ALTER FRAGMENT INIT progress

Dear All, I've tried to do ALTER FRAGMENT INIT ON TABLE that did not have any previous fragments, 25M rows, defining 3 fragments based on expression( serial column < X ...) on a quite old testing machine and it took around 20minutes to move the data and finish the operation.(more or less as expected, disks running at their full capacity for read and write etc... When I repeated this action on the production machine (much much stronger than the test one) on the same table with same indexes and 30M rows, it did not finish for 5 hours. The machine resources were idle most of the time and very few selects were running in parallel (once in few minutes). Now my questions are as follows: 1. is there any setting that i am missing in the onconfig file that causes this, or any particular issue that I am not aware of? 2. more important, is there any way to track the progress of the operation (in the manuals I've read suggestion to do an UPDSTATS and read sysgfragmenst nrows but the point is you cannot run updstats while table is locked. The only thing I've managed to catch from onstat -x of the session was the estimated rollback period which does not help much. Plus once I stopped the operation it took 10 times more to rollback the transaction than the estimation itself (while rolling back trough the logs in the onstat -x command. Thank you for any advice. Regards. Aleksandar

In both onconfig files: Wnat are the DS_ and SHM parameters set to? Is the disk layout different between the 2 servers? Which Informix version? While the alter fragment is running what is in: - sysmaster:syssesprof - sysmaster:sysscblst - onstat -g env - onstat -g ses for the session - sysmaster:sysrstcb for the threads of the session - onstat -g ioq Regards, David. > On 16 September 2016 at 09:56 ALEKSANDAR IVANOVSKI > <aleksandar.ivanovski@gmail.com> wrote: > > > Dear All, > > I've tried to do ALTER FRAGMENT INIT ON TABLE that did not have any previous > fragments, 25M rows, defining 3 fragments based on expression( serial column < > > X ...) on a quite old testing machine and it took around 20minutes to move the > > data and finish the operation.(more or less as expected, disks running at > their full capacity for read and write etc... > > When I repeated this action on the production machine (much much stronger than > > the test one) on the same table with same indexes and 30M rows, it did not > finish for 5 hours. The machine resources were idle most of the time and very > few selects were running in parallel (once in few minutes). > > Now my questions are as follows: > 1. is there any setting that i am missing in the onconfig file that causes > this, or any particular issue that I am not aware of? > 2. more important, is there any way to track the progress of the operation (in > > the manuals I've read suggestion to do an UPDSTATS and read sysgfragmenst > nrows but the point is you cannot run updstats while table is locked. The only > > thing I've managed to catch from onstat -x of the session was the estimated > rollback period which does not help much. Plus once I stopped the operation it > > took 10 times more to rollback the transaction than the estimation itself > (while rolling back trough the logs in the onstat -x command. > > Thank you for any advice. > Regards. > Aleksandar > > > ******************************************************************************* > > Forum Note: Use "Reply" to post a response in the discussion forum. >