Weka System Installation Process Using the CLI

This page describes the stages in the installation process when using the CLI.

Stage 1: Installation of the Weka Software on Each Host

Run the untar command and install.sh command, according to the instructions, on each host.

On completion of this stage in the installation process, the Weka software is installed on all the allocated hosts and running in the stem mode i.e., no cluster is attached and the Weka system is awaiting instructions.

Note: If a failure occurs during this installation stage, an error message detailing the source of the failure will be received. If possible, try to recover this error or alternatively, contact the Weka Support Team.

Stage 2: Formation of a Cluster from the Hosts

Command: weka cluster create

This stage involves the formation of a cluster from the allocated hosts. It is performed using the following command line:

weka cluster create <hostnames> [--host-ips <ips | ip+ip+ip+ip>]

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

hostnames

Space-separated strings

Hostnames or IP addresses

Need at least 6 strings, as this is the minimal cluster size

Yes

host-ips

Comma-separated IP addresses

IP addresses of the management interfaces. Use a list of ip+ip addresses pairs of two cards for HA configuration. In case the cluster is connected to both IB and Ethernet, it is possible to set up to 4 management IPs for redundancy of both the IB and Ethernet networks using a list of ip+ip+ip+ip addresses.

The same number of values as in hostnames.

No

IP of the first network device of the host

Note: It is possible to use either a host-name or an IP address; this string serves as the identifier of the host in subsequent commands.

Note: If a host-name is used, make sure that the host-name to IP resolution mechanism is reliable since a failure of this mechanism will cause a loss of service in the cluster. It is recommended to add the host-names to /etc/hosts.

Note: After successful completion of this command, the cluster is in the initialization phase, and some commands can only run in this phase.

Note: For configuring HA, at least two cards must be defined for each host.

On successful completion of the formation of the cluster, every host receives a host ID. Use of the command line weka cluster host will display a list of the hosts and IDs.

Note: In IB installations the --hosts-ips parameter must specify the IP addresses of the IPoIB interfaces.

Stage 3: Naming the Cluster (optional)

Command: weka cluster update

This command is used to give the cluster a name. Although this is optional, it is highly recommended, because the name enables cloud event notification and increases the ability of the Weka Support Team to resolve any issues that may occur. To perform this operation, use the following command line:

weka cluster update --cluster-name=<cluster-name>

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

cluster-name

String

Identifier of the cluster name

Must be a valid identifier

No

Stage 4: Enabling Cloud Event Notifications (optional)

Enabling Support via Weka Home

Command: weka cloud enable

This command enables cloud event notification (via Weka Home), which increases the ability of the Weka Support Team to resolve any issues that may occur.

To learn more about this and how to enable cloud event notification, refer to Weka Support Cloud.

For Private Instance of Weka Home

In closed environments, such as dark sites and private VPCs, it is possible to install a private instance of Weka Home.

Command: weka cloud enable --cloud-url=http://<weka-home-ip>:<weka-home-port>

This command enables the use of a private instance of Weka Home.

For more information, refer to Private Instance of Weka Home and contact the Weka Support Team.

Stage 5: Setting hosts as dedicated to the cluster (Optional)

Command: weka cluster host dedicate

It is possible to set the host as dedicated to the Weka cluster. By setting the host to dedicated, no other application is expected to run on it, and the Weka system optimizes it for performance and stability. For example, the host can be rebooted by the system at need, and all the host's memory is allocatable by the Weka processes.

Stage 6: Configuration of Networking

Command: weka cluster host net add

When PKEYs are used, the device name for InfiniBand should follow the name.PKEY convention.

Note: Although in general, devices can be renamed arbitrarily, Weka will only function correctly if the .PKEY naming convention is followed.

The networking type can be either Ethernet (direct over DPDK) or InfiniBand (IB), and can be mixed in the same host (by running multiple cluster host net add commands for the same host). A physical network device must be specified for both types. This can be a device dedicated to the Weka system or a device that is also being used for other purposes in parallel. For IP over DPDK, the standard routing parameters can be specified for routed networks.

To perform this operation, the cluster host net add command must be run for each host. The commands can run from one host configuring another host, so they can all run on a single host. The IP addresses specified using this command are the data plane IPs allocated in the planning stage. To perform this operation, use the following command line:

weka cluster host net add <host-id> <device> [--ips-type=<POOL|USER>] [--ips=<ips>]... [--gateway=<gateway>] [--netmask=<netmask>] [--label=<label>]

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

host-id

String

Identifier of the host to which a network interface will be added

Must be a valid host identifier

Yes

device

String

A device, or bond-device e.g., eth1 or bond0

Must be a valid Unix network device name

Yes

ips-type

String

POOL or USER

Must be one of the two options

No

POOL

ips

Comma-separated IP address

The data plane IP addresses for internal Weka system traffic. In IB, use the IPoIB address

Must be part of the data plane IP pool defined in the planning phase. See Weka Networking and Networking Prerequisites.

