From: Norbert B. <br...@ch...> - 2007-06-08 19:43:27
|
Hi Michael, Michael Moritz wrote: > My theory to what happens is that there are maybe two similar request - at > least they create same or similar keystr) and then one "overtakes" the other > one and ends up stealing its id. > But that's only a theory really and would need some debugging. For the moment > I'm wondering whether anybody else gets those? > It is either problem with duplicate ids being generated by AA or problem caused by not using sql transactions during cache update. The problem might be caused by two cache inserts (insert into pagecache and insert into pagecache_str2find) for each cached page which are not "encapsulated" in transaction. So under high cache update load there are several concurent cache updates for the same record. I believe that those can result in such duplicates. n. |