just tested with the latest from HEAD and the problem is still there. I need to work more on a independent test case. With all the tracing that I have on, I hope to actually recreate the situation. It is definitely reproducible (it always happens in the 3rd read transaction after I startup again, but I am thinking something goes wrong before the kill/restart)
Simon From: Simon Helsen/Toronto/IBM@IBMCA To: [email protected] Cc: Andy Seaborne <[email protected]>, [email protected] Date: 08/22/2012 12:36 PM Subject: Re: something that was written in a transaction is not found Andy, I updated https://issues.apache.org/jira/browse/JENA-301. I think that one can be closed. FYI I had created a new snapshot with the status of HEAD about 2 hours ago and with these changes (which fixed 301), the exception below remains. Moreover, I checked if that problem was also bound to direct mode, but it wasn't. Both mapped and direct mode showed the same exception. However, I just noticed a whole whack of new deliveries which I have not yet tested. I'll do that, but as for the time of writing, the exception Quad: predicate cannot be null is still a problem Simon From: Andy Seaborne <[email protected]> To: [email protected] Date: 08/22/2012 04:31 AM Subject: Re: something that was written in a transaction is not found On 21/08/12 23:32, Simon Helsen wrote: > I am seeing for each write transaction "Txn[8]/W: Add to pending > queue". I assume this is ok? Yes. It will typically do this because of JENA-256. 9 in 10 write transactions will do that in a serial work load. > Then, when I kill the process and start again, I have the following > sequence, leading to the Quad: predicate cannot be null This may be JENA-301 which is an issue across restart. > this is all on 2.7.4 from HEAD When exactly? Please try the current svn head or last night's build. Not marked fixed, but a workaround to *an* issue is, as of yesterday, in svn trunk. (also - run in mapped mode and see if it's different - that would be useful information for JENA-301). Andy
