Re: how does ODS uniquely id transactions?

Wanda, I don't have an answer for you but I can back you up with my own little investigations. I tried it myself out of curiosity and got the same results. So this is not an entry in the journal of irreproducible results. I found some other anomalies: - When I specified a user (onlog -u jake) I still got loads of checkpoint records displayed. Hey, man! I did not execute your ol' checkpoint! And all my transactions that it did display were xid:13. - When I tried onlog -x 13, it still gave me the checkpoint info. Either Wanda and I are suffering from a deep misunderstanding over exactly what the columns mean or onlog is so screwed up as to be useless as a diagnostic tool. Please, someone tell me it us and give us a better explanation than the FM (Fine Manual) tells us. Also, I do not have a log tape to play with. If Wanda would repeat the experiment with the log tape (onlog -d /dev/whatever) and see if the txid makes sense now. This would still not help for a distributed transaction, which should have a unique TXID across the whole network. BTW, I am using 7.13; I know Wanda is using a later version because dbinfo('sessionid') seems to work in her qurey and fails in mine. Any suggestions? Answers? More questions? Cold brew? Wanda Beck wrote: > > Hi, All! > > I'm a newbie trying to understand how Online Dynamic Server 7.21 > uniquely tags its transactions. I have a trigger on a table where I > need to extract session id, user id, and transaction id for the > triggering statement. > I know how to get session and user ids. To get a transaction id, I'm > doing: > select tx_id from sysmaster:systrans > where tx_addr = (select us_txp from sysmaster:sysuserthreads > where us_sid = dbinfo( 'sessionid' ) )> > In testing this using dbaccess, I find that *all* the transactions > fron a session have the same transaction ids, regardless of the fact > that some are singletons and others are bracketed by BEGIN WORK and > COMMIT WORK. Furthermore, if that user logs out, then back in again, > and if that spot in the transaction table is available, the same > transaction id is assigned. > > I've looked in the log using onlog -u <username> and see that even > there, all the transaction ids are the same, just the addresses in the > log are different. > > Can anyone tell me how I can get the unique transaction identifier for > a triggering statement? I'm really having a hard time with this one. -- -- Jake (In pursuit of undomesticated aquatic avians) +----------------------------------------------------------+ |Aside from that, how did you enjoy the play, Mrs. Lincoln?| +----------------------------------------------------------+