Hi, This is the kind of bug that is most annoying to me as a developer, so I know this is a very hard to identify report...
Basically, I have two scenarios, Sedna frozen in both and the problem went away by doing it in a huge transaction instead of doing one atomic operation per transaction. Scenario 1: I have two collections, one with around 7k documents (each one with around 2 kb), and other with around 35k documents (each one with around 2kb as well). I have a batch process that uses this two collections as input, creating a third collection with a unified document (the first collection is "household information", the second is "personal informatino", the third unifies both as "family information"). The resulting collection will be around the same size as the first in number of documents, but the size in bytes will be about the sum (ok, a bit less, because the two first collections have digitilizing control data that we don't transfer here). If I begin and commit one transaction for each "family", sedna will at some point (about 3k iterations) freeze all connections and get to a dead-lock state that will only go away after restarting the governor If I, on the other hand, do all the 7k iterations into a single transaction, it works just fine all to the end. (it might be worth mentioning that as sedna doesn't provide a cursor API, I need to have one connection open for the reads and one for the writes). Scenario 2: The second scenario is about doing some data massage from the collection generated above (some of which can only be done once we have the above collection complete), so this involves reading from the third collection in one connection, have some additional queries for each iteration and then, finally, after modifying the data, saving the result in a fourth collection (I'm doing it incrementaly in order to have checkpoints for the data). The same issue arises, If I begin and commit the transaction in the writing connection for each iteration, it will freeze at some point around 3k transactions. Having it all in a single transaction seems to work fine. End of example... Well, this kind of bug sucks, because it's hard to reproduce it in a controlled environment and i'm not in the liberty to disclose the actual data, so... I'm saying this here in the hopes that it will give any hint on what might be causing the problem. daniel ------------------------------------------------------------------------------ ThinkGeek and WIRED's GeekDad team up for the Ultimate GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the lucky parental unit. See the prize list and enter to win: http://p.sf.net/sfu/thinkgeek-promo _______________________________________________ Sedna-discussion mailing list Sedna-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sedna-discussion
