|
Tuesday's meeting cancelled
Jeff ErnstFriedman <jernstfriedman@...>
Hello members of the OpenHPC TSC community, Due to Karl's travel schedule, Tuesday's (8/16) TSC meeting will be cancelled. We will resume the following week. Thank you, Jeff ErnstFriedman 2201 Broadway #604, Oakland, CA 94612 mobile: 510.593.1367 skype: jeffrey.ernstfriedman twitter: @namdeirf |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
[openhpc-users] Need advice on how to work with developers to contribute to OpenHPC
David Brayford
I apologies for cross posting, but the post below is relevant to the
current TSC discussions.
Would it be worth considering having a second repository of components that haven't been accepted (not rejected due for being unsuitable), define it as a development, alpha, beta version. That could be used as a staging post to the official OpenHPC (release) repository and allow the application to be evaluated before it is included in the official version. This could be a method for allowing research/new software to be disseminated to the larger HPC community. David -------- Forwarded Message
--------
Hello,
This is primarily a long question for the maintainers of OpenHPC.
I have quite a few questions regarding how open-source HPC performance tool developers contribute to OpenHPC. As background, I attended, in WW32, the Scalable Tools 2016 workshop, a workshop attended by largely open-source performance tool developers and with focus on scaling for today’s petaflop systems and preparing for the exa-scale successor generation. I inquired with the tool developers about willingness to contribute to OpenHPC by mirroring their current repositories to ohpc/components/perf-tools. Most of the developers raised concerns about doing this. I have one developer from Barcelona Super Computing Center who will add one of their performance tools, but I am not sure how to get the developer started. I list out the concerns and questions below:
Concerns ‐ Basic concerns are with optional components and the RPM structure, for example Dyninst and PAPI are often optional components that one can integrate into a tool - how are optional components handled? TAU is a very good example of this which has many build optional configuration build settings. ‐ How are package dependencies handled? For example, packaging up for both OS types: CentOS 7.2 and SLES12 ‐ Are source RPMs possible? I ask because most of these developers release source code and not RPMs ‐ How are specific recipes included in the package? ‐ The case here comes from one of the national laboratories who wants to ensure tools are installed and used as specified ‐ What are the guidelines and BKMs for developers to begin contributing? ‐ How are updates to OpenHPC validated? ‐ Do developers push their updates or is there another mechanism?
I will greatly appreciate all advice and pointers on this topic.
Regards,
Rashawn Knapp
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
SC student cluster competition
Reese Baird
Hi All –
Before the TSC was up and running, OpenHPC agreed to sponsor a team from SDSU in the SC student cluster competition. Intel and Nor-Tec have loaned and donated most of the hardware and a travel budget. The last bit of kit we need is a small (8 ports would do) FDR Mellanox IB switch. Would anybody on the list know of one that could be briefly loaned to the team? I can promise a lightly irreverent team T-shirt to the donor. Thanks, Reese |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Invitation to Slack...Come Join Us!
Jeff ErnstFriedman <jernstfriedman@...>
Members of the OpenHPC TSC community Based on conversations from the most recent OpenHPC TSC meeting. I have invited all members of the TSC mailing list to join the OpenHPC Slack Channel. As I type this I can see that some people have joined so congrats to Greg, Nirmala and Todd for being the first members! To facilitate communication and build community please add a photo and fill out your profile bio info so we know who is who. But I want to make sure everyone who wants to join has that opportunity. So if you have not received an invite let me know and provide me with the email address you'd like to be included under) and I will add you directly. To facilitate adoption I have allowed any member to invite any one else. I have created other, invite only channels for private conversations for various groups, ie Governing Board, TSC members with voting rights, etc. Jeff ErnstFriedman 2201 Broadway #604, Oakland, CA 94612 mobile: 510.593.1367 skype: jeffrey.ernstfriedman twitter: @namdeirf |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
HPCSYSPROS Workshop - DRAFT submission attached
Schulz, Karl W <karl.w.schulz@...>
Hello TSC,
Under the theory that it's better late than never, attached please find a first draft for review of the HPCSYSPROS workshop paper we have talked about in previous TSC meetings. I apologize that I wasn't able to complete assembly and send it out yesterday.
This paper is of the overview variety, and folks will notice our recently minted Mission/Vision statements highlighted on the 1st page.
