[Ndn-interest] ERROR: std::bad_alloc in repo-ng

Susmit susmit at cs.colostate.edu
Thu Jun 29 10:07:13 PDT 2017


Hi Junxiao,

I can confirm that this is not related to memory. I have upgraded the
machine today and it now has 48GB RAM.

I tried what Alexander said and the crash still happens. I think they
have a deadline coming up and are scrambling to fix this, so any help
is appreciated.

I looked at the log and I am not sure what's causing this either. For
ruling out malformed data, I create a 1M file and published it.

head -c 1M < /dev/urandom > test_file.txt
ndnputfile ndn:/kisti/atmos
/kisti/atmos/hur_Amon_CESM1-CAM5_rcp60_r1i1p1_200601-204912.nc/test
test_file.txt
ndncatchunks /kisti/atmos/hur_Amon_CESM1-CAM5_rcp60_r1i1p1_200601-204912.nc/test

And exactly the same crash happens.

Thanks.


On Thu, Jun 29, 2017 at 10:39 AM, Junxiao Shi
<shijunxiao at email.arizona.edu> wrote:
> Hi Alexander
>
> The process is probably running out of memory. repo-ng keeps the whole index
> in memory, and cannot handle large datasets.
> repo-ng currently does not have a maintainer, and I'm unfamiliar with its
> internals.
>
> Have a try with repo-sql https://github.com/3rd-ndn-hackathon/repo-sql which
> places the index in a PostgreSQL database. It is designed to handle large
> dataset although it's not tested.
> Use repo-sql-insert command to insert. It does not support TCP bulk insert
> protocol or repo-ng protocol.
> All Interest selectors except Exclude are supported.
>
> Yours, Junxiao
>
> On Thu, Jun 29, 2017 at 09:19 Alexander Ni <alexni1992 at gmail.com> wrote:
>>
>> Hello,
>>
>> We moved 0.5.1 version as you recommended.
>> At exact same procedure, we now have a different error.
>> Segmentation fault (core dumped)
>> I attached the gdb backtrace.
>>
>>
>



-- 

====================================
http://www.cs.colostate.edu/~susmit
====================================


More information about the Ndn-interest mailing list