Mounting Filesystems
To use a filesystem via the Weka filesystem driver, it has to be mounted on one of the cluster hosts. This page describes how this is performed.
Overview
There are two methods available for mounting a filesystem in one of the cluster hosts:
Using the traditional method: See below and also refer to Adding Clients (Bare Metal Installation) or Adding Clients (AWS Installation), where first a client is configured and joins a cluster, after which a mount command is executed.
Using the Stateless Clients feature: See Mounting Filesystems Using the Stateless Clients Feature below, which simplifies and improves the management of clients in the cluster and eliminates the Adding Clients process.
Traditional Method for Mounting a Filesystem
Note: Using the mount command as explained below first requires the installation of the Weka client, configuring the client, and joining it to a Weka cluster.
To mount a filesystem on one of the cluster hosts, let’s assume the cluster has a filesystem called demo
. To add this filesystem to a host, SSH into one of the hosts and run the mount
command as the root
user, as follows:
The general structure of amount
command for a Weka filesystem is:
There are two options for mounting a filesystem on a cluster client: read cache and write cache. Refer to the descriptions in the links below to understand the differences between these modes:
Mounting Filesystems Using the Stateless Clients Feature
The Stateless Clients feature defers the process of joining the cluster until a mount is performed. Simplifying and improving the management of clients in the cluster, it removes tedious client management procedures, which is particularly beneficial in AWS installations where clients may join and leave in high frequency. Furthermore, it unifies all security aspects in the mount command, eliminating the search of separate credentials at cluster join and mount.
To use the Stateless Clients feature, a Weka agent must be installed. Once this is complete, mounts can be created and configured using the mount command and can be easily removed from the cluster using the unmount command.
Note: To allow only Weka authenticated users to mount a filesystem, set the filesystem --auth-required
flag to yes
. For more information refer to Mount Authentication section.
Assuming the Weka cluster is using the backend IP of 1.2.3.4
, running the following command as root
on a client will install the agent:
curl http://1.2.3.4:14000/dist/v1/install | sh
On completion, the agent is installed on the client machine.
Invoking the Mount Command
Command: mount -t wekafs
Use one of the following command lines to invoke the mount command (note, the delimiter between the server and filesystem can be either :/
or /
):
mount -t wekafs -o <options> <backend0>[,<backend1>,...,<backendN>]/<fs> <mount-point>
mount -t wekafs -o <options> <backend0>[,<backend1>,...,<backendN>]:/<fs> <mount-point>
Parameters in Command Line
Name | Type | Value | Limitations | Mandatory | Default |
| | See Additional Mount Options below | | | |
| String | IP/hostname of a backend host | Must be a valid name | Yes | |
| String | Filesystem name | Must be a valid name | Yes | |
| String | Path to mount on the local machine | Must be a valid path-name | Yes | |
Mount Command Options
Each mount option can be passed with an individual -o
flag to mount.
For All Clients Types
Option | Value | Description | Default | Remount Supported |
| None | Set mode to read cache.
