Hi Chris,
Thanks for your comments. Please see
my comments in-line in your email below.
Have a great weekend! Happy July 4th!
Regards,
Sahdev Zala
From:
Chris Lauwers <
lauwers@ubicity.com>
To:
Matt Rutkowski/Austin/IBM@IBMUS,
"tosca-interop@lists.oasis-open.org" <
tosca-interop@lists.oasis-open.org>
Date:
07/01/2015 05:43 PM
Subject:
RE: [tosca-interop]
2015-06-29 Interop. SC minutes
Sent by:
<
tosca-interop@lists.oasis-open.org>
I ran the latest CSAR file (2015-06-14
tosca_elk_DRAFT_CSAR.zip) through my validator, and I have a couple more
comments:
1. There is an indentation
error in logstash.yaml. Under the “search_endpoint” requirement, the
capability, node, and relationship keynames should all be indented equally
as follows:
node_types:
tosca.nodes.SoftwareComponent.Logstash:
derived_from: tosca.nodes.SoftwareComponent
requirements:
- search_endpoint:
capability: tosca.capabilities.Endpoint
node: tosca.nodes.SoftwareComponent.Elasticsearch
relationship:
Yes, Christopher
and I noticed this while testing the CSAR deployment and fixed it but I
didn't update new draft CSAR in order to compile possible findings as Christopher
Kaufmann and my team were in mid of testing.
2. In tosca_elk.yaml,
the mongo_db node template is missing the following required properties:
· port
· name
These properties are not always required
by certain default installation (e.g. MongoDB installation) and ideally
should be optional. In the latest spec, the port is already optional and
there is a comment to make name optional. I understand you are just pointing
to it and hopefully we will have it resolved in spec soon.
3. There are a number
of instances where operations in interfaces assign values to inputs that
haven’t been previously defined. According to the spec, it is illegal
to assign a value to an input that has not been previously defined. Here
are the instances:
· In
tosca_elk.yaml, the app_collectd node template defines the Configure operation
of the Standard interface. Specifically, it defines an implementation but
it also assigns a value to the “logstash_ip” input. However, the interface
definition does not define a “logstash_ip” input.
· similarly,
the kibana node template defines the Configure operation of the Standard
interface. Specifically, it defines an implementation but it also assigns
a value to the “kibana_ip” and “elasticsearch_ip” inputs. However,
the interface definition does not define a “kibana_ip” or “elasticsearch_ip”
input.
· Same
problem with the app_rsyslog node template. It assigns a value to an undefined
“logstash_ip” input.
· Same
problem with the mongo_dbms node template. It assigns a value to an undefined
mongodb_ip input
We (Fujitsu and
my team at IBM) have already started creating demo. I have updated a new
draft that we are using for the demo and it is locked down for now unless
we see any issue with deployment or as we discuss in the interop group
and decide to make further changes. I will be considering the above changes
after OSCON demo as they will require testing the full deployment which
is time consuming and also requires changing demo recording. I hope that's
OK.
Also, the CSAR (the latest draft I created
from) is available here -
https://github.com/openstack/heat-translator/tree/master/translator/toscalib/tests/data/CSAR/tosca_elk Chris
Original Message-----
From: tosca-interop@lists.oasis-open.org [ mailto:tosca-interop@lists.oasis-open.org ]
On Behalf Of Mr. Matthew Rutkowski
Sent: Monday, June 29, 2015 9:08 AM
To: tosca-interop@lists.oasis-open.org
Subject: [tosca-interop] 2015-06-29 Interop. SC minutes
Minutes for 2015-06-29 Interop. SC call:
[08:05] Matt Rutkowski (IBM): Attendees:
Matt, Derek, Sahdev, Chris, Karsten, Peter [08:10] anonymous: tried also
to join the conference services room 6789269 but system is saying not yet
been started ? is that only an issue on my side [08:10] Matt Rutkowski
(IBM): who is anonymous?
