If DNS is specified for hosts or ipnodes lookup in the /etc/nsswitch.conf file, and your system is running multithreaded applications, the nscd daemon might crash. This problem occurs because the nscd daemon slowly grows in size, and can consume up to 4 Gbytes of swap space. If all swap space becomes exhausted, then nscd might crash, and other new processes might suffer from various random errors.
Workaround: To avoid this problem, modify the nscd settings by following these steps.
Become superuser.
Stop the nscd daemon.
# /etc/init.d/nscd stop |
Add or modify the following lines in the /etc/nscd.conf file.
keep-hot-count hosts 0
keep-hot-count ipnodes 0
Restart the nscd daemon.
# /etc/init.d/nscd start |
Not running the nscd will cause the same leak to appear in any multithreaded applications doing host lookups.
The Solaris Point-to-Point Protocol (PPP) 4.0 daemon (pppd) might unexpectedly exit under the following conditions.
The pppd daemon is idle.
The demand and holdoff options to the pppd daemon are enabled.
A packet arrives during the holdoff period.
If the pppd daemon exits unexpectedly, a message similar to the following message is logged in the appropriate system log file.
current date hostname pppd[PID]: [ID 702911 daemon.error] unable to set IP to pass: Invalid argument current date hostname pppd[PID]: [ID 702911 daemon.error] unable to enable IPCP |
See the man page pppd(1M) for more information on the demand and holdoff options to the pppd daemon.
Workaround: Choose one of the following workarounds.
If you do not need the pppd daemon to wait before the daemon tries to re-initiate links, do not use the holdoff option with the pppd daemon.
Set the holdoff option value to 0.
Run the pppd daemon from a loop script similar to the following Bourne shell script.
#!/bin/sh while :; do /usr/bin/pppd cua/b lock idle 60 demand nodetach noauth \ 38400 10.0.0.1:10.0.0.2 holdoff 20 done
If you configure multiple IP tunnels between two IP nodes, and enable ip_strict_dst_multihoming or other IP filters, packet loss might result.
Workaround: Choose one of the following workarounds.
Configure a single tunnel between the two IP nodes and add addresses to the tunnel by using the ifconfig command with the addif option.
Do not enable ip_strict_dst_multihoming on tunnels between two IP nodes.