Note: The SMB share mount mode is always | No | Yes |
| None | Set mode to write cache | Yes | Yes |
| Number in milliseconds | After the defined time period, every metadata cached entry is refreshed from the system, allowing the host to take into account metadata changes performed by other hosts. | 1000 | Yes |
| Number in milliseconds | Each time a file or directory lookup fails, an entry specifying that the file or directory does not exist is created in the local dentry cache. This entry is refreshed after the defined time, allowing the host to use files or directories created by other hosts. | 0 | Yes |
| None | Mount filesystem as read-only | No | Yes |
| None | Mount filesystem as read-write | Yes | Yes |
| 32, 64 or auto | Size of the inode in bits, which may be required for 32-bit applications. | Auto | No |
| None | Write debug logs to the console | No | Yes |
| None | Don't show any logs to console | No | Yes |
| None | Can be defined per mount. Setting POSIX ACLs can change the effective group permissions (via the | No | No |
| None | See Object-store Direct Mount section | No | Yes |
| None | Do not update inode access times | No | Yes |
| None | Always update inode access times | No | Yes |
| None | Update inode access times only on modification or change, or if inode has been accessed and | Yes | Yes |
| Number in seconds | How much time (in seconds) to wait since an inode has been accessed (not modified) before updating the access time. 0 means to never update the access time on access only. This option is relevant only if | 0 (infinite) | Yes |
| None | Do not take | No | Yes |
| None | Do not interpret character or block special devices. | No | Yes |
| None | Do not allow direct execution of any binaries. | No | Yes |
| Numeric (octal) notation of POSIX permissions | Newly created file permissions are masked with the creation mask. For example, if a user creates a file with permissions=777 but the First, the | 0777 | Yes |
| Numeric (octal) notation of POSIX permissions | Newly created directory permissions are masked with the creation mask. For example, if a user creates a directory with permissions=777 but the First, the | 0777 | Yes |
| Numeric (octal) notation of POSIX permissions | Newly created file permissions are logically OR'ed with the mode. For example, if a user creates a file with permissions 770 but the First, the | 0 | Yes |
| Numeric (octal) notation of POSIX permissions | Newly created directory permissions are logically OR'ed with the mode. For example, if a user creates a directory with permissions 770 but the First, the | 0 | Yes |
Remount of General Options
Mount options marked as Remount Supported
in the above table can be remounted (using mount -o remount
). When a mount option has been explicitly changed previously, it should be set again in the remount operation to make sure it retains its value. For example, if you mounted with ro
, a remount without it will default to rw
, while if you mounted with rw
, it is not required to be re-specified since this is the default).
Additional Mount Options Available using the Stateless Clients Feature
Option | Value | Description | Default | Remount Supported |
| Number | Amount of memory to be used by the client (for huge pages) | 1400 MiB | Yes |
| Number | The number of frontend cores to allocate for the client. Either If none are specified, the client will be configured with 1 core. If 0 is specified then you must use | 1 | No |
| Number | Specify explicit cores to be used by the WekaFS client. Multiple cores can be specified. | No | |
| String | This option must be specified for on-premises installation, and must not be specified for AWS installations. For more info refer to Advanced Network Configuration via Mount Options section. | No | |
| Number | Network bandwidth limitation for the entire container, in Mb/s. This limitation is for all nodes running within the container, and an attempt is made to detect it automatically based on the environment e.g., when in AWS. Setting a per-node limitation can be performed in the container definition file. | Auto-select | Yes |
| Number | The number of seconds without connectivity after which the client will be removed from the cluster. Minimum value: 60 seconds. | 86,400 seconds (24 hours) | Yes |
| Number | Traces capacity limit in MB. Minimum value: 512 MB. | No | |
| None | Controls the page allocation algorithm if to reserve only 2MB huge pages or also 1GB ones | Yes | Yes |
| Number in KB | Controls the readahead per mount (higher readahead better for sequential reads of large files) | 32768 | Yes |
| String | Path to the mount authentication token (per mount) |
| No |
|
| Determine whether DPKD networking dedicates a core ( |
| No |
| Number | Preferred requests rate for QoS in megabytes per second. | No limit | Yes |
| Number | Maximum requests rate for QoS in megabytes per second. Note, the allows bursting above that limit, but will aim to keep this limit in average. | No limit | Yes |
| Number | The timeout in seconds for establishing a connection to a single host. | 10 | Yes |
| Number | The timeout in seconds for waiting for the response from a single host. | 60 | Yes |
Note: These parameters, if not stated otherwise, are only effective on the first mount command for each client.
