June 2014 - Admins - mailman3.cels.anl.gov

Re: [JLSE-Admins] Request: JLSE VLAN Availability in 240-4312
by Allen, Benjamin S. 30 Jun '14

30 Jun '14

Sounds good. Feel free to propose a place and time to meet. Ben On Jun 30, 2014, at 4:38 PM, Linda Winkler <winkler(a)mcs.anl.gov> wrote: > Let's review the requirements on Tues and develop a plan. It has been sometime since anyone paid attention to needs in that space. > lw > > > On Jun 30, 2014, at 4:11 PM, "Allen, Benjamin S." <bsallen(a)alcf.anl.gov> wrote: > >> Hello Everyone, >> >> The east part of 240-4312, what was formally named Tinkerlab, has joined JLSE. We would like to put a number of machines housed in that room on a new private VLAN that is also accessible by ftsn1, ftsn2, ftlogin1, and ftlogin2. These servers are behind ALCF’s Donald. >> >> It appears there’s a Juniper switch serving the room. There are also a number of small desktop switches in the tinkerlab area, which ideally we could replace with a single managed switch in a configuration that meets Kaz and others needs. It appears the existing uplink from the Juniper is a 10GbE port, and the second 10GbE port on the Juniper switch is used for the display wall machine. >> >> As an aside I would also suggest we get a number of additional floor tiles cut with matching floor grommets, for each work area in the thinkerlab. Then all network cabling can be run under the floor to eliminate the current tripping hazards. >> >> Please let me know what is required to make this work. >> >> Thanks, >> >> Ben >

1 0

Re: [JLSE-Admins] Request: JLSE VLAN Availability in 240-4312
by Linda Winkler 30 Jun '14

30 Jun '14

Let's review the requirements on Tues and develop a plan. It has been sometime since anyone paid attention to needs in that space. lw On Jun 30, 2014, at 4:11 PM, "Allen, Benjamin S." <bsallen(a)alcf.anl.gov> wrote: > Hello Everyone, > > The east part of 240-4312, what was formally named Tinkerlab, has joined JLSE. We would like to put a number of machines housed in that room on a new private VLAN that is also accessible by ftsn1, ftsn2, ftlogin1, and ftlogin2. These servers are behind ALCF’s Donald. > > It appears there’s a Juniper switch serving the room. There are also a number of small desktop switches in the tinkerlab area, which ideally we could replace with a single managed switch in a configuration that meets Kaz and others needs. It appears the existing uplink from the Juniper is a 10GbE port, and the second 10GbE port on the Juniper switch is used for the display wall machine. > > As an aside I would also suggest we get a number of additional floor tiles cut with matching floor grommets, for each work area in the thinkerlab. Then all network cabling can be run under the floor to eliminate the current tripping hazards. > > Please let me know what is required to make this work. > > Thanks, > > Ben

1 0

[JLSE-Admins] Request: JLSE VLAN Availability in 240-4312
by Allen, Benjamin S. 30 Jun '14

30 Jun '14

Hello Everyone, The east part of 240-4312, what was formally named Tinkerlab, has joined JLSE. We would like to put a number of machines housed in that room on a new private VLAN that is also accessible by ftsn1, ftsn2, ftlogin1, and ftlogin2. These servers are behind ALCF’s Donald. It appears there’s a Juniper switch serving the room. There are also a number of small desktop switches in the tinkerlab area, which ideally we could replace with a single managed switch in a configuration that meets Kaz and others needs. It appears the existing uplink from the Juniper is a 10GbE port, and the second 10GbE port on the Juniper switch is used for the display wall machine. As an aside I would also suggest we get a number of additional floor tiles cut with matching floor grommets, for each work area in the thinkerlab. Then all network cabling can be run under the floor to eliminate the current tripping hazards. Please let me know what is required to make this work. Thanks, Ben

1 0

[JLSE-Admins] Fwd: Quote Request: Cirrascale RM1905D
by Allen, Benjamin S. 30 Jun '14

