Harsh, IMHO, I don't see the need of doing any sort of links over name spaces or containers. If you want to have data that spans containers, you copy the data to the container, not link to it.
That's why I have a hard time understanding why someone would want to hard link over a namespace. I'm trying to understand an argument made against HDFS-3370. Thx -Mike On May 16, 2013, at 12:14 AM, Harsh J <[email protected]> wrote: > Do you see viewfs mounts coming useful there (i.e. in place of > hardlinks across NSes)? > > On Thu, May 16, 2013 at 3:49 AM, Michael Segel > <[email protected]> wrote: >> Actually creating links, symbolic or hard links makes sense in a couple of >> scenarios. >> Especially in terms of hive... ;-) >> >> So it kind of goes back to my extension of the question about that Jira >> (HDFS-3370) to see if its alive or just forgotten? >> >> The point is that one of the arguments against doing it didn't make sense. >> Creating hard links across Name Spaces. >> IMHO you'd want to create hard links within the same NN. Maybe a symbolic >> link across name spaces, but even then, I'm not so sure... still need to >> think more about the problem. >> >> On May 15, 2013, at 1:30 PM, Harsh J <[email protected]> wrote: >> >>> Namespace divides are designed with application-level separation in >>> mind. Sharing a file across namespaces does not make a whole lot of >>> sense to me. >>> >>> Anyhow, the data is on the same set of DNs, and there's HA for NN's >>> own availability (if thats really a concern), so I don't see why >>> anyone would like to _maintain_ two synced copies of files as thats >>> just data duplication when all you need is a simple path (viewfs)/URI >>> (hdfs) to access a file lying on a different NN. >>> >>> The reason you mention of metadata availability doesn't sound logical >>> - in such a case a person has to build a self failover of URIs for >>> said file, which they can simply avoid by using HDFS HA for the >>> hosting NN. >>> >>> On Wed, May 15, 2013 at 7:47 PM, Michael Segel >>> <[email protected]> wrote: >>>> Quick question... >>>> So when we have a cluster which has multiple namespaces (multiple name >>>> nodes) , why would you have a file in two different namespaces? >>>> >>>> >>> >>> >>> >>> -- >>> Harsh J >>> >> > > > > -- > Harsh J >
