postmaster dies (6.5.3) - Mailing list pgsql-hackers

From Oleg Bartunov
Subject postmaster dies (6.5.3)
Date
Msg-id Pine.GSO.3.96.SK.991215105557.10383h-100000@ra
Whole thread Raw
Responses Re: [HACKERS] postmaster dies (6.5.3)
List pgsql-hackers
Hi,

I got seriuos problem this night  with postgres which is running as 
db backend to apache. I have cron job which vacuuming database
every hour and it worked for weeks without problem
(well, there is problem with concurrent processes under high load,
but this night was very quiet)

Here is the processes currently  seen:
 167  ?  S    0:00 /usr/local/pgsql/bin/postgres localhost httpd nature idle   168  ?  S    0:00
/usr/local/pgsql/bin/postgreslocalhost httpd nature idle   169  ?  S    0:00 /usr/local/pgsql/bin/postgres localhost
httpdnature idle   170  ?  S    0:00 /usr/local/pgsql/bin/postgres localhost httpd nature idle   171  ?  S    0:00
/usr/local/pgsql/bin/postgreslocalhost httpd nature idle  
 
26578  ?  S    0:00 /usr/local/pgsql/bin/postgres localhost httpd nature idle  
29372  ?  S    0:00 /usr/local/pgsql/bin/postgres localhost httpd discovery idle


from apache's error log:
[Wed Dec 15 02:10:08 1999] [error] DBI->connect failed: connectDB() --  couldn'tsend startup packet: errno=32
Broken pipeat /opt/perl5/lib/site_perl/5.005/Apache/DBI.pm line 138

[Wed Dec 15 02:10:08 1999] [error] DBI->connect failed: pqReadData() -- backend 
closed the channel unexpectedly.       This probably means the backend terminated abnormally       before or while
processingthe request.at /opt/perl5/lib/site_perl/5.005/Apache/DBI.pm line 138
 

[Wed Dec 15 02:10:43 1999] [error] DBI->connect failed: connectDB() -- connect()failed: No such file or directory
Is the postmaster running at 'localhost' and accepting connections on Unix socke
t '5432'?at /opt/perl5/lib/site_perl/5.005/Apache/DBI.pm line 138


fortunately postmaster was started with debug option:

StartTransactionCommand
query: SET client_encoding = 'KOI8'
ProcessUtility: SET client_encoding = 'KOI8'
CommitTransactionCommand
postmaster: StreamConnection: accept: Invalid argument
/usr/local/pgsql/bin/postmaster: ServerLoop:            handling reading 6
/usr/local/pgsql/bin/postmaster: ServerLoop:            handling reading 6
FATAL 1:  ReleaseLruFile: No opened files - no one can be closed
proc_exit(0) [#0]
shmem_exit(0) [#0]
exit(0)

Aha,

>From system log files I found probable explanation - file table overflow !
Could this be a reason of postmaster dead and how to avoid this ?

Dec 15 01:47:27 zeus kernel: Unable to load interpreter
Dec 15 02:09:28 zeus xntpd[103]: kernel pll status change 89
Dec 15 02:10:08 zeus squid[133]: file_open: error opening file /d4/squid/cache/0
2/69/00001692: (23) File table overflow 
Dec 15 02:10:08 zeus squid[133]: storeSwapInStart: Failed for 'http://xyz.tvcom.
ru/99/12/100_2.jpg' 
Dec 15 02:10:12 zeus kernel: Unable to load interpreter
Dec 15 03:26:16 zeus xntpd[103]: kernel pll status change 89

Regards,
    Oleg
_____________________________________________________________
Oleg Bartunov, sci.researcher, hostmaster of AstroNet,
Sternberg Astronomical Institute, Moscow University (Russia)
Internet: oleg@sai.msu.su, http://www.sai.msu.su/~megera/
phone: +007(095)939-16-83, +007(095)939-23-83



pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: [HACKERS] Bug or feature? select, count(*), group by and empty tables
Next
From: Christof Petig
Date:
Subject: Re: [HACKERS] Volunteer: Large Tuples / Tuple chaining