On 4/12/07, tnine <[EMAIL PROTECTED]> wrote:
Excellent! Thank you for the help. I may have a race condition that I've thought of that may cause problems, but this may be taken care of my the grouping. If you could give me some advice, I would really appreciate it. It pertains to the grid processing and the nth and n-1 messages in a group in a 2 node or greater environment. Lets say we have n messages in the group, where the last message (n) sets the sequence number to 0 to close the group. When my n-1 message is consumed off the input queue, it will take a few seconds to process. The nth message is to signal the group is closed, so the receiving MDB will simply push a new message with the sequence 0 on to the output queue. The message with 0 will be on the output queue before the n-1 sequence is pushed. Won't my n-1 message get excluded from the group since its inserted after 0?
I don't quite follow. Note that the sequence number is not used to reorder the messages in the group. The zero is only used to close a group. (To be honest there's no real need to close a group since we're using hash buckets under the covers by default so there's little cost to keeping a group open).
This is an implementation detail, but an important one. Below is an ASCII diagram of my queues and MDB Standalone file reader--->Input Queue (jms/StatInput)--> MDB >-- Output Queue(jms/StatOutput) --> Consumer As you can see, I have 2 queues for round trip communication, will this cause the sequences issues I described above? If so, do you have any ideas on how I can get around them?
Each queue is separate and wil be ordered based on the order in which messages (within a group) are published; so it all sounds fine to me -- James ------- http://radio.weblogs.com/0112098/