30 Jun '14

Here is the quote from Cirrascale for two X-Gene based systems in 1U, with a 200GB SSD, 32GB RAM, and a Nvidia K20(x?) per system for $19,284.00. Their lead time is estimated at 6-weeks. For comparison SoftIron is quoting $1950 + shipping for a single server board only, i.e. would require chassis, RAM, SSD, and GPU if they support it. These should be near identical systems. Although I’m guessing Cirrascale is putting more effort into supporting GPUs. Thoughts? Ben Begin forwarded message: > From: Sal Balsamo <Sal.Balsamo(a)cirrascale.com> > Subject: RE: Quote Request: Cirrascale RM1905D > Date: June 27, 2014 at 9:24:15 PM CDT > To: "Allen, Benjamin S." <bsallen(a)alcf.anl.gov> > > Ben, > Attached is the quote you requested, please let me know if you have any questions. > > Regards, > Sal > > -----Original Message----- > From: Allen, Benjamin S. [mailto:[email protected]] > Sent: Friday, June 27, 2014 1:34 PM > To: Sal Balsamo > Subject: Re: Quote Request: Cirrascale RM1905D > > Sal, > > Go ahead and include the K20x's. The only passive cooling GPUs I have on-hand are M2070 and K40. I'm presuming the K40 hasn't been tested or is going to be over the power budget of these systems? > > Thanks, > > Ben > > On Jun 27, 2014, at 3:23 PM, Sal Balsamo <Sal.Balsamo(a)cirrascale.com> wrote: > >> Ben, >> No problem. I'll get to work on that now... that server is available with & without the K20's. Do you prefer this quoted with the k20's? Not entirely sure if an SSD can be substituted, the current dev platform has very specific hardware requirements. Let me confirm. >> >> Thanks, >> Sal >> >> -----Original Message----- >> From: Allen, Benjamin S. [mailto:[email protected]] >> Sent: Friday, June 27, 2014 1:12 PM >> To: Sal Balsamo >> Subject: Re: Quote Request: Cirrascale RM1905D >> >> Hi Sal, >> >> Lets go with #3. Does this option ship with the pair of Tesla K20x? Also would it be possible to put a SSD in-place of each of the spinning drives? Something like an Intel DC S3700 200GB or similar? >> >> Thanks, >> >> Ben >> >> >> On Jun 27, 2014, at 3:00 PM, Sal Balsamo <Sal.Balsamo(a)cirrascale.com> wrote: >> >>> Hi Benjamin, >>> Thank you for the email and for your interest in our ARM development offering. >>> >>> We have three products offered in the 1905 dev platform and I'm just looking for clarification on your preference. >>> 1) Desktop dev kit, without PCIE expansion, comes with 16GB DDR3. Availability is approximately 2 weeks. >>> 2) Rackmount 1U entry dev kit, with PCIE expansion, RAM can be upgrade. Availability is approximately 6 weeks. >>> 3) Same as #2, but its two systems in the same 1U space. >>> >>> Please let me know your preference and I will get you a quote immediately. >>> >>> Regards, >>> Sal Balsamo >>> >>> Manager of Inside Sales & Sales Operations Cirrascale Corporation >>> Mobile: 858-344-0811 >>> >>> -----Original Message----- >>> From: Allen, Benjamin S. [mailto:[email protected]] >>> Sent: Friday, June 27, 2014 11:42 AM >>> To: Sales >>> Subject: Quote Request: Cirrascale RM1905D >>> >>> Cirrascale: >>> >>> Argonne National Laboratory's Leadership Computing Facility (ALCF) is interested in purchasing a Cirrascale RM1905D with 32GB of DDR3-1600 RAM. Could you please quote a single system and provide an estimate on when you are shipping these systems? >>> >>> Kind Regards, >>> >>> >>> >>> >>> Ben Allen >>> >>> HPC Systems Administrator >>> ALCF Operations - Joint Laboratory for Systems Evaluation Argonne >>> National Laboratory >>> Email: bsallen(a)alcf.anl.gov >>> Office: (630) 252-0554 >> > >

1 0

Re: [JLSE-Admins] Need to install a new IBM XL compiler
by Allen, Benjamin S. 27 Jun '14

