A set of tools to easily develop and manage Elasticsearch and the Elastic Stack
Get from Docker Hub:
docker pull bigdataboutique/elastician
Build locally:
docker build .
docker run --net=host -v `pwd`:/data --rm bigdataboutique/elastician dump myindex --hosts http://10.63.246.27:9200
docker run --net=host -v `pwd`:/data --rm bigdataboutique/elastician ingest myindex_dump.jsonl.gz myindex --hosts http://10.63.246.27:9200
Both dump and ingest now have additional parameters to allow accessing ES using a certificate. Make sure the certificate file has the appropriate file permissions.
--crtfile
- name of the certificate file. Notice that unless you specify a folder this should sit in the working directory of the container, which is /data .
--verify-cert/--no-verify-cert
- Whether you're using a signed cert and want to verify it. By default false.
docker run --net=host --rm bigdataboutique/elastician copy source_index_name --target http://target-cluster:9200 --source http://10.63.246.27:9200
docker run --net=host -v `pwd`:/instructions --rm bigdataboutique/elastician copy-cluster /instructions/source.csv /instructions/target.csv --target http://target-cluster:9200 --source http://10.63.246.27:9200
First file name is csv with rows in the following structure:
instruction,index_name,to_del
where instruction is either copy
, dump
, delete
, to_del is X
if the index in the source is to be deleted following the action
For ingest
the format is:
ingest,file_name,index_name
Other parameters for copy-cluster:
--abort-on-failure
: stops the default is true
--dump-timeout
- default is u'1m'
--delete-timeout
- default is 60
--error-on-timeout
: whether the delete function should return an error on timeout
--preserve-index/--no-preserve-index
- whether the index saved in the dumps should be used during ingest
Dump also has a read-timeout parameter which determines request timeout (as opposed to scroll timeout).
Copy and copy-cluster also have certificate parameters, that work similarly to those above:
--crtfile-target
--verify-cert-target/--no-verify-cert
--crtfile-source
--verify-cert-source/--no-verify-cert-source
Ingest and copy cluster have an ingest-timeout parameter. Default is 10, bump it up when you're getting timeouts while ingesting.
Dump's --sliced parameter allows dumping using a sliced scroll. Not yet added into copy cluster.
The --size parameter for dump and copy-cluster determines the bulk size