On Tue, Apr 29, 2008 at 11:42 PM, Chris Hostetter <[EMAIL PROTECTED]> wrote: > > : We can probably do away with hard-links if a core swap (rename) can be made > : to work without downtime. > > One thing to keep in mind if you rely on core swapping as an integral part > of replication is that you won't be able to take advantage of > "newSearcher" events or autowarming of caches (because the cores will be > independent of eachother) .. an equivilent could probably be written to > warm a new core using info from another core with similar cache > configurations, but it's not something that exists yet. > > > > > * getFileList . Get the names of index files and their checksums. > > (NamedList response) > > * getFilePart: for 1...n of configured chunk size (simple binary > > output/http) > > you'll also want some "snapshooter" equivilent on the master to ensure an > isolated named copy of the index that is garunteed to never change no > matter how long hte client takes to copy all the parts.
The intend is to make the java replication strategy play well with the current snapshooter/snappuller/snapinstaller , so that the current users may not observe anything different. All the files/directories created on master and slave must be same. There should be a java counterpart for everything that the current scripts do. > > > > > > -- > Regards, > Shalin Shekhar Mangar. > > > > > -Hoss > >