27 Jun '14

Hi Junchao, Yea I can do that. XL is a bit of a pain to get installed in a non-standard location, i.e. /soft/compilers, but I should be able to get it installed by Monday or Tuesday. Ben On Jun 27, 2014, at 1:42 PM, Junchao Zhang <jczhang(a)mcs.anl.gov> wrote: > Hello, > I am a developer from the ANL MPICH group. I want to test a new MPICH feature that requires a new IBM xl compiler at http://www.ibm.com/developerworks/downloads/r/xlfortranlinux/. > This trial version needs root permission to install. I asked Mira support and they referred me to JLSE. Could you install it on the Power machine at JLSE, i.e, Dowd.ftm.alcf.anl.gov? > Thanks a lot. > --Junchao Zhang > _______________________________________________ > Admins mailing list > Admins(a)lists.jlse.anl.gov > https://lists.jlse.anl.gov/mailman/listinfo/admins

1 0

[JLSE-Admins] Need to install a new IBM XL compiler
by Junchao Zhang 27 Jun '14

27 Jun '14

Hello, I am a developer from the ANL MPICH group. I want to test a new MPICH feature that requires a new IBM xl compiler at http://www.ibm.com/developerworks/downloads/r/xlfortranlinux/. This trial version needs root permission to install. I asked Mira support and they referred me to JLSE. Could you install it on the Power machine at JLSE, i.e, Dowd.ftm.alcf.anl.gov? Thanks a lot. --Junchao Zhang

1 0

[JLSE-Admins] Automatic reply: JLSE Resource Access Granted
by Stevens, Rick L. 27 Jun '14

27 Jun '14

Hi, I'm currently out of the office and will return July 14th. During this time and especially from June 27th until July 7th I will not have reliable access to e-mail and responses may be considerably delayed. If you need immediate assistance, please contact Julie Smagacz at smagacz(a)anl.gov or 630-252-3378. Cheers --Rick

1 0

Re: [JLSE-Admins] login problems
by Edward May 26 Jun '14

26 Jun '14

OK thanks. Yes the old emay home directory can be removed. Ed On Thu, 26 Jun 2014, Allen, Benjamin S. wrote: > Ed, > > Let me know if I can remove your old ï¿½emay" home directory. > > No reason other than migration pains in relation to cyrptocard auth for sudo access to match your JLSE account to ALCF. > > NDA = Non-disclosure agreement. We often get early access to hardware before vendors announcement it. Thus requiring ANL to agree not to tell the world about it. It also makes the information related to that hardware Official Use Only (OUO) as far as ANL is concerned, and as a result we make it a bit harder to get to. > > Ben > > On Jun 26, 2014, at 11:31 AM, Edward May <may(a)anl.gov> wrote: > >> Thanks! >> So I infer the may account only uses SSH public key and not simple passphrass for login to >> JLSE machines. OK for me. may is a old MCS account (they have a long memory...) Historical >> note, I was an early user of Rusty Lusk's P4 parallel programming system! I can not imagine >> needing sudo. What is "NDA'ed hardware"? Luckly my files under the emay account are world >> readable so I copyed to may account. Unless there is some bearicartic reason to regularising >> or reconciling to a single account for MCS/ALCF... I am happy. >> >> Ed >> >> >> On Wed, 25 Jun 2014, Allen, Benjamin S. wrote: >> >>> Hi Ed, >>> >>> Sorry for the confusion. >>> >>> Since youï¿½re using ï¿½mayï¿½ instead of ï¿½emay", your ALCF cryptocard wonï¿½t work. The radius authentication system is naively matching your cryptocard to your ALCF username only. At this point the only reason you would actually need your cryptocard in JLSE is privileged access (i.e. sudo), or access to future NDAï¿½ed hardware. >>> >>> The passphrase you setup at https://accounts.jlse.anl.gov is currently only used for that website. In the near future itï¿½ll be used for wiki authentication as well. >>> >>> You can use SSH public key authentication to login.jlse.anl.gov, which I assume youï¿½ve figured out since I see you're logged on. >>> >>> Let me know if you run into any further issues, >>> >>> Ben >>> >>> On Jun 24, 2014, at 4:37 PM, Edward May <may(a)anl.gov> wrote: >>> >>>> I am confused over the new access method to what was previously >>>> ft.alcf.anl.gov which was working with my ALCF account emay using the cryptocard >>>> and now is not! >>>> >>>> I have initiated an account at the webpage https://accounts.jlse.anl.gov/index.php?tpl=index.tpl >>>> for id may ... note the difference with the ALCF account! >>>> I am not able to login to ssh may(a)login.jlse.anl.gov using the phasephase at https://accounts.jlse.anl.gov/ for id may >>>> >>>> I receive: Received disconnect from 140.221.69.21: 2: Too many authentication failures for may >>>> >>>> The discussion email from Ft-discuss(a)lists.alcf.anl.gov intimate that I should use >>>> a public key, but I am confused by the website https://accounts.jlse.anl.gov as to >>>> how to proceed. >>>> >>>> Ed >>>> >>>> >>>> ______________________________________________________________ >>>> Edward N. May Phone: (630)-252-6222 >>>> High Energy Physics Division FAX: (630)-252-5076 >>>> Argonne National Laboratory Internet: may(a)anl.gov >>>> Argonne, IL 60439 USA Webpage: http://www.hep.anl.gov/may >>>> _______________________________________________ >>>> Admins mailing list >>>> Admins(a)lists.jlse.anl.gov >>>> https://lists.jlse.anl.gov/mailman/listinfo/admins >>> >>> >> >> ______________________________________________________________ >> Edward N. May Phone: (630)-252-6222 >> High Energy Physics Division FAX: (630)-252-5076 >> Argonne National Laboratory Internet: may(a)anl.gov >> Argonne, IL 60439 USA Webpage: http://www.hep.anl.gov/may > > ______________________________________________________________ Edward N. May Phone: (630)-252-6222 High Energy Physics Division FAX: (630)-252-5076 Argonne National Laboratory Internet: may(a)anl.gov Argonne, IL 60439 USA Webpage: http://www.hep.anl.gov/may

