OK, then you have several things going on here if I'm reading this right.
1> you have very little spare room on your slave index. This will considerably
slow down your disk writes.
2> You are optimizing for each delta index, forcing the entire index to be
replicated each time. This is probably unnecessary, think about optimizing
once a day/week/whatever.
3> Your network is slow, although that may be a function of <1>. have you
tried copying your 80G index to another machine?
4> You will not be able to run your slave with so little extra space
reliably. I said
that replication may take *at least* double the space. Actually,
to be safe you
should have triple.
5> you may be in the range you have to consider sharding.
6> the merge factor is irrelevant on slaves, it's ignored. It only has
an effect on
the master
7> you may want to consider taking steps to reduce the size of your
index. In particular,
examine whether you need to store all the fields you store.
Here's a great place to
start answering that question:
http://lucene.apache.org/java/3_0_2/fileformats.html#file-names
Best
Erick
On Mon, Sep 5, 2011 at 12:52 AM, shinkanze <rajatrastogi.it@gmail.com> wrote:
> erik
> it takes 20 -25 minutes to transver the index::::
>
> we have a setting in solr to replicate on : startup,commit and optimize
>
> rajat
>
> --
> View this message in context: http://lucene.472066.n3.nabble.com/Problem-while-replicating-solr-tp3306228p3309925.html
> Sent from the Solr - User mailing list archive at Nabble.com.
>
|