Sounds serious. Is it possibly due to the way that you wrote the hooks? Can you post the hook? Is there code in there that is not transactional? Are you using sql in the hook or a non-transactional graphdb?
On Wednesday, June 8, 2016 at 8:22:59 AM UTC-4, Simon White wrote: > > Affecting 2.1.19 > > We have quite a large database - about a million entries for a particular > vertex. Recently I noticed by chance a duplicate, but surely this is > impossible when there is a unique index? > > I attempted to rebuild the index but unsurprisingly it fails with an > ORecordDuplicatedException. Upon investigation it appears that there are > many many duplicates - hundreds in fact. > > > I am guessing this is a race condition as we insert the entities via both > Java API and HTTP under heavy load (many threads from several separate > applications). > > Possibly unrelated but - when creating one of these entries we run an > onRecordBeforeCreate hook to do some validation and also add a 'date > created' field. This takes milliseconds but under heavy load could it be > enough to throw the index off? > > This is a very serious issue. > > Any hints to avoid the problem? > -- --- You received this message because you are subscribed to the Google Groups "OrientDB" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
