[Go-essp-tech] verification of datanode status
Eric Nienhouse
ejn at ucar.edu
Mon Dec 19 15:02:48 MST 2011
Hi All,
A long standing Gateway requirement is to provide search and discovery
of datasets (and other metadata) regardless of the state of remote
services. In the event that a data node service is unavailable, users
should still be able to identify datasets, determine what has been
published and generate download scripts.
This guiding requirement was discussed at great length by the ESG-CET
project group and was accepted as a key element in support of the
community's best interest for data discovery. Identifying "what has
been published" was a key use case driving this need. This advantage of
this approach is that it allows users to find out "what exists" during
periods of unexpected downtime or other service unavailability.
Henrik noted that the DMI data node is no longer serving datasets
publishing into ESG. In this case these datasets can be discovered at
the Gateway, however, they are inaccessible and out of sync. If these
data are no longer meant to be accessed, I'd suggest they be "retracted"
from the gateway and they will no longer appear in the search results.
Thanks,
-Eric
Estanislao Gonzalez wrote:
> T(sorry the message got cut)
> ...it's up to the publisher to define when data shouldn't be accessible
> anymore.
>
> There are some improvements that can be doen, but most I can think of
> will make the understanding of the system more complex to the end user.
>
> Datanode admin should rely on tools that help them get their nodes up
> for as long as possible (nagios & Co).
>
> My 2c,
> Estani
> On 19.12.2011 08:04, Estanislao Gonzalez wrote:
>
>> Hi Luca,
>>
>> That's not what Henrik meant. Neither the architecture retains a
>> living
>> link to a data nose (not an index one, as you've pointed out)
>>
>> This is a feature IMO as the search engine is detached from the
>> dataone. The index might indeed "prune" the data nodes down, but
>> unless
>> this is done synchroneusly it would difficult the federation
>> interaction.
>> Or to say it differently: Is up to the data node to assure data is
>> available, and if that's not desired anymore, On 19.12.2011 07:15,
>> Cinquini, Luca (3880) wrote:
>>
>>> Hi Henirik,
>>> not sure about the gateway, but this is a feature the P2P system
>>> does have: has soon as a datanode is inaccessible, the search
>>> automatically prunes that node away, so the search results never
>>> contain dead links.
>>> thanks, Luca
>>>
>>> On Dec 19, 2011, at 3:58 AM, Henrik Wiberg wrote:
>>>
>>>
>>>> Does the gateway somehow 'ping' its registered datanodes to verify
>>>> that
>>>> they are accessible? The datanode at dmi has not been running for 5
>>>> mounts still the datanodes published datasets are searchable and
>>>> displayed at the gateway cmip-gw-badc. Should not inaccessible
>>>> datasets
>>>> be removed from the search result?
>>>>
>>>> _______________________________________________
>>>> GO-ESSP-TECH mailing list
>>>> GO-ESSP-TECH at ucar.edu
>>>> http://mailman.ucar.edu/mailman/listinfo/go-essp-tech
>>>>
>>> _______________________________________________
>>> GO-ESSP-TECH mailing list
>>> GO-ESSP-TECH at ucar.edu
>>> http://mailman.ucar.edu/mailman/listinfo/go-essp-tech
>>>
>
>
More information about the GO-ESSP-TECH
mailing list