- Fix port binding check when using
--starter.host
. It used to check against theany
interface, now it checks against--starter.host
.
- For
activefailover
deployments, the message where to reach your database is now only shown for the current leader. - Added SystemD example. See
examples/systemd/README.md
. - Added
--log.dir
option to configure a custom directory to which all log files will be written. - It is no longer allowed to use
log.file
as a passthrough option. - Added
--starter.host
option, to bind the HTTP server to a specific network interface instead of the default0.0.0.0
. - Added
POST /data-auto-upgrade
support to perform a rolling upgrade of all servers (with single--database.auto-upgrade
restart) - Renamed mode option
resilientsingle
toactivefailover
. (resilientsingle
is being supported as alias for a while) - Added support for log file rotation for started server components.
- Added support for running datacenter to datacenter replication servers (
arangosync
) from the starter. - Changed increment for TCP ports (used when running multiple starters on same machine (e.g
--starter.local
)) from 5 to 10. An increment of 10 is needed to run datacenter to datacenter replication servers. If you have an existing cluster that is using--starter.local
and you also want to enable datacenter to datacenter replication, you must create a new cluster. - Added support for environment variables to act as commandline arguments.
E.g.
ARANGODB_STARTER_DATA_DIR=/tmp/foo arangodb
equalsarangodb --starter.data-dir=/tmp/foo
.
- Using shard
http.Client
to reduce the number of used file descriptors.
- Support building in a directory other than the source directory. Set
BUILDDIR
(#98). - Testing a server instance now includes testing for the expected server role.
- On linux, also look for
arangod
in/usr/local/sbin
(#93). - Fixed potential for hang in starter behavior.
- Starting with mode
resilientsingle
and--starter.local
will no longer limit the number of single servers to 2.
- Remove
server.threads
&javascript.v8-contexts
from generatedarangod.conf
files. Both settings are nolonger needed.
- Added support for
resilientsingle
mode. A configuration of 2 single servers that replicate and take over when needed. - Removed
--cluster.my-local-info
option from commandline ofarangod
servers. It is obsolete. - Fixed combination of
--starter.local
&--cluster.agency-size=1
(do not use for production!)
- Added
--version
option andversion
command to show version of the starter.
- Added
--starter.disable-ipv6
option to cope with environments where IPv6 is actively disabled.
- Update to go 1.9.0
- Fixed registration for callback (in agency) by unreachable local slaves.
- Fixed port allocation in case of using
--starter.address=127.0.0.1
with--starter.local
(#79)
- Added
--docker.imagePullPolicy
option - Allow multiple
--starter.join
arguments. - Fixed high CPU load (#75)
- Avoid using port offsets when using different
--starter.port
s that cause no overlap of port ranges. - Cluster configuration is updated to all starters at regular intervals after the starters have bootstrapped and reached a running state.
- After starters have bootstrapped, they elect a starter to be master over the cluster configuration. All changes (addition/removal) are forwarded to this master. When the master is gone for too long, a new master is elected.
- Fixed cluster setup in case where starters use different
--starter.port
s (#68). - The
--rocksdb.encryption-keyfile
is now passed through the database servers in thearangod.conf
file (it was passed as command line argument before). If you use this setting in an existing cluster, make sure the manually add this setting to allarangod.conf
files before restarting the starters. - Added
--starter.debug-cluster
option that adds a trail of status codes to the log when starting servers. (intended mostly for internal testing) - Made database image used in test configurable using
ARANGODB
make variable. - Added
--docker.tty
option for controlling the TTY flag of started docker containers. - In cluster mode the minimum agency size has been lowered to 1 (DO NOT USE IN PRODUCTION).
- Added
start
command to run starter in detached mode. - Added
stop
command to stop a running starter using its HTTP API.
- Added path containing starter executable to search path for
arangod
.
- Added
--rocksdb.encryption-keyfile
option. - Added pass through options. See README.
- Changed
--data.dir
option to--starter.data-dir
- Added
--server.storage-engine
option, used to change the storage engine of thearangod
instances (#48) - Changed option naming scheme (see
arangodb --help
for all new names). Old names are still accepted. - Renamed github repository from
github.com/arangodb-helper/ArangoDBStarter
togithub.com/arangodb-helper/arangodb
. - When an
--ssl.keyfile
(or--ssl.auto-key
) argument is given, the starter will serve it's API over TLS using the same certificate as the database server(s). - Starter will detect the name of the docker container is it running in automatically (if running in docker and not set using
--docker.container
) - Changed default master port from 4000 to 8528. That results in a coordinator/single server to be available on well known port 8529
- Added
--starter.mode=single
argument, used to start a single server database instead of a cluster (#28) - Starter will check availability of TCP ports (both its own HTTP API & Arangod Servers) (#35)
- Docker container created by the started are given the label
created-by=arangodb-starter
- Added
--starter.local
argument, used to start a local test cluster in a single starter process (#25) - When an
arangod
server stops quickly and often, its most recent log output is shown - Support
~
(home directory) in path arguments. E.g.--data.dir=~/mydata/
(#30) - Changed port offsets of servers. Coordinator -> 1, DBServer -> 2, Agent -> 3.
- Added authentication support (#10)
- Added SSL support (#11)
- Fixed various IPv6 issues (#13)
- Attach starter to existing server processes (#6)
- Use same port offsets on peers running on different machines (#3)