|Main index||Section 8||Options|
hastd allows to transparently store data on two physically separated machines connected over the TCP/IP network. Only one machine (cluster node) can actively use storage provided by hastd. This machine is called primary. The hastd daemon operates on block level, which makes it transparent to file systems and applications.
There is one main hastd daemon which starts new worker process as soon as a role for the given resource is changed to primary or as soon as a role for the given resource is changed to secondary and remote (primary) node will successfully connect to it. Every worker process gets a new process title (see setproctitle(3)), which describes its role and resource it controls. The exact format is:
hastd: <resource name> (<role>)
If (and only if) hastd operates in primary role for the given resource, a corresponding /dev/hast/<name> disk-like device (GEOM provider) is created. File systems and applications can use this provider to send I/O requests to. Every write, delete and flush operation ( BIO_WRITE, BIO_DELETE, BIO_FLUSH) is sent to the local component and replicated on the remote (secondary) node if it is available. Read operations ( BIO_READ) are handled locally unless an I/O error occurs or the local version of the data is not up-to-date yet (synchronization is in progress).
The hastd daemon uses the GEOM Gate class to receive I/O requests from the in-kernel GEOM infrastructure. The geom_gate.ko module is loaded automatically if the kernel was not compiled with the following option: options GEOM_GATE
The connection between two hastd daemons is always initiated from the one running as primary to the one running as secondary. When the primary hastd is unable to connect or the connection fails, it will try to re-establish the connection every few seconds. Once the connection is established, the primary hastd will synchronize every extent that was modified during connection outage to the secondary hastd.
It is possible that in the case of a connection outage between the nodes the hastd primary role for the given resource will be configured on both nodes. This in turn leads to incompatible data modifications. Such a condition is called a split-brain and cannot be automatically resolved by the hastd daemon as this will lead most likely to data corruption or loss of important changes. Even though it cannot be fixed by hastd itself, it will be detected and a further connection between independently modified nodes will not be possible. Once this situation is manually resolved by an administrator, the resource on one of the nodes can be initialized (erasing local data), which makes a connection to the remote node possible again. Connection of the freshly initialized component will trigger full resource synchronization.
A hastd daemon never picks its role automatically. The role has to be configured with the hastctl(8) control utility by additional software like ucarp or heartbeat that can reliably manage role separation and switch secondary node to primary role in case of the primary's failure.
The hastd daemon can be started with the following command line arguments:
|Specify alternative location of the configuration file. The default location is /etc/hast.conf.|
| ||Print or log debugging information. This option can be specified multiple times to raise the verbosity level.|
| ||Start the hastd daemon in the foreground. By default hastd starts in the background.|
| ||Print the hastd usage message.|
|Specify alternative location of a file where main process PID will be stored. The default location is /var/run/hastd.pid.|
|/etc/hast.conf||The configuration file for hastd and hastctl(8).|
|/var/run/hastctl||Control socket used by the hastctl(8) control utility to communicate with hastd.|
|The default location of the hastd PID file.|
nodeB# hastd nodeB# hastctl role secondary shared
nodeA# hastd nodeA# hastctl role primary shared nodeA# newfs -U /dev/hast/shared nodeA# mount -o noatime /dev/hast/shared /shared
|HASTD (8)||February 1, 2010|
|Main index||Section 8||Options|
Please direct any comments about this manual page service to Ben Bullock.