No

From Pool

netmask

Number

Number of bits in the netmask

Describes the number of bits that identify a network ID (also known as CIDR). Not relevant for IB / L2 non-routable networks, and must be supplied for the ethernet NICs if the cluster is set to use both ethernet and IB interfaces.

No

gateway

IP address

The IP address of the default routing gateway

The gateway must reside within the same IP network of ips (as described by netmask).

Not relevant for IB / L2 non-routable networks.

No

label

String

A label to describe the network device connectivity.

The Weka system will prefer to use paths with the same labels to send data. This is useful when the system is configured with HA networking, to hint the system to send between hosts through the same switch rather than using the ISL.

No

The number of IP addresses should be according to Weka Networking and Networking Prerequisites.

Note: Additional IP addresses may be assigned for each host if IP per core is needed. In this case, unused IP addresses are reserved for future expansions and can be automatically assigned if the number of cores assigned to the Weka system on that host is increased.

Note: For HA configurations, this command has to be run separately for each interface.

Optional: Configure default data networking

Command: weka cluster default-net set

Instead of explicit IP address configuration per each network device, dynamic IP address allocation is supported. Weka supports adding a range of IP addresses to a dynamic pool, from which the IP addresses can be automatically allocated on demand.

For Ethernet networking only, a mixed approach is supported: for certain network devices the IP addresses are assigned explicitly by the administrator, while the other devices in the cluster get an automatic allocation from the IP range. Such an approach could be useful in an environment where clients are being spawned automatically.

weka cluster default-net set --range <range> [--gateway=<gateway>] [--netmask-bits=<netmask-bits>]

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

range

IP address range

A range of IP addresses that can be used for dynamic allocation across the whole cluster

Format: A.B.C.D-E

E.g., 10.10.0.1-100

Yes

netmask-bits

Number

Number of bits in the netmask

Describes the number of bits that identify a network ID (also known as CIDR).

Yes

gateway

IP address

The IP address of the default routing gateway

The gateway must reside within the same IP network of IPs in range (as described by netmask-bits).

Not relevant for IB / L2 non-routable networks.

No

To view the current default data networking settings use the command weka cluster default-net.

If a default data networking was previously configured on a cluster and is no longer needed, it is possible to remove it using the command weka cluster default-net reset.

Stage 7: Configuration of SSDs

Command: weka cluster drive add

This stage in the installation process is used to add a local SSD to be used by a Weka filesystem. The same command can be used for adding multiple drive paths. To perform this operation, use the following command line:

weka cluster drive add <host-id> <device-paths>

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

host-id

String

Identifier of the host to which a local SSD will be added

Must be a valid host identifier

Yes

device-paths

Space-separated list of strings

List of block devices that identify local SSDs, e.g., /dev/nvme0n1 /dev/nvme1n1

Must be a valid Unix network device name

Yes

Note: If, due to some technical limitation, the use of an NVMe device through the kernel is required, contact the Weka Support Team.

Stage 8: Configuration of CPU Resources

Command: weka cluster host cores

This stage in the installation process is used to configure the number of CPU resources, which are physical rather than logical cores. To perform this operation, use the following command line:

weka cluster host cores <host-id> <cores> [--frontend-dedicated-cores <frontend-dedicated-cores>] [--drives-dedicated-cores <drives-dedicated-cores>] [--cores-ids <cores-ids>]

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

host-id

String

Identifier of the host in which a core count should be configured

Must be a valid host identifier

Yes

cores

Number

Number of physical cores to be allocated to the Weka system

Should be less than the number of physical cores in the host (leaving 1 core for the OS). Maximum 19 cores

Yes

frontend-dedicated-cores

Number

Number of physical cores to be dedicated to FrontEnd processes

The total of fe_cores and be_cores must be less than cores above

No

zero

drives-dedicated-cores

Number

Number of physical cores to be dedicated to Drive/SSD processes

The total of fe_cores and be_cores must be less than cores above

No

Typically 1 core per drive or 1/2 core per drive/SSD

cores-ids

A comma-separated list of numbers

Physical Core numbers

Specification of which cores to use.

