PSTORAGE−DEPLOYMENT

NAME
SYNOPSIS
DESCRIPTION
OPERATIONS
INPUT PARAMETERS
OUTPUT PARAMETERS
OPTIONS
ROLES AND ABBREVIATIONS
AUTHOR
SEE ALSO

NAME

pstorage-deployment-wizard − Parallels Cloud Storage deployment wizard

SYNOPSIS

pstorage−deployment−wizard [options] [root[:pass1]@]hostname1[:port1] [[root[:pass1]@]hostname1[:port1]...]

DESCRIPTION

Pstorage−deployment−wizard is a tool for scanning hardware resources and deploying all the repositories and packages needed to run Parallels Cloud Storage on a set of Linux servers. The wizard scans all provided hosts and finds all suitable hard disks and SSD disks, optionally benchmarks them and assigns one or more roles in the cluster to each disk. Only unformatted disks or disk partitions, or disk partitions with ext4 file system and with enough free space (see −m option) can be used to deploy roles. So all other locations are skipped by the cluster installer.

By default, when launched without specific action, the wizard starts in interactive mode with curses−based UI. In this mode it is possible to review and adjust cluster configuration, and deploy new cluster according to this configuration. Alternatively, it is possible to launch wizard in non−interactive mode and deploy cluster according to configuration stored in JSON format.

pstorage−deployment−wizard uses ssh to connect to each host in the list. It is recommended to use key−based authentication and do not provide hosts' passwords using command line or configuration file.

SSD disks connected through RAID controllers often cannot be identified as SSD by the Operating System. Due to this, it is recommended to perform benchmarking stage in pstorage−deployment−wizard. This allows wizard to determine bandwidth and IOPS limit of each disk and better choose cluster roles for each disk according to this information.

OPERATIONS

−a, −−advise

Recommend Parallels Cloud Storage configuration for a given set of hardware.

−i, −−install

Install Parallels Cloud Storage packages on a given set of hardware.

−b, −−ro−benchmark

Perform read−only performance benchmark of disks on a given set of hardware. This type of benchmarking is safe and will not destroy any data on the disks

−d, −−deploy

Deploy Parallels Cloud Storage according to a given configuration file.

−−remove

Remove Parallels Cloud Storage deployed according to a given configuration file. This cluster must have been deployed using pstorage−deployment−wizard before.

INPUT PARAMETERS

−H HOSTS_FILE, −−hosts−file=HOSTS_FILE

Get hosts' parameters from a file, instead of command line arguments. The file must contain hosts list in the following format: [root[:pass1]@]hostname1[:port1] [root[:pass2]@]hostname2[:port2]

−J JSON_FILE, −−json−input=JSON_FILE

Do not perform host scanning but instead take all the informantion from the JSON_FILE file.

OUTPUT PARAMETERS

With either −j or −t option provided pstorage−deployment−wizard switches to non−interactive mode, automatically scans all provided hosts and output hardware resources configuration according to provided option in JSON format or plain text format. If −a option is provided, pstorage−deployment−wizard automatically assigns cluster roles to each disk on the hosts.

−j JSON_FILE, −−json−output=JSON_FILE

Output hardware resources configuration in JSON format into JSON_FILE.

−t, −−text−output

Output hardware resources configuration in plain text format onto stdout.

−r REPLICAS, −−replicas=REPLICAS

Parallels Cloud Storage cluster replicas count.

−m SIZE, −−min−free−space=SIZE

Do not include disks, partitions and mount points with less than SIZE gigabytes free space into cluster configuration.

OPTIONS

−h, −−help

Show help message.

−v, −−verbose

Verbose mode.

ROLES AND ABBREVIATIONS

MDS

MetaData Server. MDS holds cluster metadata and must be located on fast drives (preferably SSD drives). Normally each cluster has 1−5 metadata servers. Every MDS must be located on separate host.

CS

Chunk Server. CS manages all cluster data accessible by cluster clients. Normally every disk not allocated for MDS or CSJ can have a CS.

CSJ

Chunk Server Journal. This journal is needed to increase write and flush performance of the Chunk Servers located on the same host. Single CSJ role can handle journals of several CS’s from the host. CSJ must be located on a fast drive (preferably SSD).

Phys Space

Physical Space used by all Chunk Servers in the cluster. Logical cluster space seen by all clients will be much smaller due to data replication.

AUTHOR

Copyright © 2011−2015, Parallels, Inc. All rights reserved.

SEE ALSO

pstorage(1), pstorage−overview(7), pstorage−ssd(7)