FreeBSD’s HAST

HAST or “Highly Available Storage”, is a mechanism built into FreeBSD that allows for transparent syncing of block data between two physical hosts (connected via a network).

HAST is a “primary/secondary” cluster, meaning that there will always be a single source of truthieness (the master/primary)

Advantages of a HAST Cluster

1. If you have an NFS export of a HAST’d volume and the ability to read the local disk vanishes, HAST will automatically read the data from the secondary server (assuming it’s available!)
2. RAID-like data consistency can be achieved

caveats

1. Only two machines are supported. (Primary and Master)
2. If you are operating in a “degraded” state (local disk on primary is missing), read/write speeds will be reduced as the TCP overhead / network latency will come into play (this overhead will change depending on the sync method used)

Replication Modes

1. memsync - This mode will report a successful write when the following are met:
   1. After the local write is successful.
   2. Remote node sends acknowledgment of data receipt. (Receipt is sent prior to writing data)
2. fullsync - The safest of the replication modes, fullsync waits for write of the data on both hosts prior to marking a write successful.
3. async - This is the most dangerous mode. Once data is written to the local disk, the write operation is considered successful. If you are on a high latency network, this is the option you want to used.

Example Config

For the sake of the example, assume we have two machines (fbsd_h1 and fbsd_h2) and both have a /dev/da0 disk which contains the data we want to sync.

The config file (/etc/hast.conf) must be the same on both the primary and secondary machines!

listen tcp://0.0.0.0

resource shared {
    local /dev/ada0p4

    on fbsd_h1 {
        remote tcp://10.0.0.231
    }
    on fbsd_h2 {
         remote tcp://10.0.0.233
    }
}

Now we can start up hastd on each client and tell the cluster which host is the master!

fbsd_h1# hastctl create shared
fbsd_h1# hastd
fbsd_h1# hastctl role primary shared
fbsd_h1# newfs -U /dev/hast/shared
fbsd_h1# mount -o noatime /dev/hast/shared /data

fbsd_h2# hastctl create shared
fbsd_h2# hastd
fbsd_h2# hastctl role secondary shared

If all goes to plan when you run hastctl status you should see something like this:

fbsd_h1# hastctl status
Name	 Status		Role	Components
shared	 complete	primary	/dev/ada0p4	tcp://10.0.1.233

fbsd_h2# hastctl status
Name	 Status		Role		Components
shared	 complete	secondary	/dev/ada0p4	tcp://10.0.1.231

Now you are running a memsync’d HAST cluster!

Further reading

1. hast.conf(5)
2. hastctl(8)
3. hastd(8)
4. HAST Wiki

Questions?