Note: By default, the command selects the optimal core allocation for Weka. If necessary, multiple core
parameters can be used to allocate specific cores to the WekaFS client. E.g., mount -t wekafs -o core=2 -o core=4 -o net=ib0 backend-host-0/my_fs /mnt/weka
For Example: On-Premise Installations
mount -t wekafs -o num_cores=1 -o net=ib0 backend-host-0/my_fs /mnt/weka
Running this command on a host installed with the Weka agent will download the appropriate Weka version from the hostbackend-host-0
and create a Weka container that allocates a single core and a named network interface (ib0
). Then it will join the cluster that backend-host-0
is part of and mount the filesystem my_fs
on /mnt/weka.
mount -t wekafs -o num_cores=0 -o net=udp backend-host-0/my_fs /mnt/weka
Running this command will use UDP mode (usually selected when the use of DPDK is not available).
For Example: AWS Installations
mount -t wekafs -o num_cores=2 backend1,backend2,backend3/my_fs /mnt/weka
Running this command on an AWS host will allocate two cores (multiple-frontends) and attach and configure two ENIs on the new client. The client will attempt to rejoin the cluster via all three backends used in the command line.
For stateless clients, the first mount
command installs the weka client software and joins the cluster). Any subsequent mount
command, can either use the same syntax or just the traditional/per-mount parameters as defined in Mounting Filesystems since it is not necessary to join a cluster.
It is now possible to access Weka filesystems via the mount-point, e.g., by cd /mnt/weka/
command.
After the execution of anumount
command which unmounts the last Weka filesystem, the client is disconnected from the cluster and will be uninstalled by the agent. Consequently, executing a new mount
command requires the specification of the cluster, cores, and networking parameters again.
Note: When running in AWS, the instance IAM role is required to provide permissions to several AWS APIs, as described in IAM Role Created in Template.
Note: Memory allocation for a client is predefined. Contact the Weka Support Team when it is necessary to change the amount of memory allocated to a client.
Remount of Stateless Clients Options
Mount options marked as Remount Supported
in the above table can be remounted (using mount -o remount
). When a mount option is not set in the remount operation, it will retain its current value. To set a mount option back to its default value, use the default
modifier (e.g., memory_mb=default)
.
Advanced Network Configuration via Mount Options
When using a stateless client, it is possible to alter and control many different networking options, such as:
Virtual functions
IPs
Gateway (in case the client is on a different subnet)
Physical network devices (for performance and HA)
UDP mode
Use -o net=<netdev>
mount option with the various modifiers as described below.
<netdev>
is either the name, MAC address, or PCI address of the physical network device (can be a bond device) to allocate for the client.
Note: When using wekafs
mounts, both clients and backends should use the same type of networking technology (either IB or Ethernet).
IP, Subnet, Gateway, and Virtual Functions
For higher performance, the usage of multiple Frontends may be required. When using a NIC other than Mellanox, or when mounting a DPDK client on a VM, it is required to use SR-IOV to expose a VF of the physical device to the client. Once exposed, it can be configured via the mount command.
When you want to determine the VFs IP addresses, or when the client resides in a different subnet and routing is needed in the data network, usenet=<netdev>/[ip]/[bits]/[gateway]
.
ip, bits, gateway
are optional. In case they are not provided, the Weka system tries to deduce them when in AWS or IB environments, or allocate from the default data network otherwise. If both approaches fail, the mount command will fail.
For example, the following command will allocate two cores and a single physical network device (intel0). It will configure two VFs for the device and assign each one of them to one of the frontend nodes. The first node will receive a 192.168.1.100 IP address, and the second will use a 192.168.1.101 IP address. Both of the IPs have 24 network mask bits and a default gateway of 192.168.1.254.
Multiple Physical Network Devices for Performance and HA
For performance or high availability, it is possible to use more than one physical network device.
Using multiple physical network devices for better performance
It's easy to saturate the bandwidth of a single network interface when using WekaFS. For higher throughput, it is possible to leverage multiple network interface cards (NICs). The -o net
notation shown in the examples above can be used to pass the names of specific NICs to the WekaFS host driver.
