[Starlingx-discuss] Questions related to FM containerization

Arevalo, Mario Alfredo C mario.alfredo.c.arevalo at intel.com
Wed Mar 13 21:52:12 UTC 2019


Hi Brent and team,

This mail plans to interchange information in order to get more granularity about the "Fault Manager"
Containerization process where the final objective is to deploy a pod or pods which make completely
match with the design of containerization architecture.

According from the feedback received yesterday in our meeting and taking a look to the source code
and the processes which are running in a deployed environment I can understand the possible missing
parts.

This is my understanding about the full picture and some questions about it, please let me know if something
is wrong or deviates from the original objective/result. The storyboard [1] related to FM chart will be
finished when the FM restful API and the FM manager main services are containerized and work correctly
with the rest of the StarlingX components:

-    I think the best approach is to run a container per service and both containers should live in the same pod.
-    Both services should consume the same Docker image in order to reduce the memory foot print.
-    According the previous shared information and my code exploring I understand the FM manager (fmManager
     binary) is a service which listens in the port 8001 and executes requests. It is launched by an script called "fminit".
     At this point I have a pair of questions, is this the service which the storyboard [1] makes reference of,
     intended to have an instance in bare metal and another one exposed by a container? 
     If the last questions is yes, we will have 2 instances of the same service then they should be exposed in a
     different port, and how about fm rest api service, it will just exist in a container?
     FM manager service has interaction with a PostgreSQL data base, then I wonder how this will be handled.
     Does this containerized service point to the bare metal database? Should we add a new chart which exposes
     a PostgreSQL due to the current chart tarball or just include one for MariaDB.
     Then the task 28876 makes reference to modify nfv_vim[2] in order to get the alarms information from
     the container?
-    Surfing in a deployed image I saw that is possible to remove alarms from the web interface provided by
     horizon, then the task 28878 makes reference to points to the FM manager containerized service from
     horizon, however it is not clear for me the task 28877 described in the chart storyboard[1].

These are some points and questions which will give a better understanding for the rest of tasks, however,
possibly, I will get more questions during this process I will be in contact with all of you.

Thank you for your help.

Best Regards.
Mario.

References:

[1] https://storyboard.openstack.org/#!/story/2004008
[2] https://git.starlingx.io/cgit/stx-nfv/tree/nfv/nfv-vim/nfv_vim




More information about the Starlingx-discuss mailing list