Add single-machine deployment example cfgs and scripts #7590
Conversation
docs/content/tutorials/index.md
Outdated
|
|
||
| You will need: | ||
| * Java 8 |
| # under the License. | ||
| # | ||
|
|
||
| druid.service=druid/overlord |
There was a problem hiding this comment.
Out of curiosity, why not run Coordinator asOverlord and move towards the world where the coordinator and overlord are one
There was a problem hiding this comment.
I initially felt like the simplicity benefit wasn't that high in this case (since premade configs and startup scripts are being provided), and felt that it would be closer to a full clustered deployment for eventual migration with them separate.
I updated the PR to use that setting, since it would be fine to set that in a larger cluster as well.
|
I think there are still a small set of things missing:
|
jon-wei
force-pushed
the
single_server_examples
branch
from
92db75d to
b4cdfa8
Compare
I'm working on another doc-only update PR that will provide guidance on important performance tuning properties, which will also adjust the current cluster.md docs and address these points. |
Cool. Maybe the ToC should reflect the path to production. Laptop --> Single Server --> Clustering (non-HA) --> Clustering (HA) |
Sounds good, I'll add that structure in the follow-on PR |
By the way, I don't think there's a need to separate 'laptop' and 'single server' (people can treat their laptop as a single server if they want to - there is really no meaningful difference). I also don't think there's a need to separate 'Clustering (non HA)' from 'Clustering (HA)' (HA should just be a section at the end of the clustering docs). |
|
|
||
| ### Hardware | ||
|
|
||
| Druid includes several example [single-server configurations](../operations/single-server.html), along with scripts to start the Druid processes using these configurations. |
There was a problem hiding this comment.
Please add this somewhere more prominent in a follow-up (ToC, perhaps).
- update Docker entrypoint script in respond to the new directory structure apache#7590 - update Dockerfile to allow dependency caching in multi-stage build Signed-off-by: Khwunchai Jaengsawang <[email protected]>
| @@ -34,7 +34,7 @@ else | |||
| CONFDIR="$2" | |||
| fi | |||
|
|
|||
| CONFDIR="$(cd "$CONFDIR" && pwd)/druid" | |||
| CONFDIR="$(cd "$CONFDIR" && pwd)" | |||
There was a problem hiding this comment.
Why was the /druid removed? It's a breaking change for older version users. It doesn't seem like an absolutely necessary change. This file is one of the bin scripts in the apache-druid.tar.gz distribution although the directory name is "examples". Please avoid similar changes in the future release. Thanks.
There was a problem hiding this comment.
Hi @knoguchi,
It was changed to comport with the new directory structure in this patch, where we ship with multiple potential configurations for various sizes of servers. (It no longer makes sense to add /druid to the conf dir, since we provide the full path to a Druid-specific conf dir in the supervise configs.)
I think we should have noted this in the release notes, though.
This PR adds a new set of reference configurations and launch scripts for single-machine deployments:
micro-quickstartsmallmediumlargexlargeThe
micro-quickstartis sized for small machines like laptops and is intended for quick evaluation use-cases.The other configurations are intended for general use single-machine deployments. They are sized for hardware roughly based on Amazon's i3 series of EC2 instances.
The old tutorial cluster configuration has been replaced with
micro-quickstart, with the tutorials also pointing users to the other single-server examples if they wish to use a bigger machine.This PR also adds example configurations for a clustered deployment based on the master/data/query organization, along with associated launch scripts.