http://method-r.com com_fireboard en-us Wed, 08 Sep 2010 22:04:07 -0700 http://method-r.com/components/com_fireboard/template/default_green/images/english/emoticons/fb_rsspower.gif http://method-r.com 88 http://method-r.com/forums?func=view&id=174&catid=12 You're welcome. :)... http://method-r.com/forums?func=view&id=173&catid=12 jholt wrote: But I'm not sure I'm understanding your comment about how pb and b were "a lot larger...on older databases". Generally speaking, the format of an Oracle block hasn't changed since version 7. Specifically, it has but not generally. I foolishly made a mental correlation between things that weren't tied together: When I worked at Company A on 9i, I saw x. When I worked at companies B and C on 10/11g, I saw y. Therefore the difference in results must be related to 9 vs 10/11. Plus I m... http://method-r.com/forums?func=view&id=172&catid=12 But the queries that "ASH will show" will be differentiated only if their execution rate dominates the total execution rate of all the other executions combined. That might not be true. To make matters worse is when a problem is not "recreated on a test environment." A diagnostic method that reliably works in production for 3+ tier systems is more valuable than most people will admit. The scenario I described earlier--and others--can be investigated with complete certainty only when the application is ... http://method-r.com/forums?func=view&id=171&catid=12 To be honest, I'm no expert in OPNET Panorama. When I said "we" I meant others in my team. Still, I've seen screens where they trace a particular transaction (not sure if it's a single http request) and the tool shows everything that is executed in the app layer up to and including SQL statements. I work in a development team so I've never had the "pleasure" of investigating performance issues in a production environment. That's why on my other replies I explain I investigate issues in a test environment... http://method-r.com/forums?func=view&id=170&catid=12 Ok, I think I understand more of your concerns. A block that is part of the segment does not have to have any data. So, the number of 'N' rows plus the number of 'Y' rows could be a whole lot less than the number of blocks below the HWM. Just create an empty table with a large initial extent and run the report and that will demonstrate what I trying to convey. The fact that the column is defined as "NOT NULL" means only that if a row exists, then it has to have a value. If a block has no rows, then th... http://method-r.com/forums?func=view&id=169&catid=12 Can OPNET Panorama tell you where all the time went for a specific browser's HTTP request, which had an 8 second duration? What if that request caused the app server to execute some statements thousands of times (of course, all with short durations) while all the other HTTP requests (finishing in < 1 second) were satisfied with few SQL statement executions (again, with short durations)?... http://method-r.com/forums?func=view&id=168&catid=12 Perhaps my problem simply lies in incorrect expectations... I thought that with a Y/N column that doesn't allow NULL values, the block count for Y + block count for N should be a lot closer to the total blocks below HWM (or in some cases even higher if the values were relatively evenly mixed within blocks). I just seem to recall (pb) and (b) being a lot larger numbers in relation to total blocks below HWM on older databases.... http://method-r.com/forums?func=view&id=167&catid=12 True. ASH will identify slow queries and they may not be the ones slowing down your business transaction. That's why I have a separate test environment that is a copy of production and have one user run the slow transaction. In that scenario (free from other users running different transactions) ASH proves to be very useful, if the main issue is slow queries. If no slow queries are identified, there is a possibility the issue could be outside of the database tier. In that case we use OPNET Panorama to... http://method-r.com/forums?func=view&id=166&catid=12 That looks like the code I wrote. I don't recall if Method R owns that property but it probably does not matter when it comes to fixing bugs. So, I'll definitely post a fix if required. If I temporarily 'set termout on' while executing the query on line 22 (connected to 11.1.0.7), I see that it's getting file and block number from characters 7 through 15 in the string representation of a rowid. I recall this is the proper positions to get the file and block id for RDBMS version 8+. Can you be more spe... http://method-r.com/forums?func=view&id=165&catid=12 I may be working with an old version of the script, but I've been getting incorrect results from hds.sql: [code type=sql]Table blocks below HWM Table rows (B) (R) -------------- ---------------- 116455 3,781,455 Block selectivity Block count Row count IND (pb = b/B) (b) (r) --------- ----------------- -------------- ---------------- N 22.02% 25,644 3,197,090 Y 6.16% 7,172 584,365[/code] It's not a HWM issue, I...