No

Select cores automatically

Note: cores-ids are distributed in the following order: first, all the FrontEnd processes, second, all the Compute processes, and last, all the Drive processes. By ordering the cores-ids list, it is possible to determine the exact assignment of cores to processes (e.g., for taking into account NUMA distribution).

For example: If we have 1 FrontEnd, 2 Compute, and 3 Drive, setting cores-ids to 1, 2, 4, 3, 5, 6 will put the FrontEnd on core 1, Compute on cores 2 and 4, and Drive on cores 3, 5 and 6. Assuming cores 1, 2, 3 are at NUMA 0 and cores 4, 5, 6 are at NUMA 1, we will have the following distribution of processes:

  • NUMA 0: FrontEnd, Compute, Drive

  • NUMA 1: Compute, Drive, Drive

Note: Performance can be optimized by assigning different functions to the various Weka cores. If necessary, contact the Weka Support Team for more information.

Stage 9: Configuration of Memory (optional)

Command: weka cluster host memory

As defined in the memory requirements, the fixed memory per host and the per compute/SSD cores memory are automatically calculated by the Weka system. By default, 1.4 GB is allocated per compute-core, out of which 0.4 GB is left for the capacity-oriented memory. If the host is set as dedicated, all the memory left after reductions, as described in Memory Resource Planning, is automatically allocated for the Weka system.

If capacity requirements mandate more memory, the following command should be used:

weka cluster host memory <host-id> <capacity-memory>

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

host-id

String

Identifier of the host in which the memory configuration has to be defined.

Must be a valid host identifier

Yes

capacity-memory

Number

The memory dedicated to Weka in bytes. It is possible to set the format in other units, e.g.: 1MB, 1GB, 1MiB, 1GiB.

Setting to 0 determines this value automatically

Yes

Note: This command is given the memory per-host and will later be distributed by the system per compute core. Out of this value, 1GB per compute core is reserved for other purposes (as cache) and not used for capacity.

Stage 10: Configuration of Failure Domains (optional)

Command: weka cluster host failure-domain

This optional stage in the installation process is used to assign a host to a failure domain. If the specified failure domain does not exist, it will be created by this command. If the host is assigned to another failure domain, it will be reassigned by this command.

Note: All hosts not assigned to any failure domain will be considered by the Weka system as an additional failure domain. However, it is good practice to either not define failure domains at all or to assign each host to a single failure domain.

This operation is performed using the following command line:

weka cluster host failure-domain <host-id> [--name <name>] | [--auto]

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

host-id

String

Identifier of the host in which the failure domain should be configured.

Must be a valid host identifier

Yes

name

String

The failure domain that will contain the host from now.

Yes (either --name OR --auto must be specified)

auto

Boolean

Will automatically assign a failure domain name.

Yes (either --name OR --auto must be specified)

Stage 11: Configuration of Weka System Protection Scheme (optional)

Command: weka cluster update

To configure the Weka system protection scheme, use the following command line:

weka cluster update [--data-drives=<data-drives>] [--parity-drives=<parity-drives>]

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

data-drives

Number

Protection stripe width

Between 3-16. The stripe width + the protection level cannot exceed the number of failure domains.

No

#failure domains - protection level; no more than 16

parity-drives

Number

Protection level

Either 2 or 4. The stripe width + the protection level cannot exceed the number of failure domains.

No

2

Note: This command can only be used in the initialization phase.

Stage 12: Configuration of Hot Spare (optional)

Command: weka cluster hot-spare

To configure the Weka system hot spare, use the following command line:

weka cluster hot-spare <count>

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

count

Number

Hot spare

The number of failure domains cannot be smaller than the stripe width + the protection level + hot spare

No

1

Stage 13: Applying Hosts Configuration

Command: weka cluster host apply

This command is used to apply the Weka system cluster hosts' configuration. In the install phase, all hosts need to be added, so the --all parameter can be used.

To activate the cluster hosts, use the following command line:

weka cluster host apply [--all] [<host-ids>...] [--force]

Parameters in Command Line

Name

Type

Value

Limitations

Mandatory

Default

host-ids

Comma-separated strings

Comma-separated host identifiers

Either host-ids or all must be specified

all

Boolean

Apply all hosts

Either host-ids or all must be specified

force

Boolean

Do not prompt for confirmation

No

Off

Stage 14: Running the Start IO Command

Command: weka cluster start-io

To start the system IO and exit from the initialization phase, use the following command line:

weka cluster start-io

After successful completion of this command, the system exits the initialization state and runs accept IOs from the user applications.