On 30 March 2012 11:01, Mehdi El Fadil <[email protected]> wrote: > Hi, > > I have left an application populating couchdb with tweets on a dev server > running for a few days. > > At one point there were 1.7 million documents, and the database was using > approx. 8GB. One day later there were 1.71 million documents, but the DB > had filled the whole disk (20 GB). This looks strange, only 100 000 new > docs, using more space than the 1.7 million that were previously there. All > the documents were created by the same script. > > What happened during this day was insertion of documents, one replication > (before the disk was filled), few get _all_docs queries, and I doubt that > any view has been refreshed. > > There has not been any update running on the docs. Logs show errors related > to the no space left on device, but nothing before that. > > I have copied the 20GB couchdb database file to a larger disk to get able > to trigger a replication, and the replicated version only uses a bit more > than 8 GB. > > Any explanation of the cause of the huge jump in space usage by the > database? > > couchdb version is 1.1.0 is I remember well. > > Thanks. > -- > * > Mehdi El Fadil <http://be.linkedin.com/in/elfadme> > <http://mango-is.com> > *
Hi Mehdi, 1.2.0 will have support for auto-compaction when released. Depending on what _id you choose, adding documents may require re-writing internal b~tree nodes. This can increase space which is then reclaimed post compaction or replication. Ideally your ids are ascending to minimise this, mail back a few and we'll see. A+ Dave
