On Thursday 13 February 2003 10:56, Balazs Scheidler wrote:
On Thu, Feb 13, 2003 at 09:54:52AM +0100, Peter Bieringer wrote:
Pls. don't tell me that I have to restart every service after I restart syslog-ng (already known: crond, now postfix, which one is still missing?). This would be a very unnice "feature". If bug is in a common used glibc-version, then imho a workaround by syslog-ng would be required. I don't like to loose loglines, you sure not, too.
Does this also happen if old syslogd would be used? Don't believe so.
It's all very strange...looks like I have to switch back to syslog, loosing all the nice features :-(
the same happens to old syslogd as well as this is a libc issue. just try it and you will see.
Sorry, but at least on a Linux 2.4 system this is wrong! I just tested it: With the old syslogd 1.4.1 no message will be lost: Feb 13 13:50:16 test ./syslog_test[183]: 21 Feb 13 13:50:21 test ./syslog_test[183]: 22 Feb 13 13:50:23 test exiting on signal 15 Feb 13 13:50:24 test syslogd 1.4.1: restart. Feb 13 13:50:26 test ./syslog_test[183]: 23 Feb 13 13:50:31 test ./syslog_test[183]: 24 The same with syslog-ng 1.5.25: Feb 13 13:34:03 s_local@Elwing ./syslog_test[2118]: 71 Feb 13 13:34:08 s_local@Elwing ./syslog_test[2118]: 72 Feb 13 13:34:08 s_local@Elwing syslog-ng[2204]: syslog-ng version 1.5.25 going down Feb 13 13:34:09 s_local@Elwing syslog-ng[2252]: syslog-ng version 1.5.25 starting Feb 13 13:34:18 s_local@Elwing ./syslog_test[2118]: 74 Feb 13 13:34:23 s_local@Elwing ./syslog_test[2118]: 75 Message 73 is missing! The test program "syslog_test.c": #include <syslog.h> int main( int argc, char *argv[] ) { int cnt=0; openlog( argv[0], LOG_NDELAY|LOG_PID, LOG_DAEMON ); while( 1 ) { syslog( LOG_INFO, "%d", ++cnt ); sleep( 5 ); } return( 0 ); } All programs (syslog-ng, syslogd, syslog_test) are linked with glibc-2.3.1. Achim