InstantForum.NET v2.9.0SQLServerCentralhttp://www.sqlservercentral.com/Forums/notifications@sqlservercentral.comSat, 18 May 2013 22:39:31 GMT20http://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxVery interesting information .... i have a SQL server 2005 installation on a server with 3 physical processors ( 12 logical ) , but did not face the issue during installation...... possibly because i did not attempt to run the SQL server before installing SP2 ...Thu, 11 Aug 2011 06:09:57 GMTvikram.bansalhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxI got it right in first shot. I had come across similar situation :-)Tue, 09 Aug 2011 03:34:02 GMTCOOL_ICEhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxGood question! I learned something new. I don't work with SQL 2005 anymore, but you'll never know when the knowledge is usefull.Fri, 05 Aug 2011 00:59:44 GMThakan.wintherhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxcome to knew new one.Wed, 03 Aug 2011 04:35:55 GMTDanny Oceanhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Tom.Thomson (8/2/2011)[/b][hr][quote][b]Kenneth Fisher-475792 (8/2/2011)[/b][hr]I do apologize that you felt that "Odd # of CPUs other than 1" and "RTM Install, when no service pack or hotfix was mentioned" were "unbelievable provisos" and will certainly try to be more careful of my wording in the future.Kenneth[/quote]On reflection, I think my comments were probably over the top. I was looking at the world through a haze of annoyment with my telephone provider, and it wasn't fair to allow that to prejudice my view of the question.[/quote]Perfectly understandable, every telephone provider I've ever worked with has driven me into a blind rage every now and again.I will say that I could have worded the question and answer a little more explicetly and part of it (as I look back at the KB article) has to do with my lack of knowledge about multicore processors etc.Tue, 02 Aug 2011 22:02:57 GMTKenneth.Fisherhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]shannonjk (8/2/2011)[/b][hr]The question was fine, as I see 'installation of SQL Server 2005 Enterprise edition" as being installed with no service packs. The answer however is inaccurate (though I answered it correctly) as explained in this articlehttp://support.microsoft.com/kb/954835I just encountered this last month and there are a few things. 1) No, installing SP2 or SP3 does not help this. We tried downloading the Digital versions of both from our resellers to no avail. Additionally, we tried expanding the service pack and installing via command prompt using the setup.exe hotfix=blah, solution to no avail as well. The ONLY option that would work, was forcing the server to think it only had one processor, install SQL Server and the subsequent service packs, then reboot the server into having its normal processors again. Also per what a previous person said in this post, this server was to support a websphere application, and the requirements from the vendor were to only install sql server 2005 sp1, which it is currently running at and fine.[/quote]Actually in my case the only thing that worked was to wait until the install crashed with an error (approx) of "Can't start the SQL Server Service", then when it prompted to retry/cancel I swaped out the recommended files (Sqlservr.exe and Sqlos.dll) then hit retry.KennethTue, 02 Aug 2011 21:56:20 GMTKenneth.Fisherhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Kenneth Fisher-475792 (8/2/2011)[/b][hr]I do apologize that you felt that "Odd # of CPUs other than 1" and "RTM Install, when no service pack or hotfix was mentioned" were "unbelievable provisos" and will certainly try to be more careful of my wording in the future.Kenneth[/quote]On reflection, I think my comments were probably over the top. I was looking at the world through a haze of annoyment with my telephone provider, and it wasn't fair to allow that to prejudice my view of the question.Tue, 02 Aug 2011 16:17:40 GMTL' Eomot Inverséhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxThe question was fine, as I see 'installation of SQL Server 2005 Enterprise edition" as being installed with no service packs. The answer however is inaccurate (though I answered it correctly) as explained in this articlehttp://support.microsoft.com/kb/954835I just encountered this last month and there are a few things. 1) No, installing SP2 or SP3 does not help this. We tried downloading the Digital versions of both from our resellers to no avail. Additionally, we tried expanding the service pack and installing via command prompt using the setup.exe hotfix=blah, solution to no avail as well. The ONLY option that would work, was forcing the server to think it only had one processor, install SQL Server and the subsequent service packs, then reboot the server into having its normal processors again. Also per what a previous person said in this post, this server was to support a websphere application, and the requirements from the vendor were to only install sql server 2005 sp1, which it is currently running at and fine.Tue, 02 Aug 2011 16:10:23 GMTshannonjkhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Tom.Thomson (8/1/2011)[/b][hr][quote][b]Kenneth Fisher-475792 (7/29/2011)[/b][hr]Same goes for 1 processor working. Technically I should have said "RTM installs fail on an odd number of logical or physical processers, unless its a single processor". However since in this case we weren't talking about 1 processor, logical or physical, I chose to shorten the answer slightly.