Thanks very much for your responses.
SUMMARY
Primary reasons for reboot is when there are upgrades, patches
or hardware changes. Of course when the power is shut.
Peter noted that there was a bug which hangs the server precisely
after 237 days in Solaris 2.5 (It is fixed now) and Sun recommended
to reboot monthly which is some time ago.
NO NEED TO REBOOT
There overwhelming majority response is
that there are servers runs for long time perfectly
and no need to reboot.
Mikey
David
Marc
Peter
Charlie
Tim
Todd
Rodeny
Rik
Paul
ct
Otto
Ronald
Stephen
Rsk
Graham
Al
Mike
Jay
John
Carsten
Shawn
Unknown (?)
DEPENDS ON THE APPLICATION
Couple of them responded to keep caution
depends on the nature of jobs running in that
server.
Rogerio noted that some of his server performance
suffers with no regular reboots due to the nature of
application he is running. But feels many servers are OK
with out reboots.
REGULAR REBOOT DO NOT HURT
And some suggested to reboot regularly if that
box can afford.
Rodney suggested reboots of time to time to
cleanup /tmp, process table or when applying patches.
MONTHLY REBOOT IS OK.
Bex
LIKE TO REBOOT WEEKLY IF WE CAN AFFORD
Michael
Marc likes to reboot weekly just to feel that the system
runs in "clean state" even though he believes most UNIX systems
do not need a reboot and this comes from idea of other inefficient
OS systems.
Robert feels that weekly reboots are good based on various factors
services or applications running on that system.
MY CONCLUSION FROM YOUR RESPONSES
SUN servers are pretty stable and do not need any reboots.
However depends on the type of jobs or usage on that server AND
if you can afford, schedule for periodic reboots.
There are many situtations we can not reboot the server after some
configuration change. And we may not find out the problem till the
server reboots. So periodic reboots might shed some light and it is
better to find out earlier than later because we might remember
the changes done to the system.
For example erroneous allocation of zero (0th)cylinder raw partitions
to some DB servers will screws up partition label and it shows up
after the reboot and corrupts the databases. So planned reboots might
have some sort of some backup strategy which might help us to recover
the databases with/without little data loss.
Also typically reboots are scheduled during off-hours or weekends will
provide some time window to fix the problems in case... instead user
watching you over your shoulder while fixing the problem.
Added benefits would be little cleanup of process tables,
/tmp, any memory leaks (if any), cron problems, unreleasble network
ports
(especially originates from PC networks), zombies cleanup and etc.
Some poorly developed applications chew up all cpu cycles which in turn
chews up Administrator time in troubleshooting. In my past experience I
have
seen many servers in data centers needed attention after the yearly
powerdown.
They are mostly hardware failures or Administrators mistakes.(Err is
human though)
So safe bet .... periodic reboot if you can afford .
Thanks again
Siva
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:51 CDT