[torqueusers] Inaccurate memory report in pbsnodes

Cristina Ururahy ururahy at tecgraf.puc-rio.br
Wed Mar 9 11:14:35 MST 2005


Hi,

Sorry for the misunderstanding...

I'm on a project for another company and the problem I am describing happens 
only there and not at my company. All the output I've sent you is **not**
from my computer. That said, when the problem first happened I asked 
them to send me the output from pbsnodes, which was:
bw7mpet:vget > pbsnodes -a bw7n484
totmem=2225476kb,
availmem=4070016kb,
physmem=2323256kb
Notice that physmem is smaller than totmem!

The 6G came from top:
Mem:  2323256K av, 2312728K used,   10528K free,       0K shrd,   26904K buff
Swap: 4096524K av,   37444K used, 4059080K free                 1763244K 
cached
(2323256K + 4096524K = 6419780K)
(6419780K - 2^32 = 2225476K)

They use 2 clusters bw7 and bw5. The problem occurs at the nodes, but "cat 
/proc/meminfo" was executed on the server and not on the node (bw7n484). I 
didn't notice that before. Sorry!

> > I had assumed that what you were saying was that the sum of RAM and swap
> > was 6GB.

That's right. But I still don't get why it shouldn't be a problem since 6G 
won't fit in 2^32...

I'll let you know if the changes from the latest TORQUE-1.2.0p2 snapshot
have fixed that. I am waiting for an answer from them.

Thanks a lot,
Cristina

On Wednesday 09 March 2005 13:15, Philippe Weill wrote:
> Chris Samuel wrote:
> > On Tue, 1 Mar 2005 06:13 am, Cristina Ururahy wrote:
> >>My computer has 6Gb (RAM + SWAP)
>
> this is the answer from christina about memory
>
>  >>and how many physical memory did you have    : more /proc/meminfo
>
> =================================================
> bw5mpet:
>          total:    used:    free:  shared: buffers:  cached:
> Mem:  4215582720 1453342720 2762240000        0 136871936 898838528
> Swap: 8389738496        0 8389738496
>
> bw7mpet:
>          total:    used:    free:  shared: buffers:  cached:
> Mem:  3934429184 1972264960 1962164224        0 44158976 1754578944
> Swap: 4294885376        0 4294885376
>
>
> I don't see 6G on any machine ???
>
> > I'm so sorry, I think that I've misunderstood completely what you wrote
> > here.
> >
> > I had assumed that what you were saying was that the sum of RAM and swap
> > was 6GB and not that you had 6GB of RAM and 6GB of swap for a total of
> > 12GB!
> >
> > Because of my misreading I couldn't understand why you were having this
> > problem when at least one IA32 cluster that I know with 8GB in total (4GB
> > RAM and 4GB swap) was working fine with Redhat Enterprise Linux 3 - hence
> > why I sent you off on what is probably a complete wild goose about GCC.
> > :-(
> >
> > I plead caffeine deficiency, so sorry!..
> >
> > Chris



More information about the torqueusers mailing list