[08:10] anonymous morphed into Peter Gibbels
[08:10] Matt Rutkowski (IBM): can u set your name and company please in
settings [08:10] Matt Rutkowski (IBM): ahh [08:10] Peter Gibbels morphed
into Peter Gibbels (HP) [08:10] Matt Rutkowski (IBM): Hi Peter [08:10]
Peter Gibbels (HP): Hi Matt [08:11] Matt Rutkowski (IBM): we are not screen
sharing at the moment [08:11] Peter Gibbels (HP): o.k thanks, then I am
o.k :-) just wondered [08:11] Matt Rutkowski (IBM): I can start if we need
to look at some files, but we have not had to recently [08:14] Matt Rutkowski
(IBM): Issue in create scripts identified [08:14] Matt Rutkowski (IBM):
with apt-get [08:15] Matt Rutkowski (IBM): serialization issues because
multiple create scripts running and apt-get not parallel processing aware
[08:15] Matt Rutkowski (IBM): Derek: This is not a limitation of Linux,
code does not hurt, but needs to acknowledge the real problem and impls.
must serialize appropriately [08:16] Matt Rutkowski (IBM): Derek: on same
Compute (node) you should not parallelize scripts (too much assumed is
dangerous) should serialize typically [08:17] Matt Rutkowski (IBM): Sahdev:
collectd and rsyslog are running in parallel in same node. Others run on
their own nodes.
[08:18] Matt Rutkowski (IBM): Derek: Having
the loop is never a good idea. Cant rely on scripts to consider order and
parallel processing [08:18] Matt Rutkowski (IBM): Derek: All TOSCA orchestrators
should handle serialization and ordering [08:20] Matt Rutkowski (IBM):
Derek: recommend serialization in the same node. Transient states possible.
Parallelism is a goal, but safer to serialize [08:21] Matt Rutkowski (IBM):
Matt: please annotate scripts as a Heat-Translator workaround (not able
to yet serialize) and can be removed by other TOSCA impls [08:21] Matt
Rutkowski (IBM): Derek: Why do we have python and bash scripts in different
folders?
[08:22] Matt Rutkowski (IBM): Matt: TOSCA
needs to show declarative and not have these kinds of workaround code or
no value over imperative [08:22] Matt Rutkowski (IBM): Sahdev: just a logical
separation [08:22] Matt Rutkowski (IBM): Derek: when we extract we move
scripts to different places, would need to adjust to this paradigm [08:23]
Matt Rutkowski (IBM): Derek: so CSAR structure is just the designers preference
[08:23] Matt Rutkowski (IBM): Derek: Do we have MIME info?
[08:25] Matt Rutkowski (IBM): Matt: we
SHOULD support filling in MIME type info, it is in spec. for both BASH
and Python [08:25] Matt Rutkowski (IBM): Matt: SHOULD be in template def.
for artifacts, if not we SHOULD lookup based upon file extension [08:26]
Matt Rutkowski (IBM): Derek: not big deal, his developers can adapt ifnot
there [08:30] Matt Rutkowski (IBM): Matt: Simple profile should be able
to auto. identify .py and .sh extensions and assign a MIME type based upon
norm. types defined in spec.
[08:32] Matt Rutkowski (IBM): Matt: Sahdev
continue to work with Chris on making sure our 2 demos have variety [08:39]
Matt Rutkowski (IBM): Matt: Sahdev will upload a video to the TC by EOD,
tuesday July 7th; Final by July 9th for TC review [08:39] Matt Rutkowski
(IBM): Matt: this helps Fujitus and others using same HEat-Translator codebase
see what we intent to show [08:46] Matt Rutkowski (IBM): Derek: will we
splice videos?
[08:46] Matt Rutkowski (IBM): Matt: not
a goal, more jukebox approach [08:46] Matt Rutkowski (IBM): Karsten: Eurocloud
we rotated videos, IBM, HP< Huawei etc.
[08:46] Matt Rutkowski (IBM): Derek: we
may want to focus on tooling [08:47] Matt Rutkowski (IBM): Peter: Presenter
(whoever was up at the time) could choose videos they were most comfortable
showing [08:51] Matt Rutkowski (IBM): Matt: IBM is shooting for 3 mins
max.
[08:52] Matt Rutkowski (IBM): Matt: MPEG4
that works on MS Player and Quicktime player [08:54] Matt Rutkowski (IBM):
Matt: upload to TC working directory, in past we had new directory but
no one used so just use the one we use at TC for everything [08:54] Matt
Rutkowski (IBM): Derek: In future we should work to show diff script types
for different platforms (e.g. RHEL vs Ubuntu) [08:59] Matt Rutkowski (IBM):
Matt: we are adjourned