Storage preparation¶
Prepare video storage on each Indexer or Standalone node before installation. The video storage directory defaults to /var/lib/forensiq/video_storage.
Storage requirements¶
ZFS is required when hardware RAID is not used. Place the video storage directory on a ZFS dataset and give ZFS direct access to the disks.
Deployments using hardware RAID may use another file system. A controller operating in passthrough or HBA mode does not count as hardware RAID; ZFS is still required in that case.
This requirement applies to video storage. The operating system and the rest of /var/lib/forensiq can use other file systems.
ZFS settings and memory¶
Apply these settings when using ZFS.
| Property | Value |
|---|---|
atime |
off |
xattr |
sa |
sync |
standard |
Keep sync=standard so Kafka's requests to persist data are honored. Do not use sync=disabled. This setting alone does not guarantee that every acknowledged Kafka message has reached disk.
Limit the ZFS memory cache (ARC) so enough RAM remains for the services. A 4 GiB limit was used in our tests; choose the limit from the node's available memory and service requirements.
Performance tuning (optional)¶
SLOG and special vdevs are optional performance tuning for ZFS pools backed by HDDs. Plain ZFS can be used without either.
Video storage uses Kafka. Most video writes are buffered, but Kafka also synchronizes its controller metadata log to disk. These small synchronous writes can compete with video traffic on HDDs. An SSD can reduce those waits.
Choose the SSD layout based on the installed OpenZFS version:
| OpenZFS version | SSD option |
|---|---|
| 2.4 or newer | Consider a mirrored SSD special vdev. It stores ZFS metadata and can also hold the intent log when no separate log device is present. |
| Before 2.4 | Consider an SSD SLOG to reduce synchronous-write delays. These versions normally cannot use a special vdev for the intent log. |
The OpenZFS 2.4 release notes describe the new intent-log support. Our upload tests passed with both plain ZFS and ZFS with a SLOG. The special-vdev option still needs validation with the intended hardware and workload.
For both options, use enterprise SSDs with power-loss protection. You can use whole SSDs or dedicated partitions, including partitions on OS SSDs that meet these requirements. File-backed devices are not supported.
For a special vdev, use a mirror with at least the same fault tolerance as the HDD pool. Losing the special vdev can lose the entire pool. Size it for the pool's metadata with room to grow. Leave special_small_blocks=0 unless a tested configuration calls for it: this setting moves data blocks by size, and can also move video data onto SSDs. Kafka's metadata log is application data and is not automatically placed there as ZFS metadata.
For a SLOG, 16 GB is a starting point for this workload. This size does not apply to a special vdev.
Choose an installation method¶
Once the prerequisites are ready, choose the installation method that matches your server's internet access.
-
Online installation
If your server has internet access, proceed to the installation guide for instructions on how to install ForensiQ.
-
Offline installation (experimental)
If your server does not have internet access, you can use the offline installation method to install ForensiQ.