For example, the following command will allocate two cores and two physical network devices for increased throughput:
Using multiple physical network devices for HA configuration
Multiple NICs can also be configured to achieve redundancy (refer to Weka Networking HA section for more information) in addition to higher throughput, for a complete, highly available solution. For that, use more than one physical device as previously described and, also, specify the client management IPs using -o mgmt_ip=<ip>+<ip2>
command-line option.
For example, the following command will use two network devices for HA networking and allocate both devices to four Frontend processes on the client. Note the modifier ha
is used here, which stands for using the device on all processes.
Advanced mounting options for multiple physical network devices
With multiple Frontend processes (as expressed by -o num_cores
), it is possible to control what processes use what NICs. This can be accomplished through the use of special command line modifiers called slots. In WekaFS, slot is synonymous with a process number. Typically, the first WekaFS Frontend process will occupy slot 1, then the second - slot 2 and so on.
Examples of slot notation include s1
, s2
, s2+1
, s1-2
, slots1+3
, slot1
, slots1-4
, where -
specifies a range of devices, while +
specifies a list. For example, s1-4
implies slots 1, 2, 3 and 4, while s1+4
specifies slots 1 and 4 only.
For example, in the following command, mlnx0
is bound to the second Frontend process whilemlnx1
to the first one for improved performance.
For example, in the following HA mounting command, two cores (two Frontend processes) and two physical network devices (mlnx0
, mlnx1
) are allocated. By explicitly specifying s2+1
, s1-2
modifiers for network devices, both devices will be used by both Frontend processes. Notation s2+1
stands for the first and second processes, while s1-2
stands for the range of 1 to 2, and are effectively the same.
UDP Mode
In cases where DPDK cannot be used, it is possible to use WekaFS in UDP mode through the kernel. Use net=udp
in the mount command to set the UDP networking mode, for example:
Note: A client in UDP mode cannot be configured in HA mode. However, the client can still work with a highly available cluster.
Note: Providing multiple IPs in the <mgmt-ip> in UDP mode will utilize their network interfaces for more bandwidth (can be useful in RDMA environments), rather than using only one NIC.
Mounting Filesystems Using fstab
Note: This option works when using stateless clients and with OS that supports systemd
(e.g.: RHEL/CentOS 7.2 and up, Ubuntu 16.04 and up, Amazon Linux 2 LTS).
Edit /etc/fstab
file to include the filesystem mount entry:
A comma-separated list of backend hosts, with the filesystem name
The mount point
Filesystem type -
wekafs
Mount options:
Configure
systemd
to wait for theweka-agent
service to come up, and set the filesystem as a network filesystem, e.g.:x-systemd.requires=weka-agent.service,x-systemd.mount-timeout=infinity,_netdev
Any additional
wekafs
supported mount option
Reboot the machine for the systemd
unit to be created and marked correctly.
The filesystem should now be mounted at boot time.
Note: Do not configure this entry for a mounted filesystem before un-mounting it (umount
), as the systemd
needs to mark the filesystem as a network filesystem (occurs as part of the reboot
). Trying to reboot a host when there is a mounted WekaFS filesystem when setting its fstab
configuration might yield a failure to unmount the filesystem and leave the system hanged.
Mounting Filesystems Using autofs
It is possible to mount a Weka filesystem using the autofs
command.
To get started, install autofs
on the host:
Then run the following commands to create the autofs
configuration files for Weka filesystems:
Or run the following commands for stateless clients (which require the backend names as parameters):
Finally, restart the autofs
service:
The configuration is distribution-dependent, and it is necessary to ensure that the service is configured to start automatically after the host is rebooted. To verify that the autofs
service automatically starts after restarting the server, run the following command: systemctl is-enabled autofs.
If the output is enabled
the service is configured to start automatically.
In Amazon Linux, for example, autofs
service can be verified with chkconfig
command. If the output is on
for the current runlevel (can be checked with runlevel
command), autofs
will be enabled upon reboot.
It is now possible to access Weka filesystems using thecd /mnt/weka/<fs-name>
command.
Last updated