Yesterday, I gave a webinar about managing Ceph and Storage with openATTIC for the SUSE Monthly Open Source Talks. In case you missed it, a recording of the session is now available from the "on demand" section of that web site.

I really enjoyed giving this presentation and sharing an update about the current state of development with the audience. In fact, I got somewhat carried away, so I nearly ran out of time! Unfortunately this left no time for doing Q&A after the presentation, I apologize for that. However, we took note of all your questions, which I will answer below.

Will a link to the RPM builds for SUSE be added to the openATTIC download page?

Very good suggestion, thanks! The installation on SUSE was already documented in the installation instructions, we have now added a link to the SUSE repos to the download page as well.

Besides software package downloads, is there a software (VM) appliance in the works and what requirements might there be for such an evaluation?

Not at this time, sorry. We have an open issue for that (OP-505), but have not yet gotten around to implement this. If you don't want to perform an installation, you can use our live demo for getting a first impression (it will be refreshed every top of the hour).

Another idea would be to provide a docker image (for a pure Ceph management instance of openATTIC) - we're open for suggestions or contributions here!

When it comes to requirements, this depends heavily on your intended use case. If you want to evaluate the storage management functionality, your test system should have at least one dedicated disk drive for data. If you're just interested in the Ceph management part, you don't actually need an additional disk. Other system requirements are outlined in the documentation.

Is openATTIC modular enough to be used to manage/monitor various storage technologies (NAS/SAN/Ceph) from a single deployed instance?

Yes, it is. In fact you can manage multiple openATTIC instances (and Ceph clusters) from the same WebUI. The only requirement is that all openATTIC instances have shared access to the PostgreSQL database that hosts the openATTIC runtime data.

Are there any plans to integrate with hardware management such as libstoragemgmt?

We're evaluating if it makes sense to replace our custom storage management code with calls to libstoragemgmt, this is tracked in OP-1155. One concern is that this framework needs to be available and fully supported on all the distributions we support, this requires further evaluation/investigation.

Can you elaborate on the proposed "Salt integrations" going forward?

As a first step, we intend to start using Salt Open for the remote management and automation of tasks related to managing Ceph nodes. In particular, we will create support for SUSE's Salt-based Ceph management framework that is currently in the works. This work and related subtasks is tracked in issue OP-1243 in our public Jira. However, once the basic infrastructure support for Salt is in place, we could also use it for performing other storage-related management tasks, for example as a replacement for the openATTIC systemd process.

Any plans for OSD and PG troubleshooting? I am thinking even as much sane configuration suggestions based on errors and warning from ceph status.

That would indeed be a good idea! As a first step, we will start with giving the administrator more insight into various performance and runtime information via monitoring and graphing. Once we can keep track of this information, we're in a better position to analyze it and come up with warnings and suggestions on how to fix potential issues based on patterns. In addition to that, we're thinking of adding an editor similar to the PG calculator to generate sane PG numbers for all pools (see OP-1072 for details).

Is CephFS support on the roadmap?

It certainly is - this work is tracked in issue OP-925 and its subtasks. What CephFS management features would you most be interested in? Please let us know.

Will openATTIC allow for management of individual devices (start, stop, in, out, create, destroy)?

If you're referring to Ceph OSDs here, then yes - this depends on the Salt-based management framework to be complete and is tracked via the subtasks of OP-1004.

Will openATTIC make it easier to manage lrbd configurations and include such concepts as LUN masking?

Making it easier to manage lrbd configurations is one of the goals we just discussed with the SUSE developers during last week's hackweek. Again we'll likely be using Salt here, the integration of lrdb support is tracked in OP-775.

Will the nagios/icinga integration also monitor the throughput of the various networks (public/cluster) and individual interfaces?

As a first step, we'll add monitoring for all resources that can be queried via librados locally. In a next step, we plan to use Salt to deploy and configure collectd on all Ceph nodes, to collect this kind of node-specific performance data. See OP-1244 for details.

Is there support for a Ceph cluster configuration report to identify differences in node configurations?

Not that I am aware of - would you mind elaborating on your idea by submitting an issue in our public Jira tracker for this, or by contacting us via IRC or the Google Group?

Thank you!

Comments