If no chunking was requested by the user, the co-ordinating node buffers all
results in RAM before emitting a single result. However buffering was not
merging results for rows which had data for the same series. This change fixes this.
Fixes issue #3242.
* Capitalize first letter of message
* Log all services staring consistently
* Remove some extraneous log statements in meta.Store
* Log data dirs for meta, data and hinted handoff
The env var overrides panic if we if peers was set in the config.
The type of that value is a []string and NumFields is not a valid
for that type. We don't currently support this slice fields via
the env variable settings. This particular one can be set with the
-join flag already so we just skip these fields for now.
This allows all config variable to be set via environment variables using
a similar naming convention for the toml config. For example, to change the
HTTP API port using the config, you would set:
[http]
bind-address = ":8086"
To change it with an environment variable, you would use:
HTTP_BIND_ADDRESS=":8086" influxd
The section name is used as the env variable prefix and the config key
name is the suffix. The only change to the config name is that "-" should
be replaced with "_" to avoid shell interpretation issues.
This makes it much easier to configure docker instances within a docker container
or adhoc instances at the command-line.
For slice config sections like graphite, you can currently only override the first
entry since the default config only has 1 entry. To do that use, GRAPHITE_0 as the
prefix. You cannot currently add new entries like GRAPHITE_1. A future PR might
address this issue.
The environment variable values should be the same as the config values.
The order that configuration values are applied is as follows:
* Default config
* Config file
* Environment variables
* Command-line arguments
Fixes#3246
When starting a influxd in a docker container, the processess needs to know
the hosts address and port in order to create its NodeInfo correctly. -hostname
previously only allowed us to change the hostname and the port would always be 8088
which may not be correctly if running multiple containers on the same host.