Node Maintenance
[Previous] [Next]
[First] [Last]
Node Maintenance
A regular schedule of Calendar Server node maintenance is the best protection
against unscheduled down time and loss of data. By following the procedures
outlined below, an administrator should minimize problems and ensure the
smooth and uninterrupted running of the Calendar application within his
enterprise.
This chapter outline the following maintenance tasks:
Maintenance procedures
Daily monitoring procedures
The following system monitoring procedures should be performed on a
daily basis:
-
Check that all relevant daemons/services are operational.
-
UNIX
-
Select Server Status | Show Server Status from the Calendar Server
Manager menu.
-
Select the option(s) you wish to invoke by clicking the appropriate check
box(es).
-
Click OK to run the script.
-
This script can also be run from the command line. See unistatus
in Appendix G for complete instructions
and syntax.
-
NT
-
Open Properties from the Services menu. You will find the five Calendar
Services listed in the Services display box with a flag indicating the
service as "started" or no entry signifying that the service is stopped.
-
Check that ample space is left on the /users/unison directory
or file system. See Appendix A for more
information on calculating the storage requirements for your node.
-
Verify that the previous night's backup has run.
-
Check for the existence of a /users/unison/log/dbv.log. This file
is created only if there is a problem. If the file is present and is not
empty, you might analyze the contents and utilize the unidbfix
tool (documented in Appendix G), consult
your support provider, or refer to the Release Notes for Netscape's current
point of contact.
-
NT
-
Windows NT's Performance Monitor tool can be used to chart or log the performance
and activity of the Calendar Server services; while the Windows NT's Event
Viewer records any problems in running the application.
Daily maintenance procedures
A nightly backup of the Calendar Server database (i.e. /users/unison/db/
) is your best protection against database corruption, which may occur
as a result of a power failure or disk crash. While database corruption
is very rare, even under the aforementioned conditions, nightly backups
serve as a safeguard in the event that your database cannot be restored.
-
Warning:
-
Netscape Calendar Server must be shut down prior
to making a backup.
Monthly maintenance procedures
The following system maintenance procedures should be done after hours
on a monthly basis:
-
Verify the consistency of the Calendar Server database(s). See the unidbfix
script in Appendix G for complete instructions.
-
Archive the log files. Remember to shut down the Calendar Server before
archiving the log files and to restart the Calendar Server once the task
is completed.
Other maintenance procedures
-
The utility unirmold, documented in Appendix
G, should be used as needed to remove all events and tasks
that are older than a specified value from a Calendar Server node.
-
To view the elapsed time and CPU statistics for each client connection,
set the parameter stats = TRUE under [ENG] in the /users/unison/misc/unison.ini
file. To set this parameter, select Server Preferences| Manage Calendar
Server to open the "Manage Calendar Server" form, and follow the on-line
instructions. When a client connection is closed, stats results are appended
to the /users/unison/log/stats.log file. After analysis, set stats
= FALSE to disable logging.
Backup procedures
-
Warning:
-
The Calendar Server database must be shut down prior
to performing a backup.
Once the backup has been completed, the Calendar Server may be restarted.
To further protect valuable information and to minimize the amount of time
that the Calendar Server is unavailable to users during backups, it is
recommended that Calendar administrators consider the use of mirrored disks.
-
UNIX
-
The optimal backup procedure for UNIX servers:
-
Use RAID stripe/mirroring (called RAID 1+0 or RAID 10). This is important
for performance, reliability, and to enable Calendar Server backups without
down time.
-
Before the backup starts, you have to shut down the Calendar Server. Have
everyone log off before shut down, as it takes a long time to "disconnect"
each and every active user.
-
After the server is down, "break" one of the sub-mirrors. In ODS (Solaris-2.x)
you do that with metaoffline d2 d0 where "d2" is the
mirrored/striped disk, and "d0" is the striped sub-mirror to off-line.
-
Bring the Calendar Server up again. The Calendar Server is completely functional,
but only writing to one half of the mirror. The server will record what
is changing while the mirror is broken.
-
Mount the off-line sub-mirror on the filesystem, read only. For
example:
% mount -o ro /dev/md/dsk/d0 /mnt
-
Now perform the backup of /mnt. This disk is a snapshot of the
Calendar data. The backup is consistent, even though the Calendar Server
is still running on the "broken mirror".
-
After the backup finishes, unmount the sub-mirror:
umount /mnt
-
Put the sub-mirror on-line again, and it will automatically synchronize
with the changes made to Calendar data.
metaonline d2 d0
-
Note:
-
This solution means down time of only a few minutes.
Users should log off before the nightly (or weekly) backups, since it will
reduce the time required to shut down the Calendar Server.
If you cannot use mirrored disks, you must have the Calendar Server
down during the entire backup.
-
NT
A backup procedure for Windows NT:
-
Select Properties from the Services menu.
-
Select the Calendar Lock Manager and click Stop to stop all Calendar services.
-
Run Windows NT backup or another third party backup tool to perform a weekly
full copy backup of the /users/unison/ directory.
-
Perform a daily incremental backup of the /users/unison/ directory.
-
Repeat the process described in the second step above, but click Start
to restart each of the five Calendar Services.
Shutdowns
-
On the Server Administration Page, to the left of the server's name, is
an icon that can be toggled between OFF/ON. If the server is running, ON
is illuminated by a green light.
-
Using the Calendar Server Manager form:
-
Click the server button on the Server Administration Page to open the Calendar
Server Manager.
-
Select Server Preferences| Server On/Off to open the "Server On/Off"
form.
-
Click Start Server or Stop Server as appropriate.
-
Once the shutdown has been initiated, you can follow the progress of the
invoked script at the bottom of the form. The working status and any problems
encountered in stopping the Calendar Server daemons/services is displayed
during the execution of the scripts.
-
Click the hyperlink "Return to process control" to return to the "Server
On/Off" form.
-
UNIX
-
Shell script -- unistop. For full instructions on the use of this
script, see Appendix G.
-
NT
-
From the Services Control Panel -- stop the five Calendar Server services
(uniengd, unilckd, unicwsd, unisncd and unidasd)
-
Warning:
-
As a result of the number of user connections, the
number of processes running, and the volume of network traffic, Calendar
Server may take over a minute to stop.
[Previous] [Next]
[First] [Last]