Return-Path: carole@computer-science.manchester.ac.uk Delivery-Date: Thu, 25 Mar 93 10:37:27 -0800 Return-Path: carole@computer-science.manchester.ac.uk Received: from sun2.nsfnet-relay.ac.uk by postgres.Berkeley.EDU (5.61/1.29) id AA27343; Thu, 25 Mar 93 10:36:08 -0800 Via: uk.ac.manchester.computer-science; Thu, 25 Mar 1993 18:20:03 +0000 Date: Thu, 25 Mar 93 18:01:43 GMT From: Carole Goble Message-Id: <9303251801.AA06622@r1o.cs.man.ac.uk> To: aoki@postgres.berkeley.edu Subject: Re: sun problem Cc: boonteet@computer-science.manchester.ac.uk, cag@computer-science.manchester.ac.uk, rick@computer-science.manchester.ac.uk Paul Okay, we have a crash. Neil McQuaig doesn't help us as we are bsd. The first part of these notes has the Postmaster process failing but we couldn't get a stack trace. We rebooted the machine and increased the number of file descriptors. Then we hit ANOTHER problem where the postmaster keeps going but the postgres process exits when an attempt to REPLACE or similar is made without dumping core. Any help/guidance/debugging hints appreciated. Only one user (Mr Boontee) on who did the following: ---------------------------------------------------- * define rewrite rule sharon_too_much is on replace to emp.salary where current.name="sharon" and new.salary > emp.salary and emp.name = current.manager do instead replace emp (salary=current.salary) where current.name="sharon" \g Query sent to backend is "define rewrite rule sharon_too_much is on replace to emp.salary where current.name="sharon" and new.salary > emp.salary and emp.name = current.manager do instead replace emp (salary=current.salary) where current.name="sharon" " APPEND 30048DEFINE Go * replace emp (salary=500) where emp.name="sharon" \g Query sent to backend is "replace emp (salary=500) where emp.name="sharon" " Error: No response from the backend, exiting... On the server: --------------- PostMaster: fork failedprogram terminated by signal ILL (bad stack) I typed (dbx) trace instead of (dbx) where which gave me a segmentation fault. Sorry. That was finger trouble. t6% dbx postmaster Reading symbolic information... Read 3481 symbols warning: core file read error: address not in data space warning: core file read error: address not in data space attempt to read stack failed - bad frame pointer Processes now running on server: -------------------------------- Note the lack of Mr Boontee's postgres backend and the postmaster process: t6% ps aux USER PID %CPU %MEM SZ RSS TT STAT START TIME COMMAND root 0 0.0 0.0 0 0 ? D Mar 15 0:47 swapper root 1 0.0 0.0 52 0 ? IW Mar 15 0:00 /sbin/init - root 2 0.0 0.0 0 0 ? D Mar 15 0:04 pagedaemon root 108 0.0 0.0 56 0 ? IW Mar 15 0:11 rpc.bootparamd root 61 0.0 0.0 68 0 ? IW Mar 15 1:23 portmap root 3429 0.0 0.0 24 0 ? IW Mar 20 0:00 sh -c /usr/lib/sendmail -oeq -oi r root 68 0.0 0.0 40 0 ? IW Mar 15 0:00 keyserv root 79 0.0 0.0 16 0 ? I Mar 15 0:00 (biod) root 78 0.0 0.0 16 0 ? I Mar 15 0:00 (biod) bin 66 0.0 0.0 36 0 ? IW Mar 15 0:00 ypbind root 168 0.0 0.0 40 0 co IW Mar 15 0:00 - std.9600 console (getty) root 90 0.0 0.0 60 0 ? IW Mar 15 0:01 syslogd root 117 0.0 0.2 108 56 ? S Mar 15 0:06 automount root 101 0.0 0.0 68 0 ? IW Mar 15 2:55 rpc.mountd -n root 102 0.0 0.0 28 0 ? I Mar 15 9:18 (nfsd) root 104 0.0 0.0 28 0 ? I Mar 15 9:43 (nfsd) root 105 0.0 0.0 48 0 ? IW Mar 15 0:14 rarpd -a root 106 0.0 0.0 24 0 ? IW Mar 15 0:03 rarpd -a root 6715 0.0 0.0 56 0 ? IW Mar 22 0:00 cron root 111 0.0 0.0 84 0 ? IW Mar 15 0:00 rpc.lockd root 112 0.0 0.0 52 0 ? IW Mar 15 0:00 rpc.statd root 9931 0.0 0.1 24 24 ? S 12:10 0:00 in.rlogind root 146 0.0 0.0 100 0 ? IW Mar 15 3:51 /usr/oracle/bin/orasrv oracle 138 0.0 0.8 268 232 ? S Mar 15 0:10 ora_pmon_oracle4 root 154 0.0 0.0 12 4 ? S Mar 15 96:16 update root 6716 0.0 0.0 24 0 ? IW Mar 22 0:00 sh -c /usr/lib/sendmail -oeq -oi r oracle 139 0.0 1.0 240 304 ? S Mar 15 1:46 ora_dbwr_oracle4 oracle 140 0.0 0.8 232 236 ? S Mar 15 1:03 ora_lgwr_oracle4 oracle 141 0.0 0.0 296 0 ? IW Mar 15 23:02 ora_smon_oracle4 root 157 0.0 0.2 56 60 ? S Mar 15 0:02 cron root 163 0.0 0.0 48 0 ? IW Mar 15 0:03 inetd oracle 9920 0.0 4.8 296 1488 ? I 12:07 0:00 oracleoracle4 T:I,,5 root 166 0.0 0.0 52 0 ? IW Mar 15 0:00 /usr/lib/lpd root 3428 0.0 0.0 56 0 ? IW Mar 20 0:00 cron postgres 17338 0.0 0.0 48 0 p0 IW Mar 17 0:00 -csh (csh) root 173 0.0 0.0 144 0 ? IW Mar 15 1:19 rpc.rquotad root 17337 0.0 0.0 24 0 ? IW Mar 17 0:00 in.rlogind root 6599 0.0 0.7 2364 208 ? S Mar 21 0:05 -(null) To t6 (sendmail) root 3430 0.0 0.7 3036 208 ? S Mar 20 0:06 -(null) To t6 (sendmail) postgres 9936 0.0 1.4 188 420 p1 R 12:12 0:00 ps aux oracle 9923 0.0 0.0 184 0 ? IW 12:08 0:00 oracleoracle4 T:I,,5 root 6717 0.0 0.7 1688 208 ? S Mar 22 0:03 -(null) To t6 (sendmail) root 6598 0.0 0.0 24 0 ? IW Mar 21 0:00 sh -c /usr/lib/sendmail -oeq -oi r postgres 9932 0.0 0.7 48 216 p1 S 12:10 0:00 -csh (csh) oracle 7881 0.0 0.0 424 0 ? IW Mar 23 0:02 oracleoracle4 T:I,,5 postgres 18119 0.0 0.0 256 0 p0 IW Mar 17 0:00 dbx oracle 9899 0.0 0.0 308 0 ? IW 11:37 0:00 oracleoracle4 T:I,,5 root 6597 0.0 0.0 56 0 ? IW Mar 21 0:00 cron We reboot the server and increase the number of file descriptors. So now we get.... Welcome to the C POSTGRES terminal monitor Go * retrieve (emp.all) \g Query sent to backend is "retrieve (emp.all) " ------------------------------------------------------------------------------------- | location | age | name | salary | manager | dept | ------------------------------------------------------------------------------------- | (11,10) | 20 | bill | 1000 | sharon | toy | ------------------------------------------------------------------------------------- | (11,10) | 31 | carole | 6000 | sharon | toy | ------------------------------------------------------------------------------------- | (15,12) | 25 | sharon | 500 | sam | shoe | ------------------------------------------------------------------------------------- | (10,5) | 30 | sam | 300 | bill | candy | ------------------------------------------------------------------------------------- Go * replace emp (salary=5000) where emp.name="sharon" \g Query sent to backend is "replace emp (salary=5000) where emp.name="sharon" " NOTICE:Mar 25 17:37:58:I have been signalled by the postmaster. NOTICE:Mar 25 17:37:58:Some backend process has died unexpectedly and possibly NOTICE:Mar 25 17:37:58:corrupted shared memory. The current transaction was NOTICE:Mar 25 17:37:58:aborted, and I am going to exit. Please resend the NOTICE:Mar 25 17:37:58:last query. -- The postgres backend Go * replace emp (salary=5000) where emp.name = "sharon" \g Query sent to backend is "replace emp (salary=5000) where emp.name = "sharon" " r1o> The postgres process has been removed, the postmaster is still running but the postgres process hasn't dumped core.