Bug #62436 | auto increment in MySQL Cluster always increases - even on insert error! | ||
---|---|---|---|
Submitted: | 14 Sep 2011 20:47 | Modified: | 23 Jan 2012 8:22 |
Reporter: | J H | Email Updates: | |
Status: | Not a Bug | Impact on me: | |
Category: | MySQL Cluster: Cluster (NDB) storage engine | Severity: | S2 (Serious) |
Version: | 5.1.56-ndb-7.1.15a | OS: | Linux |
Assigned to: | Martin Skold | CPU Architecture: | Any |
Tags: | auto increment, id, insert, ndb |
[14 Sep 2011 20:47]
J H
[14 Sep 2011 20:57]
J H
Forgot to mention: another acceptable solution may be to make this work with auto_increment_offset, because each node can individually reserve given id space, allowing it to 'release' the 'reserved' increment (or "roll-back" or whatever). In other words, with auto_increment_offset, each node can control given increment values, and therefore there should be no need to talk to other nodes. Just a thought; obviously not a perfect (or even preferred) one, but still a thought.
[21 Jan 2012 11:30]
J H
Is anyone even looking at this? A simple, "here is where we are at" will do.... But seriously, four months with no response? This is sad....
[23 Jan 2012 8:22]
Martin Skold
There is no claim that auto_increment should work the same in all engines. The only thing we guarantee is that it is an a monotonically increasing number series. No guarantees that all values are used or that there will not be holes. Auto_increment in cluster is much more complex than in local non-distributed storage engines since it is a distributed counter with prefetch of several values. This might be fixed in the future, but it is really a feature request.