Some reminders and comments:
* the submission deadline is today. I don't see a time listed, so I'll assume by end of day. I'm in the US/Central time zone and would like to shoot for submission by 8pm if possible (or earlier).
* author opt-in: please let me know if you would like to be included in the author list. As the list expands, the template recommends we use an additional authors section so we can add more here to accommodate interested TSC members.
* the page limit is 5 pages (+ refs), so we have a little wiggle room to add more, but not too much without cutting elsewhere
* reviews/suggestions: please feel free to send me comments/suggestions in whatever form is convenient. These can be pull requests, patches (thanks Yiannis for your patches to date), email, etc, and I'll do my best to incorporate. If you feel something
is particularly egregious, please advise on that as well.
For those who wish to build the paper locally, you will need Latex and should, in theory, be able to use "make" to generate a .pdf". The latest commits are on the master branch in the docs/papers/HPCSYSPROS subdirectory.
Thanks in advance for any feedback,
-k
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
FW: HPCWire Readers Choice awards voting has started
Schulz, Karl W <karl.w.schulz@...>
FYI, Derek noticed that OpenHPC is apparently listed in this years HPCWire
toggle quoted message
Show quoted text
readers choice awards. -k On 8/26/16, 10:45 AM, "Derek Simmel" <dsimmel@...> wrote:
Noticed OpenHPC in one of the categories - we should encourage members to |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Re: HPCSYSPROS Workshop - DRAFT submission attached
Schulz, Karl W <karl.w.schulz@...>
Hi folks,
Many thanks to all who had a chance to review the draft from earlier today and provide excellent feedback. Attached is the latest updated DRAFT with changes based on input received (and an additional paragraph in the related work section).
There was one question from Nirmala asking if a better title might be in order? I have left it as "Cluster Computing with OpenHPC" for now. One of her alternative suggestions was "High Performance Computing with OpenHPC". I don't have strong opinions
other than the fact that it should probably have "OpenHPC" somewhere in the title. Presumably we will have some time to tweak a bit if it were to be accepted, but if folks have ideas now or desperately want to see something different for today's submission,
please chime in.
I'm going to review a bit more and await any additional (hopefully small) suggestions you might have. I'll plan on waiting another 3 hours (till 9:30pm US/Central) before pulling the trigger.
-k
From: <openhpc-tsc-bounces@...> on behalf of "openhpc-tsc@..."
<openhpc-tsc@...>
Reply-To: "Karl W. Schulz" <karl.w.schulz@...> Date: Friday, August 26, 2016 at 9:19 AM To: "openhpc-tsc@..." <openhpc-tsc@...> Subject: [OpenHPC-TSC] HPCSYSPROS Workshop - DRAFT submission attached
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Re: HPCSYSPROS Workshop - DRAFT submission attached
Schulz, Karl W <karl.w.schulz@...>
And here is draft #3 (caught a few minor things on a fresh read).
-k
From: "Karl W. Schulz" <karl.w.schulz@...>
Date: Friday, August 26, 2016 at 6:26 PM To: "openhpc-tsc@..." <openhpc-tsc@...> Subject: Re: [OpenHPC-TSC] HPCSYSPROS Workshop - DRAFT submission attached
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
questions for OpenHPC issues
yiannis georgiou <yiannis.georgiou@...>
Hi all,
just being curious... let's say that I want to deal with issue: https://github.com/openhpc/ohpc/issues/268 Is there an official procedure or should I just go ahead and make a patch/pull request to correct it? I don't know if this is written somewhere or if it has already been discussed within the TSC but I couldn't find anything related. Concerning the technical choices of the solution: Is there a general plan for advanced configuration of components? Should we extend the existing "recipe.sh" script or introduce new ones? Currently Slurm has a pretty basic configuration by default so I expect people asking for more advanced configuration and tuning. Should I just propose a solution based on my judgement and then we review and decide or should this be discussed in advance? Thanks a lot for the clarifications, Yiannis |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Re: questions for OpenHPC issues
Schulz, Karl W <karl.w.schulz@...>
It's a good topic that we haven't really hit yet in our TSC discussions,
toggle quoted message
Show quoted text
so I can queue it ups for the next meeting. -k On 9/2/16, 9:25 AM, "openhpc-tsc-bounces@... on behalf of yiannis georgiou via OpenHPC-TSC" <openhpc-tsc-bounces@... on behalf of openhpc-tsc@...> wrote: Hi all, |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Re: questions for OpenHPC issues
Derek Simmel
Seems to me that this example belongs in the installation and/or best practices documentation for that component, explaining to deployer of the component how to configure that component (or related service and other components) for the desired result, rather than a "fix" to change default behavior of the installed component.
