I have a syslogd issue. More detail is as follows.

I have a hos with uptime 247 days. And got a syslogd memory leak I think. and the syslogd consume memory as follows.

# ps -ef -o fname=Name -o vsz=vmem -o rss=ram | grep syslo
syslogd 2826384 123728

According to this the total size of the syslogd process is about 2,8GB and the memory resident is 123MB.

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++=
If I look the corresponding parts from an other host the figures are totally different. Also the uptime of the troublesome host is not a problem as this host, without problems, has been up for 677 days compared to the 247 days.

# uptime
9:00am up 677 day(s), 5:38, 2 users, load average: 10.09, 12.52, 12.74 # ps -ef -o fname=Name -o vsz=vmem -o rss=ram | grep syslo
syslogd 3568 1688


I wonder why the host with uptime 247 days will over-use the memory.
How can I troubleshoot it? use gcore to generate a process core file? Can you give me some advices?

What Solaris version is this ?
Are both machines the same patch level ?

You can restart syslogd, use libumem with it to investigate further, see http://developers.sun.com/solaris/ar...m_library.html

Sorry to told you more details about system.

Solaris 8 2/02
KUP is 108528-27
syslogd patch level is 110945-08 which is the latest one.

I am not sure whether KUP upgrade will help.

I have seen your recommended libumem which is shipped with solaris5.9. I wonder whether it is applicable to 5.8.

Kernel update would unlikely change anything, as this isn't a kernel leak.

libumem isn't applicable to Solaris 8.

You can workaround the problem by restarting syslogd with something like: /etc/rc.d/init.d/syslog restart