[/quote]That, in my view, is appalling indifference to accuracy. [/quote]Probably true. What can I say, I'm not perfect[quote]It's a rotten question, because it leaves out essential information. Why should I, or anyone else, assume in late 2011 that an install of SQLS 2005 is not an install of SQL 2005 with SP2 or a higher service pack? Any DBA installing an earlier version than SP2 at this date would be irresponsible - the potential problems with such an old version are enormous.[/quote]First you did understand that this was a DR install right? You install what was in production. If RTM was in production (lets certainly hope not) then you install RTM. That being said I was starting at RTM with an intent to patch. Now oddly enough I work with a team of 8 DBAs ranging in experience from just a couple of years up through around 30 and none of us have ever used a "slipstreamed" install. We all install RTM then patch. Its certainly an interesting idea and I'm going to figure out how to do it and probably we will implement it at our company, but obviously assuming that a DBA will know how and will start with a patched install is just as poor an idea as assuming they will start with RTM.[quote]It's an even more rotten answer, because it is just plain wrong even assuming an install of the RTM version. What can that weasel word "technically" in your words quoted above purport to mean other than "actually what I said was correct was just plain wrong"?[/quote]oops, sorry, like I said above, not perfect[quote] BTW, I got the point because I guessed the simple answer would be wrong, did some hunting, found the story, and reached the conclusion that it was a ridiculous trick question. But most people don't have nasty suspicious minds like mine, and will assume that the question means what it asks with no unbelievable provisos thrown in and that the answers offered mean what they say. SO you can't take this comment as sour grapes.[/quote]Actually I'm glad you got the point. I posted the question, not to trick anyone, not to upset anyone, or cheat someone out of a point. I posted it because it was an interesting problem, and I thought there were people out there who could benefit from the information. I personally always try to use the QoD to learn something. If I get the question right or wrong I generally follow the link, when provided, and read at least something on the subject. I do apologize that you felt that "Odd # of CPUs other than 1" and "RTM Install, when no service pack or hotfix was mentioned" were "unbelievable provisos" and will certainly try to be more careful of my wording in the future.KennethTue, 02 Aug 2011 10:48:09 GMTKenneth.Fisherhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Tom.Thomson (8/1/2011)[/b][hr][quote][b]Kenneth Fisher-475792 (7/29/2011)[/b][hr]Same goes for 1 processor working. Technically I should have said "RTM installs fail on an odd number of logical or physical processers, unless its a single processor". However since in this case we weren't talking about 1 processor, logical or physical, I chose to shorten the answer slightly.[/quote]That, in my view, is appalling indifference to accuracy. The answer you wanted to be correct was actually incorrect. Why should anyone tick "it won't install on an odd number" when in fact it will install cleanly and without problem on 1 processor? 1 was an odd number last time I looked. Your "correct" answer is just plain "incorrect".It's a rotten question, because it leaves out essential information. Why should I, or anyone else, assume in late 2011 that an install of SQLS 2005 is not an install of SQL 2005 with SP2 or a higher service pack? Any DBA installing an earlier version than SP2 at this date would be irresponsible - the potential problems with such an old version are enormous.It's an even more rotten answer, because it is just plain wrong even assuming an install of the RTM version. What can that weasel word "technically" in your words quoted above purport to mean other than "actually what I said was correct was just plain wrong"?BTW, I got the point because I guessed the simple answer would be wrong, did some hunting, found the story, and reached the conclusion that it was a ridiculous trick question. But most people son't have nasty suspicious minds like mine, and will assume that the question means what it asks with no unbelievable provisos thrown in and that the answers offered mean what they say. SO