Обсуждение: 7.2b2 problem using like 'XXX%' sequential scan
Ciao, I'm testing 7.2b2 and I see it is much slower when I use like on column with indexes, the explain sais it is a sequantial scan (same as 7.1.3), but the time it get to complete is much longer. I got the problem only on SMP machine: explain on 7.1.3 say it is a seq scan with cost 18837 explain on 7.2b2 say it is a seq scan 22.5 on single cpu 7.2b2 says: Index scan with cost 29 and it is very fast!! All platform are linux/i386 using redhat 7.2 and latest libraries I'm testing latest snapshot just now. thanks -- ------------------------------------------------------- Giuseppe Tanzilli g.tanzilli@gruppocsf.com CSF Sistemi srl phone ++39 0775 7771 Via del Ciavattino Anagni FR Italy
Yes, I vacuumed/reindexed the db, just tested the latest snapshot, same problem. bye Antonio Fiol Bonnín wrote: >Did you VACUUM ANALYZE on your SMP machine? > >Antonio > >Giuseppe Tanzilli - CSF wrote: > >>Ciao, >>I'm testing 7.2b2 and I see it is much slower when I use like on column >>with indexes, >>the explain sais it is a sequantial scan (same as 7.1.3), but the time >>it get to complete is much longer. >> >>I got the problem only on SMP machine: >>explain on 7.1.3 say it is a seq scan with cost 18837 >>explain on 7.2b2 say it is a seq scan 22.5 >> >>on single cpu 7.2b2 says: >>Index scan with cost 29 and it is very fast!! >> >>All platform are linux/i386 using redhat 7.2 and latest libraries >> >>I'm testing latest snapshot just now. >> >>thanks >> >> >> >>-- >>------------------------------------------------------- >>Giuseppe Tanzilli g.tanzilli@gruppocsf.com >>CSF Sistemi srl phone ++39 0775 7771 >>Via del Ciavattino >>Anagni FR >>Italy >> >>---------------------------(end of broadcast)--------------------------- >>TIP 5: Have you checked our extensive FAQ? >> >>http://www.postgresql.org/users-lounge/docs/faq.html >> > > -- ------------------------------------------------------- Giuseppe Tanzilli g.tanzilli@gruppocsf.com CSF Sistemi srl phone ++39 0775 7771 Via del Ciavattino Anagni FR Italy
... > All platform are linux/i386 using redhat 7.2 and latest libraries I'm guessing that one platform has "locale" enabled (or some combination of multibyte parameters?), whereas the other does not. SMP will not make a difference in configuration or optimizer choices. The good news is that you have *one* machine which does what you want, so we know that we can get the other one doing that too :) - Thomas
(back on-list) > You got it, I have locale anebled on all platform, > but I see a different locale setting: > > it is working where I have LC_* = it_IT > not working where I have LC_* = it_IT@euro > what I can do ? > I tried to start postgresql 7.2b2 with the working locale but nothing > appened, I must initdb ?? I would try that. If you stay on list, then someone who actually *knows* is likely to answer. I'm just guessing ;) - Thomas
I did a fresh initdb after setting LC_* to it_IT but nothing seems to change bye Thomas Lockhart wrote: >(back on-list) > >>You got it, I have locale anebled on all platform, >>but I see a different locale setting: >> >>it is working where I have LC_* = it_IT >>not working where I have LC_* = it_IT@euro >>what I can do ? >>I tried to start postgresql 7.2b2 with the working locale but nothing >>appened, I must initdb ?? >> > >I would try that. If you stay on list, then someone who actually *knows* >is likely to answer. I'm just guessing ;) > > - Thomas > -- ------------------------------------------------------- Giuseppe Tanzilli g.tanzilli@gruppocsf.com CSF Sistemi srl phone ++39 0775 7771 Via del Ciavattino Anagni FR Italy
>> it is working where I have LC_* = it_IT >> not working where I have LC_* = it_IT@euro >> what I can do ? >> I tried to start postgresql 7.2b2 with the working locale but nothing >> appened, I must initdb ?? Yes. The database locale is determined at initdb time. regards, tom lane