On Thu, 2005-05-19 at 11:01 -0700, Mike Tremaine wrote:
On Thu, 2005-05-19 at 08:12, Mike Tremaine wrote:
On Wed, 2005-05-18 at 23:51, Balazs Scheidler wrote:
On Wed, 2005-05-18 at 17:53 -0700, Mike Tremaine wrote:
In looking through yesterdays logs I can confirm that this is only happening to sendmail logs. But like I said I just installed syslog-ng 2 days ago and was not having this issue with the standard syslog.
Just to follow up a little. I tried tweaking a few configuration options [sync, fifo, max connections] but with out success. I also tried two older version syslog-ng-1.6.5-4 and syslog-1.6.5-2. The problem [for me atleast] exists under these version also.
Very odd, almost like sendmail is not giving a proper EOL. I looked in sources.c [do_read_line] but my c is weak and nothing jumped out at me.
I'm open to suggestions, I'm also willing to move up to 1.9 if this is not worth debugging.
If your log volume is bearable, it would help if you could attach strace to the syslog-ng process with a large -s parameter while a broken message is received. e.g. something like: strace -s 4096 -o syslog-ng.trace -p <syslog-ngpid> That would help to identify whether syslog-ng or sendmail is at fault. Standard syslogd uses unix-dgram sockets, but if sendmail uses the standard syslog() routines from libc it should simply work with unix-stream() as well. -- Bazsi