you can't take this comment as sour grapes.EDIT: Steve, if you feel this comment is OTT, please delete it. But I think you should awatd point sto all who answered "it works OK"[/quote]I have to completely disagree.First, I just encountered this issue this month (I had encountered it year ago, but had forgot). I was installing SQL 2005 on a server with 2 CPU's, 6 cores/CPU, and Hyperthreaded. It initially looked like 24 cores, but after I got on ILO to do a hard reset I saw the configuration.Second, I always install SQL...then the SP. I don't slipstream.Third, I would not call running SP1 "irresponsable", there are vendor packages that may require it.Fourth, it's not so much an error of the question, but how MSFT documents it. Perhaps the documentation should say an odd number of cores, where the number of cores is greater than one. However, considering a workaround is to seet boot.ini to a single processor, perhaps they though it was unnecessary to point out that the install will succeed with one cpu.Mon, 01 Aug 2011 16:58:00 GMTGregoryFhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Kenneth Fisher-475792 (7/29/2011)[/b][hr]Same goes for 1 processor working. Technically I should have said "RTM installs fail on an odd number of logical or physical processers, unless its a single processor". However since in this case we weren't talking about 1 processor, logical or physical, I chose to shorten the answer slightly.[/quote]That, in my view, is appalling indifference to accuracy. The answer you wanted to be correct was actually incorrect. Why should anyone tick "it won't install on an odd number" when in fact it will install cleanly and without problem on 1 processor? 1 was an odd number last time I looked. Your "correct" answer is just plain "incorrect".It's a rotten question, because it leaves out essential information. Why should I, or anyone else, assume in late 2011 that an install of SQLS 2005 is not an install of SQL 2005 with SP2 or a higher service pack? Any DBA installing an earlier version than SP2 at this date would be irresponsible - the potential problems with such an old version are enormous.It's an even more rotten answer, because it is just plain wrong even assuming an install of the RTM version. What can that weasel word "technically" in your words quoted above purport to mean other than "actually what I said was correct was just plain wrong"?BTW, I got the point because I guessed the simple answer would be wrong, did some hunting, found the story, and reached the conclusion that it was a ridiculous trick question. But most people son't have nasty suspicious minds like mine, and will assume that the question means what it asks with no unbelievable provisos thrown in and that the answers offered mean what they say. SO you can't take this comment as sour grapes.EDIT: Steve, if you feel this comment is OTT, please delete it. But I think you should awatd point sto all who answered "it works OK"Mon, 01 Aug 2011 16:36:17 GMTL' Eomot Inverséhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Slick84 (8/1/2011)[/b][hr]bogus question...waste of timeshould have specific the service pack if you answer was going to be service pack specific.[/quote]Next time you have to do an install which fails due to service pack issues, let us know if this type of question is a waste of time. I see it as a very practical concern.Mon, 01 Aug 2011 10:50:54 GMTjeff.masonhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxThanks for the question.Mon, 01 Aug 2011 10:33:57 GMTTrey Stakerhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxbogus question...waste of time[b]should specify the service pack if your answer is going to be service pack specific.[/b]Mon, 01 Aug 2011 10:33:01 GMTSlick84http://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxGood questionMon, 01 Aug 2011 08:50:56 GMTSQLRNNRhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxThanks for the question.Sun, 31 Jul 2011 23:44:09 GMTKoen Verbeeckhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxNice question, good to know info.Fri, 29 Jul 2011 15:40:04 GMTKWymorehttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxRan into this issue years ago with an even number of processors but not a power of 2.4 sockets with 6 cores each = 24 logical.Fri, 29 Jul 2011 11:51:13 GMTJeremyEhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxThank you! Seeing it written as 12:3 made it 'click'; I get it now.Fri, 29 Jul 2011 11:31:00 GMTscunningham 38251http://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]scunningham 38251 (7/29/2011)[/b][hr]3 quad core processors = 12 or 24 CPU's (24 since I'm not sure if the CPUs were HT or not, or how SQL Server 'sees' CPUs with HT)So, the ratio of CPUs to sockets is 12/3=4 or 24/3=8.2^2 = 4, and 2^3 = 8 so both are a power of two.What am I missing?