to navigation Jump to search

Disk reservation is a technique to reserve nodes' hard disks, in order to locally store large datasets between your reservations, and avoid needing to move data to nodes at the beginning of each node reservation.

It takes benefit from nodes which feature a bunch of disks.

Note.png Note

During the beta testing phase, disks reservation will only be available on the grimoire cluster. The service will be extended to other clusters later.

The table below shows the cluster compatibility with disk reservation.

Site Cluster Number of nodes Number of reservable disks per node
Nancy grimoire 8 5
Warning.png Warning

Disk reservation is a service for experimentation: data integrity is not guaranteed. You MUST backup your important data.

How it works

Disk reservation works with jobs of type noop and jobs of type deploy. You can reserve some disks in the long run with a job of type noop. Then, you may reserve a job of type deploy on the same nodes as those where you reserved your disks.

When the job of type deploy starts, the disks you reserved are enabled by the RAID card of the node, and the other disks are disabled. Reserved disks can only be accessed by the user who reserved them (see also Security issues).

Usage

The main commands to reserve disks are given below. Note that disk reservation only works with a node reservation of type deploy. By default, reserving a node on a cluster with disk reservation enabled will only grant access to the main disk (additional disks need to be reserved).

The maximum duration of a disk reservation is defined in the Usage Policy.

Reserve disks and nodes at the same time

Reserve a node with only the main disk (none of the additional disks), on the grimoire cluster
Terminal.png fnancy:
oarsub -I -t deploy -l {"cluster='grimoire'"}/host=1

(no change to the way a node was to be reserved in the past, before the disk reservation mechanism existed)

Reserve a node with all its disks
Terminal.png fnancy:
oarsub -I -t deploy -l {"type='disk' or type='default' and disk_reservation_count > 0"}/host=1

disk_reservation_count is equal to the number of disks which can be reserved on a node.

Reserve nodes grimoire-1 and grimoire-2 with one reservable disk per node
Terminal.png fnancy:
oarsub -I -t deploy -l {"type='default' and host in ('grimoire-1.nancy.grid5000.fr','grimoire-2.nancy.grid5000.fr')"}/host=2+{"type='disk' and host in ('grimoire-1.nancy.grid5000.fr','grimoire-2.nancy.grid5000.fr')"}/host=2/disk=1

or, same but using environment variables:

Terminal.png fnancy:
HOSTNAMES="'grimoire-1.nancy.grid5000.fr','grimoire-2.nancy.grid5000.fr'"
Terminal.png fnancy:
HOSTCOUNT=2
Terminal.png fnancy:
oarsub -I -t deploy -l {"type='default' and host in ($HOSTNAMES)"}/host=$HOSTCOUNT+{"type='disk' and host in ($HOSTNAMES)"}/host=$HOSTCOUNT/disk=1

Reserve disks and nodes separately

First : reserve the disks

For example, you may decide to reserve some disks for one week, and the nodes where your disks are located only when you want to carry out an experiment. Jobs of type noop are used to reserve disks. They cannot be interactive and are necessarily reserved in advance.

Reserve two disks on grimoire-1 for one week, starting on January 1st 2018:

Terminal.png fnancy:
oarsub -r "2018-01-01 00:00:00" -t noop -l {"type='disk' and host='grimoire-1.nancy.grid5000.fr'"}/host=1/disk=2,walltime=168

Reserve the first two disks on grimoire-2:

Terminal.png fnancy:
oarsub -r "2018-01-01 00:00:00" -t noop -l {"type='disk' and host='grimoire-2.nancy.grid5000.fr' and disk in ('1', '2')"}/host=1/disk=2,walltime=168

Or reserve all disks on two nodes:

Terminal.png fnancy:
oarsub -r "2018-01-01 00:00:00" -t noop -l {"type='disk' and cluster='grimoire'"}/host=2/disk=ALL,walltime=168

Secondly: reserve the nodes

You can reserve nodes grimoire-1 and grimoire-2 for 3 hours, in the usual way:

Terminal.png fnancy:
oarsub -I -t deploy -l {"host in ('grimoire-1.nancy.grid5000.fr', 'grimoire-2.nancy.grid5000.fr')"}/host=2,walltime=3

You must respect this order : reserve the disks first, then reserve the nodes. Otherwise the disks you reserved will not be available on your nodes.

Security issues

The mechanism used to enable/disable disks is designed to avoid mistakes from other users. However, a malicious user could take control of the RAID card, enable all disks, and access your data. If you have sensitive data, you should use cryptographic mechanisms to secure it (contact the Grid'5000 team if you need help).

Also, the data on reserved disks is not automatically erased at the end of your job. If you don't want the next user to access it, you should erase it yourself.