Crail on shared volume block device

Crail-blkdev is an extension of the crail project to enable it to run on shared volume block devices.

Building

Clone and build jaio dependency. Clone and build the project using:

mvn -DskipTests install

Then copy the jar files crail-blkdev-1.0.jar and its dependencies from the targetfolder into$CRAIL_HOME/jars/.

Alternatively you can also put these files in your custom classpath.

Configuration parameters

There are two sets of parameters for a blk-device: client and target

The target parameters are as follows (shown with default values):

crail.storage.blkdev.storagesize	1073741824
crail.storage.blkdev.allocationsize	1048576
crail.storage.blkdev.ip         12.12.12.63
crail.storage.blkdev.port		12345

The client parameters are as follows (shown with default values):

crail.storage.blkdev.path		/dev/vdev1,/dev/vdev2
crail.storage.blkdev.ipport		12.12.12.62:54321,12.12.12.63:12345
crail.storage.blkdev.storagelimit	2147483648
crail.storage.blkdev.queuedepth 	8

The columns in in the path and ipport parameters reflect the mapping (iSCSI) that client has already setup between a target block device and local virtual block device; this is why there is no device specified on the target side. For example, in the configuration above, on the target, a block device is exposed at ip address 12.12.12.63 and port 12345. The client has already mapped this remote block device to path /dev/vdev2, with a protocol such as iSCSI.

Each column in ipport and path reflect at different block device target. Additionally, the storagelimit should be a sum of the capacity of all the block-device storage targets.

You can put these values in $CRAIL_HOME/conf/crail-site.conf.

Starting a crail-blkdev datanode

To start a crail-blkdev datanode, start a datanodes as

$CRAIL_HOME/bin/crail datanode -t com.ibm.crail.storage.blkdev.BlkDevStorageTier

in order for a client to automatically pick up connection to a new datanode type, you have to add following class to your list of datanode types in the $CRAIL_HOME/conf/crail-site.conf file. An example of such entry is :

crail.storage.types  com.ibm.crail.storage.rdma.RdmaDataNode,com.ibm.crail.storage.blkdev.BlkDevStorageTier

Please note that, this is a comma separated list of datanode types which defines the priorty order as well in which the blocks from a datanode will be consumed by the namenode.

Setting up automatic deployment

To enable deployment via $CRAIL_HOME/bin/start-crail.sh use the following extension in the crail slave file ($CRAIL_HOME/conf/slave.):

hostname1 -t com.ibm.crail.storage.blkdev.BlkDevStorageTier
...

Note: A crail-blkdev datanode does not serve data requests from clients, but only registers the block device storage information to the namenode. Such that clients can directly access shared volume block devices with the offset information provided by the namenode.

Alignment requirements

The crail block device client depends on jaio which requires buffers to be aligned to block size (512B). For any application using the crail block device client you need to set the aligned direct memory option of the JVM:

-Dsun.nio.PageAlignDirectMemory=true

For example, in spark you can set the property spark.executor.extraJavaOptions in spark-defaults.conf.

Contributions

PRs are always welcome. Please fork, and make necessary modifications you propose, and let us know.

Contact

If you have questions or suggestions, feel free to post at:

https://groups.google.com/forum/#!forum/zrlio-users

or email: [email protected] block device target.