Fixes should be reserved for things that are broken, i.e. problems that make the component unreliable (errors), or unable to operate as intended, or unable to support required policies (build options/code that enable necessary configuration options to comply). If the component as supplied cannot support the desired configuration (feature request?) because it is missing essential code or missing other required components (e.g. a database), then modification of the component to satisfy the request should go through some review (prioritization, evaluation for side effects of the change - will adding this break things for existing users?, etc.) before being modified and released. Regardless of what default configuration is chosen for components, installation documentation and "recommended best practices" documentation should be provided to inform deployers how to get the behavior they require from the component. Perhaps we need a documentation librarian (enforcer?)... - Derek On Sep 2, 2016, at 10:39 AM, Schulz, Karl W via OpenHPC-TSC <openhpc-tsc@...> wrote:--- Derek Simmel Pittsburgh Supercomputing Center (412) 268-1035 |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Tuesday's meeting (9/6): Roll call vote
Jeff ErnstFriedman <jernstfriedman@...>
To the participants of the OpenHPC TSC, At the next meeting, the OpenHPC TSC will be holding a roll call vote on the communication path for the project: Public | Private | or Hybrid. This topic was discussed at length at during the Aug. 30 meeting, the details of which you can find here. Being of a matter pertaining to the regular business of the OpenHPC, we will be following the voting procedures as outlined in the OpenHPC charter. Only the official member roles and their designees with voting rights will be asked to provide a vote. Those members have been notified in a separate communication of the exact language of the vote. If you feel you are a member with voting privileges and did NOT receive the communication, please contact me immediately. See you all on Tuesday, Jeff ErnstFriedman 2201 Broadway #604, Oakland, CA 94612 mobile: 510.593.1367 skype: jeffrey.ernstfriedman twitter: @namdeirf |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Apologies
Jeff ErnstFriedman <jernstfriedman@...>
My dog was attacked this morning and we are in the emergency clinic.
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
meeting down
David Brayford
It appears that the meeting is down for me.
David |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Re: Apologies
Schulz, Karl W <karl.w.schulz@...>
Folks,
toggle quoted message
Show quoted text
Sorry for the technical issues on the meeting today. Let's cancel and I will follow up shortly with a related email discussion with the topics that were on the docket. -k On 9/13/16, 10:03 AM, "openhpc-tsc-bounces@... on behalf of Jeff ErnstFriedman via OpenHPC-TSC" <openhpc-tsc-bounces@... on behalf of openhpc-tsc@...> wrote: My dog was attacked this morning and we are in the emergency clinic. |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Re: Apologies
Schulz, Karl W <karl.w.schulz@...>
Following up on what we originally had on tap for today, attached are the
toggle quoted message
Show quoted text
slides we hoped to cover. There were 3 primary issues I was hoping to get feedback on: (1) with the component selection process mapped out, are we in general agreement that we should exercise the process and try to target a next release for SC'16? If so, I worked backwards from SC to put together a high-level schedule. The key bit here is that we would presumably want to advertise the existence of the component selection process, and give the community a small bit of time (11 days in this example) to make submissions. We would have to turn around the initial reviews quickly and then have roughly 1 month to finalize the release. (yes, this is a short turn around time, motivated by a desire to be able to highlight the release at SC'16). For completeness, here is a copy of the proposed aggressive schedule from Slide #9: Action, Proposed: Announce component selection process, 9/15/16 First component submission deadline, 9/26/16 Finalize component selection for 1.2 branch, 10/4/16 Target 1.2 Release, 11/4/16 SC'16 starts on November 14, 2016 (2) if there is general agreement on (1), we need to publicly highlight the process. Slide #5 in the attached has a starting draft for this. For editing/commenting purposes it is in a google doc at: https://docs.google.com/document/d/1CAwIhJ3ZS7Iai2H46n26rXuemYFQ20XHmkfiyIN 8xPU/edit?usp=sharing I'd like to finalize this overview and post on a GitHub wiki page this week (proposed to do this on Thursday; thanks to Craig already for his edits) (3) To take formal submissions and finalize the overview in #2, we also need to finalize the underlying submission infrastructure. Right before the call dropped earlier today, I