1 0

Re: [JLSE-Admins] login problems
by Allen, Benjamin S. 26 Jun '14

26 Jun '14

Ed, Let me know if I can remove your old “emay" home directory. No reason other than migration pains in relation to cyrptocard auth for sudo access to match your JLSE account to ALCF. NDA = Non-disclosure agreement. We often get early access to hardware before vendors announcement it. Thus requiring ANL to agree not to tell the world about it. It also makes the information related to that hardware Official Use Only (OUO) as far as ANL is concerned, and as a result we make it a bit harder to get to. Ben On Jun 26, 2014, at 11:31 AM, Edward May <may(a)anl.gov> wrote: > Thanks! > So I infer the may account only uses SSH public key and not simple passphrass for login to > JLSE machines. OK for me. may is a old MCS account (they have a long memory...) Historical > note, I was an early user of Rusty Lusk's P4 parallel programming system! I can not imagine > needing sudo. What is "NDA'ed hardware"? Luckly my files under the emay account are world > readable so I copyed to may account. Unless there is some bearicartic reason to regularising > or reconciling to a single account for MCS/ALCF... I am happy. > > Ed > > > On Wed, 25 Jun 2014, Allen, Benjamin S. wrote: > >> Hi Ed, >> >> Sorry for the confusion. >> >> Since you’re using “may” instead of “emay", your ALCF cryptocard won’t work. The radius authentication system is naively matching your cryptocard to your ALCF username only. At this point the only reason you would actually need your cryptocard in JLSE is privileged access (i.e. sudo), or access to future NDA’ed hardware. >> >> The passphrase you setup at https://accounts.jlse.anl.gov is currently only used for that website. In the near future it’ll be used for wiki authentication as well. >> >> You can use SSH public key authentication to login.jlse.anl.gov, which I assume you’ve figured out since I see you're logged on. >> >> Let me know if you run into any further issues, >> >> Ben >> >> On Jun 24, 2014, at 4:37 PM, Edward May <may(a)anl.gov> wrote: >> >>> I am confused over the new access method to what was previously >>> ft.alcf.anl.gov which was working with my ALCF account emay using the cryptocard >>> and now is not! >>> >>> I have initiated an account at the webpage https://accounts.jlse.anl.gov/index.php?tpl=index.tpl >>> for id may ... note the difference with the ALCF account! >>> I am not able to login to ssh may(a)login.jlse.anl.gov using the phasephase at https://accounts.jlse.anl.gov/ for id may >>> >>> I receive: Received disconnect from 140.221.69.21: 2: Too many authentication failures for may >>> >>> The discussion email from Ft-discuss(a)lists.alcf.anl.gov intimate that I should use >>> a public key, but I am confused by the website https://accounts.jlse.anl.gov as to >>> how to proceed. >>> >>> Ed >>> >>> >>> ______________________________________________________________ >>> Edward N. May Phone: (630)-252-6222 >>> High Energy Physics Division FAX: (630)-252-5076 >>> Argonne National Laboratory Internet: may(a)anl.gov >>> Argonne, IL 60439 USA Webpage: http://www.hep.anl.gov/may >>> _______________________________________________ >>> Admins mailing list >>> Admins(a)lists.jlse.anl.gov >>> https://lists.jlse.anl.gov/mailman/listinfo/admins >> >> > > ______________________________________________________________ > Edward N. May Phone: (630)-252-6222 > High Energy Physics Division FAX: (630)-252-5076 > Argonne National Laboratory Internet: may(a)anl.gov > Argonne, IL 60439 USA Webpage: http://www.hep.anl.gov/may

