[ http://opensource.atlassian.com/projects/hibernate/browse/HHH-3586?page=c... ] Krasimir Chobantonov commented on HHH-3586: ------------------------------------------- Also HiLoOptimizer seems that have the same issue.. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://opensource.atlassian.com/projects/hibernate/secure/Administrators.... - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ http://opensource.atlassian.com/projects/hibernate/browse/HHH-3586?page=c... ] Krasimir Chobantonov commented on HHH-3586: ------------------------------------------- Robin, That is the problem - the multiplication takes the id to go far more that it is allowed - it is not in the range of the "reserved" ids. in the above code you are basically saying that the id that have to be generated is equal to the current sequence number multiplied by the step + 1 - imagine that the current sequece is 1000 and the step is 50 then the next generated id should be in the range of 1000 and 1050 but because of the multiplication you will end up with the following result id = 1000 * (50 + 1) = 51000 as you see the 51000 is far more ahead than the 1050 limit - and yes in cluster environment you will have a big problem with that ... in a single machine env you will just "eat" the much more ids and you will see huge "holes" in the ids that are beeing generated. Krasimir -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://opensource.atlassian.com/projects/hibernate/secure/Administrators.... - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ http://opensource.atlassian.com/projects/hibernate/browse/HHH-3586?page=c... ] Robin Sander commented on HHH-3586: ----------------------------------- @Krasimir: you got me wrong, I should better have written: "The multiplication is needed to "reserve" a sequence of identifiers without increasing the sequence for every one". I used quotes for reserve, because those values aren't really reserved in the database. To use your example: 1. sequence has current value 1000 2. fetch nextval and "reserve" 50 values, so in memory we would have values from 1001 to 1051 3. sequence has current value 1001 now What happens now if the sequence is used for another type or by another cluster node? You would use values from 1002 to 1052 so we would have a big problem. I think what you got in mind is to increase the value of the sequence by (step+1) and hence really reserve those values in the database. I don't know whether this is possible or not (performance?), maybe some core developers could shed some light on this topic? (I've got nothing to do with the code nor the algorithm, just stumbled accross this issue because I wondered about those gaps in ID generation myself) -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://opensource.atlassian.com/projects/hibernate/secure/Administrators.... - For more information on JIRA, see: http://www.atlassian.com/software/jira