Обсуждение: 7.3.4 freezing
Hi All, After PostgreSQL freeze some times, I am moving from 7.3.4 to 7.4.1, trying to solve this problem. When the daemon get frozen, I can't even use psql to browse a database (as if the database was down for some reason, or the daemon is not running). In these cases, a "pg_ctl -D /var/db stop" followed by a "pg_ctl -D /var/db start" fix the problem. I was not saving the logs in a file, so I can't show you the messages. Now I am saving it. Do you know some bug on this version (7.3.4)??? I will try 7.4.1 and seek for bugs! Regards, Marcelo ______________________________________________________________________ Yahoo! GeoCities: a maneira mais fácil de criar seu web site grátis! http://br.geocities.yahoo.com/
On Wed, 28 Jan 2004, MaRCeLO PeReiRA wrote: > Hi All, > > After PostgreSQL freeze some times, I am moving from > 7.3.4 to 7.4.1, trying to solve this problem. > > When the daemon get frozen, I can't even use psql to > browse a database (as if the database was down for > some reason, or the daemon is not running). > > In these cases, a "pg_ctl -D /var/db stop" followed by > a "pg_ctl -D /var/db start" fix the problem. > > I was not saving the logs in a file, so I can't show > you the messages. Now I am saving it. > > Do you know some bug on this version (7.3.4)??? > > I will try 7.4.1 and seek for bugs! Is the machine becoming unresponsive? I.e. is it taking several seconds for things like ls to run? What does top show for postgresql memory and cpu usage? Is postgresql really freezing, or just running REALLY slow? I.e. sometimes if you wait long enough you can get in to what seems to be a frozen database
Hi, > Is the machine becoming unresponsive? I.e. is it > taking several seconds > for things like ls to run? It isn't the case! I can do everything without problem, and without losing performance. The main problem is the postmaster daemon, that die. (really ;-) > What does top show for postgresql memory and cpu > usage? The top output is something like this: top - 16:44:24 up 2 days, 6:15, 10 users, load average: 1.24, 1.16, 1.05 Tasks: 123 total, 4 running, 119 sleeping, 0 stopped, 0 zombie Cpu(s): 96.0% user, 4.0% system, 0.0% nice, 0.0% idle Mem: 256292k total, 243696k used, 12596k free, 14024k buffers Swap: 530104k total, 82568k used, 447536k free, 114312k cached > Is postgresql really freezing, or just running > REALLY slow? I.e. > sometimes if you wait long enough you can get in to > what seems to be a > frozen database No way! It really die!! :( Regards, Marcelo ______________________________________________________________________ Yahoo! GeoCities: a maneira mais fácil de criar seu web site grátis! http://br.geocities.yahoo.com/
On Wed, 28 Jan 2004, MaRCeLO PeReiRA wrote: > Hi, > > > > Is the machine becoming unresponsive? I.e. is it > > taking several seconds > > for things like ls to run? > > It isn't the case! I can do everything without > problem, and without losing performance. The main > problem is the postmaster daemon, that die. (really > ;-) > > > What does top show for postgresql memory and cpu > > usage? > > The top output is something like this: > > top - 16:44:24 up 2 days, 6:15, 10 users, load > average: 1.24, 1.16, 1.05 > Tasks: 123 total, 4 running, 119 sleeping, 0 > stopped, 0 zombie > Cpu(s): 96.0% user, 4.0% system, 0.0% nice, > 0.0% idle > Mem: 256292k total, 243696k used, 12596k free, > 14024k buffers > Swap: 530104k total, 82568k used, 447536k free, > 114312k cached > > > Is postgresql really freezing, or just running > > REALLY slow? I.e. > > sometimes if you wait long enough you can get in to > > what seems to be a > > frozen database > > No way! It really die!! :( Dang. When you run top, what do the postgres owned processes show?
=?iso-8859-1?q?MaRCeLO=20PeReiRA?= <gandalf_mp@yahoo.com.br> writes:
> It isn't the case! I can do everything without
> problem, and without losing performance. The main
> problem is the postmaster daemon, that die. (really
> ;-)
That's pretty hard to believe (we've not had a postmaster-crashing bug
in a long time), and it's even harder to believe that the symptoms would
be as you describe if it did. if the postmaster wasn't there, psql
connection attempts would fail immediately with "connection refused" or
similar messages.