[/quote]That's a good question and all I can say is that a ratio is not always divided. The ratio of CPUS to sockets should be written as 12:3 not 12/3.An example:If orange juice concentrate is to be diluted with water in the ratio 1:4, then one part of concentrate is mixed with four parts of water, giving five parts total; the fraction of concentrate is 1/5 and the fraction of water is 4/5. Notice it's not 25% orange juice as you would think.Fri, 29 Jul 2011 10:23:38 GMTcengland0http://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxFirst, I would like to say that as it said in the question, this was during a DR test. I've never used a "slipstreamed" install before, and if it is similar to method 1 in the article where you use the HOTFIXPATCH parameter on the install, this didn't work for me. I had to run the install, then replace Sqlservr.exe and Sqlos.dll while on the "Retry or Cancel" button, then hit "Retry". All that being said, I didn't think to mention that it was RTM, if for no other reason then if you don't mention a patch level software is generally assumed to be at RTM. Or at least I always do. Next time I do a QOD though I will make sure to add RTM, since this was supposed to be an interesting learning question, not a trick one.Same goes for 1 processor working. Technically I should have said "RTM installs fail on an odd number of logical or physical processers, unless its a single processor". However since in this case we weren't talking about 1 processor, logical or physical, I chose to shorten the answer slightly.I'm actually surprised to hear that anyone else has had this problem, let alone multiple times since I was assured by my server team that this was highly unusual.And last but not least, scunningham you shouldn't feel dumb at all. The important thing here though is that there were THREE quad cores. Or in this case 3 physical processors, each one split into 4 logical.I have to admit I posted the question more because I thought it was an interesting problem, and didn't want anyone else to get caught out unawares, particularly since when all you have to go on is [quote]The SQL Server service failed to start. For more information, see the SQL Server Books Online topics, "How to: View SQL Server 2005 Setup Log Files" and "Starting SQL Server Manually."[/quote]It can be a real bear to find a solution.KennethFri, 29 Jul 2011 10:02:20 GMTKenneth.Fisherhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]cengland0 (7/29/2011)[/b][hr][quote][b]Carlton Leach (7/29/2011)[/b][hr]SQL can't be installed on a single socket machine?Carlton.[/quote]It can. One is a power of two (2^0).[/quote]My point exactly, "can't be installed with an [b]odd[/b] number of physical or logical processors." Carlton.Fri, 29 Jul 2011 09:27:36 GMTCarlton Leachhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]cengland0 (7/29/2011)[/b][hr][quote][b]Carlton Leach (7/29/2011)[/b][hr]SQL can't be installed on a single socket machine?Carlton.[/quote]It can. One is a power of two (2^0).According to the reference:[quote]The ratio between logical processors and physical sockets is not a power of 2. For example, the computer has a single socket together with a triple-core processor. [/quote]Besides, the problem was fixed in SP2.[/quote]I have the feeling I'm going to feel dumb once this is explained to me, but I gotta ask - why did the install fail in the author's case?3 quad core processors = 12 or 24 CPU's (24 since I'm not sure if the CPUs were HT or not, or how SQL Server 'sees' CPUs with HT)So, the ratio of CPUs to sockets is 12/3=4 or 24/3=8.2^2 = 4, and 2^3 = 8 so both are a power of two.What am I missing?Fri, 29 Jul 2011 09:13:02 GMTscunningham 38251http://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Meet George Jetson (7/29/2011)[/b][hr]Not to nitpick, but even the article referenced stated - "On a computer that has a multicore processor, you [b]may [/b]be unable to install SQL Server 2005"The optimal word being "may."[/quote]You took that out of context. Sure that's the title of the article but the rest of the article describes the conditions where the installation will fail. In short, the title is correct that not all multicore processor installations will fail so the word "may" is right.Fri, 29 Jul 2011 07:49:33 GMTcengland0http://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxI have to say that the correct answer is not worded in such a way as to make much sense. It says "Install fails because SQL 2005 can't be installed with an odd number of physical or logical processors" which is just plain wrong. It WILL install with a single processor with no problem. The article at least clearly states what is defined as the correct reason it fails. The reason I chose that it would install correctly is because I had eliminated the even processor count as incorrect because I know that it will install with a single processor with no issue. Interesting question on a old version of sql, horrible wording on the correct answer. Seems that my sneek-o-meter failed me today.Fri, 29 Jul 2011 07:35:43 GMTSean Langehttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxKeep in mind that this