was mentioning that the issue template mechanism in GitHub is useful in this regard, but it appears to apply to all issues in a given repository. This is not really ideal for us as we really only want this for a subset of issues (ie. new submissions). On Slide #6 in the attached, I laid out what I think are two reasonably quick options to move forward: * (3a) create web-form that creates Github ticket with our desired features in existing OpenHPC repo * (3b) create a dedicated component selection repository under the top-level OpenHPC Github project and use an issue template to customize the process My personal leaning at the moment is towards (3b) and having a dedicated submission repository that we use to manage all the submissions. This would allow us to have all the overview information sitting right up front when a user visits the submission repo and any issue they create will be customized to request the info we want. If you have comments/suggestions for the above, please chime in and hopefully we can work thru email channels this week. Thanks, - On 9/13/16, 10:18 AM, "Schulz, Karl W" <karl.w.schulz@...> wrote:
Folks, |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Re: Apologies
Schulz, Karl W <karl.w.schulz@...>
On 9/13/16, 10:57 AM, "Schulz, Karl W" <karl.w.schulz@...> wrote:
Hello TSC, Following up from yesterday's thread and the component submission infrastructure in particular, I have created a prototype submissions repository on GitHub. This approach corresponds to option 3b above. This repo is private at the moment, but if we like it, we can flip the switch to use this as the starting submission site. In this case, the top-level submissions URL would be: https://github.com/openhpc/submissions Since it's currently a private repo, you won't see anything just yet, but I am attaching two relevant screenshots. The first screenshot (submissions-overview) is the welcome page that contains slightly modified text based on what was posted in the google doc yesterday. It includes links to an example issue that shows all the submission elements, and also has a direct link to create a new request (using the template associated with this repository). The second screenshot (submission-issue) is an example submission issue highlighting all the requirements. This is the template submitters will be asked to adhere to and contains the elements we've discussed in previous TSC meetings. I think this gives us a reasonable starting path to organize future submissions. Please advise if you have additional suggestions/feedback on the approach. Note that I've asked Jeff to secure submissions@... to have a receiving location for submitters who want to optionally provide their email. If there are no strong objections, I would like to propose announcing this submission site to the community email lists this week. I had originally proposed tomorrow (9/15), but how about Friday (9/16) to give a little more time to review? If folks have strong objections to going forward with this approach, please let me know and I'll hold off and we can queue up more discussion at next week's meeting. Thanks, -k |
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Re: Apologies
Jeff ErnstFriedman <jernstfriedman@...>
I would also like to add a link to the Website and once it is appropriate invite the members of the Outreach committee to share the content through their various channels to build awareness. I can delay the "announcement" for a few days if we want to do a soft launch of Karl's suggestion so that we can work out any potential kinks. Thank you, Jeff ErnstFriedman 2201 Broadway #604, Oakland, CA 94612 mobile: 510.593.1367 skype: jeffrey.ernstfriedman twitter: @namdeirf On Wed, Sep 14, 2016 at 10:41 AM, Schulz, Karl W via OpenHPC-TSC <openhpc-tsc@...> wrote:
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Re: Apologies
Schulz, Karl W <karl.w.schulz@...>
Hello TSC,
toggle quoted message
Show quoted text
Many thanks to those who had a chance to look at the submission screenshots earlier in the week to give their go ahead on the approach. Thus far, I haven't received any strong objections and have just toggled the new repository to be publicly accessible. If anyone wants to take a look, go to: https://github.com/openhpc/submissions I will hold off until the end of day today (5pm US/central) before posting an announcement to the mailing lists that the submission site is open. Of course, if you have strong objections, please let me know and we can postpone to have more discussion at next week's meeting. Thanks, -k On 9/14/16, 12:41 PM, "Schulz, Karl W" <karl.w.schulz@...> wrote:
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
Conference Invite from Jeff ErnstFriedman: OpenHPC TSC meeting (Weekly) on Tuesday September 20, 2016 @ 8:00 am (PDT)
UberConference <noreply@...>
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
Try our other product,