1 0

Re: [JLSE-Admins] login problems
by Edward May 26 Jun '14

26 Jun '14

Thanks! So I infer the may account only uses SSH public key and not simple passphrass for login to JLSE machines. OK for me. may is a old MCS account (they have a long memory...) Historical note, I was an early user of Rusty Lusk's P4 parallel programming system! I can not imagine needing sudo. What is "NDA'ed hardware"? Luckly my files under the emay account are world readable so I copyed to may account. Unless there is some bearicartic reason to regularising or reconciling to a single account for MCS/ALCF... I am happy. Ed On Wed, 25 Jun 2014, Allen, Benjamin S. wrote: > Hi Ed, > > Sorry for the confusion. > > Since youï¿½re using ï¿½mayï¿½ instead of ï¿½emay", your ALCF cryptocard wonï¿½t work. The radius authentication system is naively matching your cryptocard to your ALCF username only. At this point the only reason you would actually need your cryptocard in JLSE is privileged access (i.e. sudo), or access to future NDAï¿½ed hardware. > > The passphrase you setup at https://accounts.jlse.anl.gov is currently only used for that website. In the near future itï¿½ll be used for wiki authentication as well. > > You can use SSH public key authentication to login.jlse.anl.gov, which I assume youï¿½ve figured out since I see you're logged on. > > Let me know if you run into any further issues, > > Ben > > On Jun 24, 2014, at 4:37 PM, Edward May <may(a)anl.gov> wrote: > >> I am confused over the new access method to what was previously >> ft.alcf.anl.gov which was working with my ALCF account emay using the cryptocard >> and now is not! >> >> I have initiated an account at the webpage https://accounts.jlse.anl.gov/index.php?tpl=index.tpl >> for id may ... note the difference with the ALCF account! >> I am not able to login to ssh may(a)login.jlse.anl.gov using the phasephase at https://accounts.jlse.anl.gov/ for id may >> >> I receive: Received disconnect from 140.221.69.21: 2: Too many authentication failures for may >> >> The discussion email from Ft-discuss(a)lists.alcf.anl.gov intimate that I should use >> a public key, but I am confused by the website https://accounts.jlse.anl.gov as to >> how to proceed. >> >> Ed >> >> >> ______________________________________________________________ >> Edward N. May Phone: (630)-252-6222 >> High Energy Physics Division FAX: (630)-252-5076 >> Argonne National Laboratory Internet: may(a)anl.gov >> Argonne, IL 60439 USA Webpage: http://www.hep.anl.gov/may >> _______________________________________________ >> Admins mailing list >> Admins(a)lists.jlse.anl.gov >> https://lists.jlse.anl.gov/mailman/listinfo/admins > > ______________________________________________________________ Edward N. May Phone: (630)-252-6222 High Energy Physics Division FAX: (630)-252-5076 Argonne National Laboratory Internet: may(a)anl.gov Argonne, IL 60439 USA Webpage: http://www.hep.anl.gov/may

1 0