Validation script/command line

Command line to validate a snapup and get statistical information.

Resolution

The basic command line to validate a snapup and get statistics is:

% avtar --validate --allnode --stats --account= [ --id= --password= ]

This  validates the most recent snapup for the specified host. You can use  the --label= or --sequence= flags  to specify a specific snapup for the host. For example:

% avtar --validate --allnode --stats --account=/clients/test --seq=10 --id=root --password=abc

This  will cause the entire contents of snapup #10 from /clients/test to be  transferred across the network to the client, but it will suppress  writing the data to the local disk.

The --allnode flag causes  the client to establish one socket connection with each storage node  (five connections in this case) so that all data can be moved directly  from the storage node holding the data to the client. Without --allnode  the client opens a single connection to a randomly chosen storage node,  and when data must be fetched from a different node it is relayed  through that node. This, of course, restricts throughput.

The  --stats flag dumps various statistics at the end of the run. The most  important statistic is probably "total bytes/second", which is the  average rate at which data was moved.

Also, you can optionally add the flag "--verbose=2" to see the name and size of each file as it is being validated.

Some things you might want to try:

1.  Run a validate on the Axion Services Node. Since the Axion Services  Node and all of the Storage Nodes have Gig-E connections, this should  yield the maximum rate that the server can restore a snapup.

2.  Run a validate of a different snapup from one of your clients. If the  client has sufficient bandwidth from the Axion Server, the validate  performance will be comparable to what was seen on the Services Node.  Obviously, if there are network bottlenecks they will impact the  validation rate. Also, be aware that if you validate the same snapup  multiple times in a row you will likely see artificially fast results  since the five Axion storage nodes have 20GB of combined RAM for  caching.

3. Run multiple validations of different snapups  simultaneously on different client systems (or the Services Node) to  confirm that throughput is, in fact, being limited by the Axion Server,  and not by any client limitations.

4. After using validate to assess the server and  network performance to a particular client, run a restore by replacing  the "--validate" flag with "--extract". You might also want to add  something like "--target=/tmp/restore" to specify the destination  location of the restore.  The difference in performance will indicate  the overhead incurred by the local file system and disk subsystem.