[j-nsp] T4000 power architecture

Tom Storey tom at snnap.net
Sat Sep 27 18:41:26 EDT 2014


Found an ebay listing with some pictures of a T1600 power supply
(T640/T1600/T4000, its all the same chassis so the pinout should be
the same) that also includes a description of each pin on the power
supply...

http://www.potomacescrap.com/ebay/images/STORE-5409-0005.JPG
http://www.potomacescrap.com/ebay/images/STORE-5409-0003.JPG

Seems there are some pins on one connector that relate to each FPC
(enable pins, perhaps detects when an FPC is inserted?) and there are
-48 outputs from 0 to 8 (FPC 0-7 + 1 more for ____?)

Perhaps the power supply detects when an FPC is inserted and enables
the related output, or all of the outputs are enabled regardless but
individually traced to each FPC slot rather than sharing a common
trace on the backplane. It is possible to configure on or off the
power supply to individual FPC slots, so it seems plausible that each
power supply has individually switchable power outlets?

Though TBH after nearly 3 weeks I'd just do the maintenance and see if
swapping the power supply is simple enough to fix the problem. If you
havent got an answer from Juniper by now, how many more weeks are you
potentially going to wait before you either cant turn up more
capacity, or a power feed failure takes down your FPC and the capacity
on it? :-)

