+1 Noble. I think we should be careful about using the word snap or
snapshot.
The reason I say its confusing is that: at the top of the wiki, it said
"no snapshots" - but indeed there were snapshots, just not in the same
sense as before. Shalin also mentioned that
this new replication method doesnt need to take a snapshot to replicate,
but in my mind, the commit point is a snapshot of the index (terminology
thats been used in the Lucene world). The only difference from the old
scripts replication snapshot is that you dont copy it out to a different
folder with this new method - to me, its a snapshot maintained in the
same folder.
So in my mind, snapshot is just so overloaded, and almost more confusing
than helpful.
Its also not very clear that you don't need to call snappull unless you
want backups. Its also not clear where those backups will go (to someone
reading the wiki), or how they are managed, or rather you can replicate
from a backup
rather than the live index, etc. Its also not very clear how
"enable/disable" replicate relate to "enable/disable" polling. I know
because I saw the email about it, but a user reading the wiki would have
trouble. Its also not clear that replication defaults to on as well as
polling (I think?).
It might also talk about what happens if you shutdown during
replication. Will Solr wait for the replication to finish? Will it be
cut off? If its cut off, are the temp files cleaned up later?
Not to be over critical, or come down all at once, I was just going over
the page with someone trying to learn it recently and these are the type
of issues that came up. Just some suggestions of what we could address,
but I'm not nearly as familiar with it as you guys are.
- Mark
Noble Paul നോബിള് नोब्ळ् wrote:
how about renaming the 'snappull' command to 'synchronize' or 'sync'
this is the only place where it uses the term 'snap'
On Wed, Jun 10, 2009 at 11:37 PM, Mark Miller<[email protected]> wrote:
Shalin Shekhar Mangar wrote:
2009/6/10 Noble Paul നോബിള് नोब्ळ् <[email protected]>
<str name="snapshot">startup</str>
<str name="snapshot">commit</str>
Please note that snapshots are needed only if you want backups.
Replication
does not need to create snapshots.
The wiki is quite confusing on this.
Doesnt it talk about the slave pulling a snapshot (or snap pull)? What does
it pull if you don't create a snapshot?
I think the terminology around snapshot needs to be cleaned up.
--
- Mark
http://www.lucidimagination.com
--
- Mark
http://www.lucidimagination.com
