lucene index on shared disk (NFS)

[Al Faller, geändert vor 14 Jahren.]

Hi All -

Is it acceptable to have clustered servers all pointing at the same lucene index on an NFS mounted volume?

Thanks!

Al

[William Fleming, geändert vor 13 Jahren.]

did you get this working? I am in the process of trying to set this up and would be very interested to see if you got it working. Did you move the whole data folder or just the lucene folder?

Thanks

William

[Al Faller, geändert vor 13 Jahren.]

I went a different route. I ended up using clusterlink with an index on each application server.

[Mika Koivisto, geändert vor 13 Jahren.]

This only works if you have only one node that is writing to the index. Other nodes must be read only. This is because lucene on the first node will lock the index and other nodes can't write to it.

[William Fleming, geändert vor 13 Jahren.]

Thanks for the info.

William

[Stephan Huber, geändert vor 13 Jahren.]

Hi Mika,

what happens in a clustered environment when for all of the three servers:

lucene.store.type=jdbc
index.read.only=false

Is it a problem? The DBMS (Oracle) should handle read and write from different servers, isn't it?
How much slower is using the database for lucene instead of a file system?

Thanks in advance,
Stephan

[Mika Koivisto, geändert vor 13 Jahren.]

Sorry for not answering sooner.

With jdbc several servers can write to the index. I haven't done any performance tests agains jdbc index so I don't have any numbers on how much slower it is. If you have a lot of data in the index (documents are also indexed) and you rely on the search then I would recommend using Solr instead. It's a dedicated search server which can be scaled much easier than Lucene by itself.

[Hitoshi Ozawa, geändert vor 13 Jahren.]

I think I was getting errors when I did that. Especially when on server did a recreate index.

Setting only one node to have write priviledge would solve the problem but then, other nodes would not be able to accept any posts including document libray uploads.