could be for DR. On many sites, we use original media, pre-SP2, to do a DR install on the fly. In fact, the only time I have used slipstream media (which I had to make myself) was installing SQL on Windows 2008 clusters in an environment where you could not even start to install without the SP itself being slipstreamed. So whether you pick at the nit or not, it still is a very good piece of information to know and therefore a good question.Fri, 29 Jul 2011 07:32:05 GMTjeff.masonhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxI answered that it would install correctly. I was sure that a company as big as Microsoft could figure out how to get a database program to count the number of CPUs in the server hardware it would be running on. Fri, 29 Jul 2011 07:22:33 GMTpatrickmcginnis59http://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Meet George Jetson (7/29/2011)[/b][hr]Not to nitpick, but even the article referenced stated - "On a computer that has a multicore processor, you [b]may [/b]be unable to install SQL Server 2005"The optimal word being "may."[/quote]On a server with an odd number (physical CPU * core), I have *never* been able to do a successful install of SQL 2005 without making modifications. There is no may about it, it just plain does not work as isFri, 29 Jul 2011 07:20:43 GMTGregoryFhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxNot to nitpick, but even the article referenced stated - "On a computer that has a multicore processor, you [b]may [/b]be unable to install SQL Server 2005"The optimal word being "may."Fri, 29 Jul 2011 07:17:25 GMTMeet George Jetsonhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Shark Energy (7/29/2011)[/b][hr]In the absence of an SP listed in the question for the SQL install, I'd assume he is referring to vanilla flavour.Totally agreeing to disagree :-P[/quote]You can't assume RTM anymore than a specific SP as he didn't specify RTM either.Fri, 29 Jul 2011 07:09:29 GMTmtassinhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Shark Energy (7/29/2011)[/b][hr]No excuses people I mean if follow the question in your head, it fails before you get to installing any SPx so follow the question and its all in front of you.[/quote]Not if you or a helpful netadmin slipstreamed the sp onto the RTM install media, or the disk you purchased to install came with SP2.Fri, 29 Jul 2011 07:08:20 GMTmtassinhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]anthony.green (7/29/2011)[/b][hr]I would have to agree with Murray on this, it would have been nice to know which level of the product it relates to.All the media I have been working with included SP2 as part of the disk so I have never encountered this issue.[/quote]Same here. Stating RTM SQL 2005 or SP1 2005 would have clarified it better. Alas...Fri, 29 Jul 2011 07:07:00 GMTmtassinhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxIt would be nice to have more questions like these. Good question. Thanks.Fri, 29 Jul 2011 07:03:55 GMTmohammed moinudheenhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxThis has bit me numerous times (including last week), now I make a habbit of changing the boot.ini file to one processor before all SQL 2005 installs on machine with an odd number of total processors or cores. That being said, one processor is an odd number, yet the install worksFri, 29 Jul 2011 06:39:03 GMTGregoryFhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxWent for the obvious answer of installs correctly given the information provided.Fri, 29 Jul 2011 06:18:18 GMTBritt Cluffhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxThank-you for the question. It is good to know that such a bug ever existed with SQL Server.Fri, 29 Jul 2011 05:43:41 GMTNakul Vachhrajanihttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxI figured the question came up because of a real life problem - and therefore assumed it probably DIDN'T install correctly... but I still picked that answer assuming I'd learn something new this morning. Thank you for the information and link!I think the question was fair. Good job.Fri, 29 Jul 2011 05:30:49 GMTRich Weisslerhttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspx[quote][b]Carlton Leach (7/29/2011)[/b][hr]SQL can't be installed on a single socket machine?Carlton.[/quote]It can. One is a power of two (2^0).According to the reference:[quote]The ratio between logical processors and physical sockets is not a power of 2. For example, the computer has a single socket together with a triple-core processor. [/quote]Besides, the problem was fixed in SP2.Fri, 29 Jul 2011 05:23:40 GMTcengland0http://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxAh! an assumption... he he! not so clear then ;-)Fri, 29 Jul 2011 04:05:52 GMTDugyChttp://www.sqlservercentral.com/Forums/Topic1150652-1186-1.aspxIn the absence of an SP listed in the question for the SQL install, I'd assume he is referring to vanilla flavour.Totally agreeing to disagree :-PFri, 29 Jul 2011 04:02:43 GMTShark Energy