[Tendrl-devel] Tendrl planning meeting 15-Nov-2017

Mrugesh Karnik mrugesh at brainfunked.org
Tue Nov 14 12:59:18 UTC 2017


Now that the foundations have been laid, there are a few areas where
I'd like to see tendrl improve in. Moving forward, I propose
maintaining the current line of 1.x codebase for the gluster 3.x
releases and in parallel starting work on the 2.x codebase for gluster
4.x releases. However, the latter needs more clarity and conversations
with gluster at this point and so, for this particular planning phase,
I propose that we focus on the following:


Enhancements to the currently supported deployment scenarios:

* Standard deployments: bare-metal or virtualised. This is the
currently supported default scenario. However, we should look at how
we can make the deployment composable by properly modularising our
existing components and formalising some interfaces
(monitoring-integration especially) by allowing custom integrations
with other infrastructure components for monitoring, logging etc.

* Co-located deployments. In this scenario, we need to be able to
co-locate the tendrl server components with gluster nodes. The
architecture already supports this, but we need to explicitly test out
co-location scenarios. This scenario may require additional
performance optimisation as well. Additionally, in a gluster
deployment, all gluster nodes are treated as equal in terms of their
role in the cluster. We need to specifically support and test out
scenarios for being able to dynamically relocate tendrl components
from a failed gluster node to another.


General feature additions:

* Improve the monitoring and alerting functionality. We need to look
at whether we can shave off some latency from the notifications. We
also need to support more granular configuration of the alert
destination and the corresponding filtering of the alerts based on the
metadata. As mentioned above, monitoring-integration could use some
modulerisation to be able to support different types of monitoring
systems such as prometheus. Prometheus' alert manager also needs to be
re-considered as a (stand-alone) replacement for our notifier since it
seems to already provide some of the functionality we'd like to in our
notifier.

* Bringing back the provisioning and management features. We had coded
and (partially?) tested these. Would this be a good time to start
re-incorporating these old features back into the 1.5.x codebase? This
time, however, it would be good to focus on a specific use case and
make the features available to support that use case fully. This leads
to the next point..

* Supporting networked file system protocols such as NFS and Samba. We
could look at supporting a use case for Windows administrators who
would want to be able to consume gluster volumes mounted via Samba. We
need to expand on this use case with details that enable us to plan
the re-incorporation of the management features.


User interface:

* In terms of the UI, one critical part missing is the complete lack
of health and deployment information about the tendrl components
themselves.

* Our current UI caters to an end user who wants to manually look at
and manage the nitty-gritties of a gluster, including volume and brick
configuration. If we are to serve a user such as a Windows admin, we
need to look at providing a storage management UI, rather than a
gluster management UI. This means that we need to shift our focus to
be able to cater to 'how do I provide <storage capacity> to be
consumed over <protocol> for <use case> quickly and painlessly.' This
involves a higher level management view that takes away the necessity
to micro-manage the deployment via our current cluster context.

--
Mrugesh

On 14 November 2017 at 13:50, Rohan Kanade <rkanade at redhat.com> wrote:
> Hello,
>
> Please join me for the Tendrl planning meeting.
>
> Date: 15-Nov-2017
> Time: 1415-1615 hrs (Indian standard time)
>
> Agenda:
> - Tendrl beyond 1.5.x, project roadmap
> - Per component roadmap discussion
> - Open discussion
> _______________________________________________
> Tendrl-devel mailing list
> Tendrl-devel at redhat.com
> https://www.redhat.com/mailman/listinfo/tendrl-devel




More information about the Tendrl-devel mailing list