Cockroach Admin UI Suggestions

[Max Lang]

Hey!

I recently joined the CockroachDB team to build our Admin UI, and I wanted to solicit feedback from the community as I start to build it out.

What features / dashboards would you like to see and what user interfaces (admin/monitoring or otherwise) do you enjoy (or hate) using?

Thanks,

Max

[jhart...@gmail.com]

Enjoy using: RethinkDB's UI. Fantastically simple and well executed interface. The features that it does include (though not exhaustive) work really well, and they hit the important ones (query builder & data explorer, cluster overview / health checks, basic database admin).

Re: Cockroach, it'd be nice to see topology overview in a drill-down fashion, starting at the data center.

Exciting to follow Cockroach's progress, keep up the good work.

[Max Lang]

I think a good topology diagram will be a great way to allow people to understand/explore their cluster. While it's not currently in the works for beta we've been discussing what such a diagram would look like and how it will scale to clusters with 100s or 1000s of nodes. If you know any good examples there I'd love to check them out.

I took a look at RethinkDB's UI and it looks great. It's easy to see your cluster is up and healthy, and I liked how easy it was to change configuration settings in place (eg with sharding/replication.) We'll definitely have to take inspiration from them as we go.

I'll check back with the google group soon, once we have a good baseline, but feel free to follow along in Github as well, and thanks for the feedback!

-Max

--
You received this message because you are subscribed to the Google Groups "Cockroach DB" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cockroach-db...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[win.ma...@gmail.com]

I agree with Max that a good topology diagram that identifies DCs and nodes would be very useful. Maybe have a list view on the side of the topology view organized by DC with some basic stats and a health indicator to get the best of both worlds. Selecting a DC or node could overlay the list sidebar with just that DC or node's details replacing the list (this is a web UI pattern we have used many times with great success). Having a dynamic dashboard with the ability to select widgets like a topology view, graphs, and cluster health indicators and organize appropriately is pretty handy for filtering what you want to see.

An overlooked feature is having an alarm that can make an audible alert so we can put the dashboard up in our NOC and have our people call out when something is awry. Also useful is to expose your alerts via SNMP so monitoring systems can consume state and do alerting and keep metrics.

Provisioning nodes would be very useful as well, using user/pass or pre-shared key authentication to gain access and install a monitoring agent or the full blown system can speed adoption.

Not really a UI issue but integration with saltstack, ansible, and puppet for provisioning would be very welcome.

[Max Lang]

Good idea - putting DCs between cluster and nodes as the 2nd hierarchical grouping is something we aren't doing currently but probably should do. Drilling down into specific DCs, nodes, and timescales will be important for anyone attempting to root-cause an issue.

Interesting idea about the audible alarm, I actually haven't heard that proposal before, but sounds like it could be fun/useful if you're willing to keep an ops dashboard up.

Personally, I think provisioning new nodes directly from the UI would be a killer feature but out of scope for the early versions. I think we'd need to make it pluggable somehow to make it easy to add support for multiple cloud providers. I think there are some tricky security issues to resolve there as well. Still it's something I'd like to see eventually.

 And yeah, finally, I agree that we'll need integrations with other systems. I don't know if we'll build these all out explicitly, but hopefully if we expose good enough APIs it'll be easy enough to wire up your tool of choice.

Thanks again for the input and I hope to follow up with you guys in a few months :)

[Daniel Theophanes]

As a DBA or developer, I would want to look at a history of queries sorted by cost and execution time. I would want a view that showed queries that had to be re-run N number of times. I would want a dashboard to show currently running queries with the option to abort them if the dev forgot to put in an appropriate bounding clause.

[ramku...@gmail.com]

A good example of executing SQL queries from Web can be found here: http://rickbergfalk.github.io/sqlpad/

Tagging and saving queries is really nice and helpful feature.

[Ajaya Agrawal]

Hi,

You could also look for ideas at UI built for Cassandra by Datastax.