System administration

##Backing up Cassandra data store##

Backing up Cassandra is essentially making a snapshot of occurrence store.

Cassandra comes with some command-line tools that we use for this task:

• cassandra-cli
• nodetool

###Synopsis on Cassandra occurrence store### Where indexed occurrence data is stored:

$ cd /data/cassandra/data/occ

(to provide directory description)

Connect to Cassandra and have a glimpse of stored records:

$ cassandra-cli
[default@unknown] use occ;
[default@occ] list occ limit 1;

In the last line an UUID is generated automatically by Cassandra. As long as a record is contained by the same data resource determined by druid, the UUID will be stable.

It's also possible if you want to print out a specific record:

[default@occ] get occ where uuid = 'e47e0e31-ff9c-4f31-b598-34f452cb023f';

###Making a snapshot### Assuming it's the first time we make a snapshot, this directory should be empty before we do:

$ cd /data/cassandra/data/occ/occ && ls

(By default this directory is owned by root so you will need to sudo.)

Now make a snapshot of occ, which we store occurrence data:

$ nodetool snapshot occ

The terminal returns:

Requested creating snapshot for: occ 
Snapshot directory: 1406163740504

A directory 1406163740504 is created under /data/cassandra/data/occ/occ/snapshots. If you list the files under the 1406163740504 directory, you'll notice it has the same files in /data/cassandra/data/occ/occ. 1406163740504 is where you backup occurrence store and /data/cassandra/data/occ/occ is where backed up occurrence data would be restored.

(@todo explain location when biocache sampling is running)

###Pointing to a remote Cassandra instance### Chances are you want to use a remote Cassandra instance. To do this, update listen_address: localhost in /etc/cassandra/cassandra.yaml by replacing 'localhost' with the domain name of the remote Cassandra.

##Backing up Solr index## To be written.