Red Hat Bugzilla – Bug 6676
Non printing of timestamp if run through a named pipe
Last modified: 2014-03-16 22:11:01 EDT
I have a setup on my syslog such that it writes to a named pipe on the disk, which is read by a PERL script, and printed onto a tty. It would seem that some entries on this don't get the timestamp or computer name printed (I use the -r option to syslogd as well). It seems random as to whether a given entry gets printed with the timestamp or not. If this happens, then it would appear not to get logged to the messages file. In addition, I have had problems where log events appear to be queued in /dev/log, and the system suddenly prints them all when I restart syslog.
This seems to happen most under the logging done by the sshd on the machine, with the 1.2.27-5i releases from ftp.replay.com . I understand that these are unsupported, but I don't understand how the messages are appearing without timestamps if they are coming through the syslog mechanism, as they must be to appear on the display.
Of course, if more log data than will fit in the pipe buffer accumulates before the pipe is read, some of the data won't be put in the pipe. Aside from that, I can't reproduce this here. What does the perl script look like?