regards, tom lane
Hi Tom, I have just moved to 7.4.1. I hope to fix that problem. And, yes, I was receiving "connection refused" in the cases when the postmaster was gone! It has been exactly 36min the new version is running. Let us see the postmaster's behavior on the next few days! Thanks in advance and Regards, Marcelo --- Tom Lane <tgl@sss.pgh.pa.us> escreveu: > =?iso-8859-1?q?MaRCeLO=20PeReiRA?= > <gandalf_mp@yahoo.com.br> writes: > > It isn't the case! I can do everything without > > problem, and without losing performance. The main > > problem is the postmaster daemon, that die. > (really > > ;-) > > That's pretty hard to believe (we've not had a > postmaster-crashing bug > in a long time), and it's even harder to believe > that the symptoms would > be as you describe if it did. if the postmaster > wasn't there, psql > connection attempts would fail immediately with > "connection refused" or > similar messages. > > regards, tom lane ______________________________________________________________________ Yahoo! GeoCities: a maneira mais fácil de criar seu web site grátis! http://br.geocities.yahoo.com/
On Wed, 28 Jan 2004, Tom Lane wrote: > =?iso-8859-1?q?MaRCeLO=20PeReiRA?= <gandalf_mp@yahoo.com.br> writes: > > It isn't the case! I can do everything without > > problem, and without losing performance. The main > > problem is the postmaster daemon, that die. (really > > ;-) > > That's pretty hard to believe (we've not had a postmaster-crashing bug > in a long time), and it's even harder to believe that the symptoms would > be as you describe if it did. if the postmaster wasn't there, psql > connection attempts would fail immediately with "connection refused" or > similar messages. I'm pretty sure, from his first message, what he means is that the postmaster process is going into some kind of loop (100% CPU) and not responding, not necessarily dying. I'm reasonably certain English is not Marcelo's first language, so it's understandable he'd miss the mark on the meaning of a few words here and there.
Hi Tom and Scott, Thanks for your support. I hope the upgrade fix the problem. And, Scott, you are right, english isn't my first language, I'm a brazilian and I know brazilian portuguese. When I try to write english in a fast way, and without doing gramatical correction, I use to go into lots of mistakes. ;-) Thou I am really trying to improve my english. Well, once again, thanks for your support. Regards, Marcelo Pereira Campinas' University - Brazil Math Depth --- "scott.marlowe" <scott.marlowe@ihs.com> escreveu: > On Wed, 28 Jan 2004, Tom Lane wrote: > > > =?iso-8859-1?q?MaRCeLO=20PeReiRA?= > <gandalf_mp@yahoo.com.br> writes: > > > It isn't the case! I can do everything without > > > problem, and without losing performance. The > main > > > problem is the postmaster daemon, that die. > (really > > > ;-) > > > > That's pretty hard to believe (we've not had a > postmaster-crashing bug > > in a long time), and it's even harder to believe > that the symptoms would > > be as you describe if it did. if the postmaster > wasn't there, psql > > connection attempts would fail immediately with > "connection refused" or > > similar messages. > > I'm pretty sure, from his first message, what he > means is that the > postmaster process is going into some kind of loop > (100% CPU) and not > responding, not necessarily dying. I'm reasonably > certain English is not > Marcelo's first language, so it's understandable > he'd miss the mark on the > meaning of a few words here and there. > ______________________________________________________________________ Yahoo! GeoCities: a maneira mais fácil de criar seu web site grátis! http://br.geocities.yahoo.com/
On Wed, 28 Jan 2004, MaRCeLO PeReiRA wrote: > Hi, > > > > Is the machine becoming unresponsive? I.e. is it > > taking several seconds > > for things like ls to run? > > It isn't the case! I can do everything without > problem, and without losing performance. The main > problem is the postmaster daemon, that die. (really > ;-) > > > What does top show for postgresql memory and cpu > > usage? > > The top output is something like this: > > top - 16:44:24 up 2 days, 6:15, 10 users, load > average: 1.24, 1.16, 1.05 > Tasks: 123 total, 4 running, 119 sleeping, 0 > stopped, 0 zombie > Cpu(s): 96.0% user, 4.0% system, 0.0% nice, > 0.0% idle > Mem: 256292k total, 243696k used, 12596k free, > 14024k buffers > Swap: 530104k total, 82568k used, 447536k free, > 114312k cached > > > Is postgresql really freezing, or just running > > REALLY slow? I.e. > > sometimes if you wait long enough you can get in to > > what seems to be a > > frozen database > > No way! It really die!! :( > Can you perhaps build it with debugging symbols and run it until it crashs again and then send us the core file? You should check that the core file will be generated first, by checking things like the ulimit for core files that the postgresql server runs with. -- Nigel Andrews
=?iso-8859-1?q?MaRCeLO=20PeReiRA?= <gandalf_mp@yahoo.com.br> writes:
> And, yes, I was receiving "connection refused" in the
> cases when the postmaster was gone!
You didn't say that, you said it froze. Please be more accurate about
your bug reports in future. A minimum requirement is that you give the
exact text of any error messages you get.
It does sound like you were seeing a postmaster crash. As Nigel
suggested, it would be very useful to see a debugger stack trace
from such a crash.
regards, tom lane
Hi guys, > =?iso-8859-1?q?MaRCeLO=20PeReiRA?= > <gandalf_mp@yahoo.com.br> writes: > > And, yes, I was receiving "connection refused" in > the > > cases when the postmaster was gone! > > You didn't say that, you said it froze. Please be > more accurate about > your bug reports in future. A minimum requirement > is that you give the > exact text of any error messages you get. Sorry!! I will be more accurate next time. To tell you the truth, I wasn't running in troubles when I sent the first message, so I did not had the error messages to send to the list. I will be better next time! ;-) > It does sound like you were seeing a postmaster > crash. As Nigel > suggested, it would be very useful to see a debugger > stack trace > from such a crash. Sure I will do that!! Thanks!! Regards, Marcelo Yahoo! Mail - 6MB, anti-spam e antivírus gratuito. Crie sua conta agora: http://mail.yahoo.com.br