Re: Packing, again
Toby Dickenson <[email protected]>
| Newsgroups | gmane.comp.web.zope.zodb.dirstorage |
|---|---|
| Message-ID | <[email protected]> |
On Thursday 02 Sep 2004 04:35, Richard Waid wrote: > Preparing local storage.... > Fetching increment.... > ReplicaError: storage has been packed since the last replica > ( 2004-05-26 13:02:55 GMT > 2004-05-26 11:10:20 GMT ) > ReplicaError: Error from remote replica Thats strange.... are your clocks wrong? this is saying that the master storage was most recently packed at 2004-05-26 13:02:55, and the most recent transaction on the replica is from 2004-05-26 11:10:20 GMT. The normal solution to this problem is to keep 'enough' days of history on the master. when packing, dont choose "0 days". > If it just picks up where it leaves off ... wouldn't that mean that > you'd have to pack the replica separately anyway? yes > (in which case, the rsync might simply be quicker anyway). Maybe, unless your rsync is over a low bandwidth link. Your solution has less redundancy..... Suppose something happens to corrupt one old file. Your solution will copy the corrupt version onto the replica at the next rsync operation, and (I assume) delete the only good one. If you only replicate using the replica tool, then all data on the replica is sure to be the authentic version from the first replication event after the data was first created. Also beware that a naive implementation of your tar+rsync+untar solution will not be atomic. If something crashes half way through your untar operation then your replica will not be self-consistent. If you have scripts that defend against this type of problem then I would be keen to include them in the DirectoryStorage distribution. ------------------------------------------------------- This SF.Net email is sponsored by BEA Weblogic Workshop FREE Java Enterprise J2EE developer tools! Get your free copy of BEA WebLogic Workshop 8.1 today. http://ads.osdn.com/?ad_id=5047&alloc_id=10808&op=click