On 25 September 2014 00:21, Sam Silvester <sam.silvester at gmail.com> wrote:
> Hi Fred,
>
> Thanks for the reply. Your suggestions from step 1 - 3.2 makes sense.
>
> Having said that, rather than swapping the PEMs in step 4, I can just swap
> PEM1 with a replacement (we have support on the chassis). The difficulty
> for me is we have to get electricians onsite out of hours to do this -
> which I'm trying to avoid if it's unlikely the PEM is actually at fault.
> This would be the case if the PEM output is carried to the FPCs via a
> common bus arrangement, at which point I would suggest there is no way the
> PEM can be at fault considering all the other FPCs DO show power being
> received from PEM1. If I can confirm (via JTAC or otherwise) what the power
> architecture is, then we can save this step, bite the bullet and move on
> with getting the chassis replaced without wasting time on replacing the
> PEM, then having to get a second window booked with electricians etc to
> swap out the whole chassis.
>
> I suppose ultimately part of the problem here has been that my coworkers
> and myself would have assumed back in July when we first lodged this issue
> to JTAC that asking a relatively straightforward question ("how is the
> power distributed from the PEMs to the FPCs in a T4000?") that we would
> have actually had a fairly clear answer without too much fuss.
>
> To date, that hasn't happened - we can't work out why this is; is it that
> nobody runs T4000s? Is it that nobody in Juniper has this information? Is
> it 'commercially sensitive' enough that nobody wants to tell us?
>
> The response from JTAC so far has basically been hand-waving and avoiding
> the question. That doesn't give me much confidence in either the platform
> or JTAC.
>
> Cheers,
>
> Sam
>
>
>
> On Thu, Sep 25, 2014 at 2:33 AM, Fred Quan <fquan at juniper.net> wrote:
>
>> Hi Sam,
>>
>> Maybe you can try the below steps to isolate the problem.
>>
>> 1.  run "show chassis environment pem" to check which FPC has 0 Voltage
>> and Current on PEM.
>>         ( Here, I assume that,  FPC 0 has 0 Voltage and Current on PEM 1 )
>> 2. check FPC 0 is Online state with Dual PEM power on.
>> 3. power off PEM #0 in slot 0, then check FPC 0 state.
>> 3.1 if FPC 0 is still Online state with single PEM #1 power on, then PEM
>> #1 in slot 1 can provide power to FPC 0, it is a display problem in STEP 1.
>> 3.2 if FPC 0 is Offline state with single PEM #1 power on, then PEM #1 in
>> slot 1 cannot provide power to FPC 0, then continue STEP 4.
>> 4. power off both PEMs and swap PEM #0 and PEM #1, so PEM #0 in slot 1 and
>> PEM #1 in slot 0.
>> 5. power on PEM #1 in slot 0 only. It looks like the PEM #1 problem if FPC
>> 0 is Offline state.
>> 6. power on PEM #0 in slot 1 only. It looks like the Mid plane problem if
>> FPC 0 is Offline state.
>>
>>
>> Regards,
>>
>> -Fred
>>
>>
>> ________________________________________
>> From: juniper-nsp <juniper-nsp-bounces at puck.nether.net> on behalf of Sam
>> Silvester <sam.silvester at gmail.com>
>> Sent: Tuesday, September 23, 2014 9:40 PM
>> To: juniper-nsp at puck.nether.net
>> Subject: Re: [j-nsp] T4000 power architecture
>>
>> Bumping for good measure...
>>
>> So JTAC are suggesting it's either the PEM or the midplane, but I was
>> hoping to get some more information so I can validate what I'm being told -
>> which, currently, is very little. Basically the suggestion is "replace the
>> PEM, if that doesn't work replace the chassis". I'd like to avoid
>> whack-a-mole if possible; for example, if I knew that all of the FPCs were
>> fed via the PEM via a central bus on the midplane, then we don't have to
>> worry about replacing the PEM and can go straight ahead with organising a
>> chassis swap.
>>
>> I don't suppose somebody happens to have a spare T4000 chassis or PEM lying
>> around and could take a photo of the connectors between the PEM and the
>> midplane? I'm still trying to get an idea of how power is distributed from
>> the PEM to the actual FPC slots.
>>
>> The exact same questions has been posed to JTAC multiple times, but no
>> answer has been forthcoming. I really hope the nagging suspicion I have
>> that nobody in Juniper knows this information is incorrect but in the
>> meantime I'm posting here in the hope I can figure it out from some
>> pictures.
>>
>> Google image search isn't helping me find any clear photos of the rear of
>> the PEM or the midplane :/
>>
>> Sam
>>
>>
>> On Thu, Sep 11, 2014 at 8:06 AM, Sam Silvester <sam.silvester at gmail.com>
>> wrote:
>>
>> > Hi Aqeel - thanks for the reply.
>> >
>> > Agree 100% - the problem is, we only seem to be getting power to FPC0
>> from
>> > one PEM... have a look below (and also note that FPC1 is fine, as are the
>> > rest of the FPCs).
>> >
>> > The question is - are we looking at a PEM fault here, or a midplane
>> fault?
>> > As per previous discussion, the FPC itself seems fine, as moving it to
>> > another slot resolves the issue. Putting another card into slot 0 yields
>> > the same result as below.
>> >
>> > PEM 0 status:
>> >   State                      Online
>> >   Temperature                32 degrees C / 89 degrees F
>> >   DC Input:                  OK
>> >                       Voltage(V)  Current(A)  Power(W)  Load(%)
>> >       INPUT 0           54.750       4.312      236        9
>> >       INPUT 1           54.500       6.000      327       13
>> >       INPUT 2           54.625      11.750      641       26
>> >       INPUT 3           54.750       6.125      335       13
>> >       INPUT 4           54.250      10.500      569       23
>> >       INPUT 5           54.500       6.062      330       13
>> >   DC Output           Voltage(V)  Current(A)  Power(W)  Load(%)
>> >       FPC 0             55.062       8.625      474       31
>> >       FPC 1             55.250       4.062      224       14
>> >
>> > <...snip...>
>> >
>> > PEM 1 status:
>> >   State                      Online
>> >   Temperature                30 degrees C / 86 degrees F
>> >   DC Input:                  OK
>> >                       Voltage(V)  Current(A)  Power(W)  Load(%)
>> >       INPUT 0           54.500       3.250      177        7
>> >       INPUT 1           54.625       3.375      184        7
>> >       INPUT 2           54.500      12.437      677       28
>> >       INPUT 3           54.500       5.125      279       11
>> >       INPUT 4           54.625      12.062      658       27
>> >       INPUT 5           54.375       2.750      149        6
>> >   DC Output           Voltage(V)  Current(A)  Power(W)  Load(%)
>> >       FPC 0              0.000       0.000        0        0
>> >       FPC 1             55.125       4.500      248       16
>> >
>> > <...snip...>
>> >
>> >
>> >
>> > On Wed, Sep 10, 2014 at 8:48 PM, aqeel ahmed <aqeelpk at yahoo.com> wrote:
>> >
>> >> Hi
>> >>
>> >> Though aimed for redundancy If system has both power supplies installed
>> >> then it will automatically load balance and in case one power supply
>> goes
>> >> down then whole system will be on single power supply left working.
>> >>
>> >> For further details you can refer to following juniper document.
>> >>
>> >>
>> >>
>> http://www.juniper.net/techpubs/en_US/release-independent/junos/information-products/topic-collections/hardware/t-series/t4000/hwguide/t4000-hwguide.pdf
>> >>
>> >> Regards
>> >>
>> >>
>> >>
>> >>   On Wednesday, September 10, 2014 5:16 AM, Sam Silvester <
>> >> sam.silvester at gmail.com> wrote:
>> >>
>> >>
>> >> Howdy,
>> >>
>> >> Can anybody shed any light on how the PEMs on a T4000 actually
>> distribute
>> >> power to each FPC slot?
>> >>
>> >> Have the case of a single FPC slot that is showing power being received
>> >> from only one of the PEMs, whilst all the other FPC slots are load
>> sharing
>> >> as expected.
>> >>
>> >> Replacing the FPC shows the same issue, so we're pretty happy that it's
>> >> "slot specific".
>> >>
>> >> What I'm curious about is if the midplane has individual 'traces' (for
>> >> lack
>> >> of a better term) for supplying power to each FPC from the two PEMs, or
>> if
>> >> there is a common bus shared between all the FPCs from each PEM. The
>> >> reason
>> >> I ask is if the PEM only has a single connection to the midplane,
>> >> replacing
>> >> it seems pointless and instead it looks like we're better off replacing
>> >> the
>> >> midplane. If the PEM has individual outputs to each slot, then replacing
>> >> the PEM seems like a reasonable approach.
>> >>
>> >> I've been pointed at the following document (
>> >>
>> >>
>> http://www.juniper.net/documentation/en_US/release-independent/junos/topics/concept/power-supply-t4000-description.html
>> >> )
>> >> which is very light-on in terms of detail. Does anybody know if there
>> is a
>> >> more detailed document available (or even internally?) that we can ask
>> >> about?
>> >>
>> >> Thanks!
>> >> _______________________________________________
>> >> juniper-nsp mailing list juniper-nsp at puck.nether.net
>> >> https://puck.nether.net/mailman/listinfo/juniper-nsp
>> >>
>> >>
>> >>
>> >
>> _______________________________________________
>> juniper-nsp mailing list juniper-nsp at puck.nether.net
>> https://puck.nether.net/mailman/listinfo/juniper-nsp
>>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp


More information about the juniper-nsp mailing list