tag:blogger.com,1999:blog-5124641802818980374.post8839164758946599819..comments2023-09-06T11:58:30.752+02:00Comments on Oracle related stuff: ASSM bug reprise - part 1Unknownnoreply@blogger.comBlogger7125tag:blogger.com,1999:blog-5124641802818980374.post-55893784411394329832011-05-20T09:30:18.539+02:002011-05-20T09:30:18.539+02:00Henry,
I think it is simply sbout a shortcoming/b...Henry,<br /><br />I think it is simply sbout a shortcoming/bug in the ASSM implementation that has been fixed in 11.2.<br /><br />As you've pointed out, MSSM and ASSM handle the "freeness" of blocks differently internally, so you'll end up with those little details that might make a big difference in some special circumstances.<br /><br />RandolfRandolfhttps://www.blogger.com/profile/13463198440639982695noreply@blogger.comtag:blogger.com,1999:blog-5124641802818980374.post-44741105487234796312011-05-19T19:32:45.068+02:002011-05-19T19:32:45.068+02:00Randolf,
Why ASSM and not MSSM? The first reason ...Randolf,<br /><br />Why ASSM and not MSSM? The first reason that comes to mind is that MSSM will take the block off the free list when all ITL slots are full. However I don't know if that is really the case.<br /><br />Interesting stuff.Henryhttps://www.blogger.com/profile/16660217870358681323noreply@blogger.comtag:blogger.com,1999:blog-5124641802818980374.post-47093365209582862942011-05-19T08:43:34.471+02:002011-05-19T08:43:34.471+02:00Randolf,
yes, I remember the thread. Rereading the...Randolf,<br />yes, I remember the thread. Rereading the discussing, I think I can recognize some of the patterns - but still I have no idea what role the flushing of the shared pool (mentioned in the xing thread) could play in this context. I look forward to part 2.<br />MartinMartin Preisshttps://www.blogger.com/profile/06388592214305009761noreply@blogger.comtag:blogger.com,1999:blog-5124641802818980374.post-18823045364906198682011-05-18T12:15:52.234+02:002011-05-18T12:15:52.234+02:00Hi Martin,
you might remember that thread some ti...Hi Martin,<br /><br />you might remember that thread some time ago on a German forum where also someone hit the variation of the bug - that was the first time I realized that there is more to this bug than the basic one described here.<br /><br />The recent OTN post finally pushed me to complete at least this part 1 note about the basic bug, and hopefully I'll be able to complete part 2 sometime soon, since it is much more relevant and also very likely explains what has been described in the "long running update" thread.<br /><br />RandolfRandolfhttps://www.blogger.com/profile/13463198440639982695noreply@blogger.comtag:blogger.com,1999:blog-5124641802818980374.post-46458639534302369002011-05-17T22:50:46.053+02:002011-05-17T22:50:46.053+02:00Randolf,
thank you for the concise explanation. An...Randolf,<br />thank you for the concise explanation. An interesting addition to the current <i>long running update</i> thread in OTN (http://forums.oracle.com/forums/thread.jspa?threadID=2215565&start=30&tstart=0).<br /><br />MartinMartin Preisshttps://www.blogger.com/profile/06388592214305009761noreply@blogger.comtag:blogger.com,1999:blog-5124641802818980374.post-85618063077266584582011-05-17T17:03:10.840+02:002011-05-17T17:03:10.840+02:00Hi Hemant,
yes, it is true that it can be reprodu...Hi Hemant,<br /><br />yes, it is true that it can be reproduced in any block size - but you have to consider that it takes quite an exotic setup in those smaller block sizes, therefore it is quite unlikely to be hit in real life since most of the time the migrated rows will be longer than 24 bytes (as discussed above).<br /><br />However, and that is going to be covered in part 2, the variation of the bug introduced in 11.1 can happen with longer row sizes in combination with the default block size of 8K, so it is much more relevant to real-life cases.<br /><br />RandolfRandolfhttps://www.blogger.com/profile/13463198440639982695noreply@blogger.comtag:blogger.com,1999:blog-5124641802818980374.post-39103943811235025442011-05-17T16:31:43.445+02:002011-05-17T16:31:43.445+02:00"The bug can show up with any block size"..."The bug can show up with any block size" is significant.<br />I do remember following the original discussion a long time ago (going by the timestamp on Greg Rahn's blog, Aug-Sep '08). Then, part of the "conclusion" was that it was manifest in 16K blocks. (I believe that it was in this discussion that comments like "don't use non-standard block sizes because Oracle doesn't test them well" were also repeated).<br /><br /><br />So, you can reproduce this in 8K blocks as well.Hemant K Chitalehttps://www.blogger.com/profile/07369112096230549250noreply@blogger.com