Since I answered the first time, let me see if I can continue the dialog
intelligently, even though it is Friday afternoon!
I agree - I like the the randomization, and would like to see something more
than 50% too. Turning off randomization would mean that most of your backups
would be occuring at the same time, causing contention on both the network and
the server. You would also have to allow for the maximum number of scheduled
sessions, but that's a different issue.
I have seen a few rare situations where a scheduled session is interrupted, but
it is not marked as failed, and when the session is reestablished, if the window
is open, the session will take place.
Also to confuse more, the 50% number also is based on the remainder of the
window, not the total window. For example. assume a two hour window starting
at 10:00. The normal 50% randomization would cover the time of 10-11. However,
if I start a scheduled session at 11:00, the 50% randomization is now applied to
the remainder of the window, meaning that the time available for randomization
will be from 11:00 till 11:30.
Also, sometimes the randomization process will "pass" on the randomizing, if the
amount of window left to work on is too small. For example, using the same
10:00 to 12:00 (PM) window example as before, let's assume a scheduled session
is established at 10:15 AM. Randomization **could** fit into the window, but
the randomization process will sometimes "pass" on this, and schedule a "retry
in 12 hours" pass. Then, at 10:15PM, the randomization will occur, based on 50%
of the remaining 1hour and 45 minutes.....
Let's look at some examplesfrom client 2....
> 07/19/97 10:12:01 Command will be executed in 10 hours and 41 minutes.
> 07/20/97 08:56:13 Schedule will be refreshed in 12 hours.
> 07/20/97 20:56:13 Command will be executed in 1 hour and 57 minutes.
In line 1, the scheduler is started at 10:12 - The window, which starts at 7:00
(19:00) is within range, and so the server tells the client to go play in the
traffic, and come back in 10 hours and 41 minutes, which works out to be 20:59
In line 2, the scheduler is started at 08:56. The window, which starts at
19:00, is within range, but the server evidently feels that the amount of
randomization available is not sufficient tio schedule a session. (At this
time, 1 hour and 56 minutes of a 12 hour window is available; 50% would yield 58
minutes... so therefore the decision is made to set a 12 hour retry.
In line 3, we see the client coming back 12 hours later. At this time, 10 hours
and 4 minutes of window remain, so the return time of 1 hour and 57 minutes from
then (22:53 if my math is correct) is set.
BTW, I have seen situations also, where the amount of time remaining in the
window is so small, that the backup commences immediately.
Lastly, assuming that the backups for these machines do not drag on and on, the
message "will try again in 12 hours should occur at the end of the scheduled
backup.
Hope this helps clarify....
Jerry Lawson
jlawson AT thehartford DOT com
PS - one last point.... The retry in 20 minutes is not generated from the server
it comes from teh client when he cannot make a connection to the server. For
example, the client wakes up and calls the server, but the server is down for
maintenance. The client should then do the will try again in 20 minutes
(referred to as the "20 minute vamp" here) until the server is back in business.
At that time, if the window is still open, the session should be established,
based on randomization of the remaining window.
JL
______________________________ Forward Header __________________________________
Subject: Randomizing Schedule Start Times
Author: INTERNET.OWNERAD at SNADGATE
Date: 7/25/97 1:07 PM
Hi y'all,
On Wed, 23 Jul 1997 16:45:03 -0400 Jerry Lawson wrote:
> If you take the default (Which is Schedmode=polling), then what happens is
this: > > 1. You start the scheduler task on the client. > 2. The scheduler
task contacts the server. > 3. The server either tells him to come back in 12
hours (if the schedule window > is more than 12 hours away), or tells him when
to come back to do the backup. >
If randomization is not 0, does the server tell him to come back in 12 hours
IF THE RANDOMLY SCHEDULED TIME THAT THE SERVER CALCULATES FOR COMMAND EXECUTION
is more than 12 hours away?
For example, these 2 clients are associated with a schedule that starts at 7PM
and has a duration of 12 hours. Randomization is 50%.
client #1's daily "will be execute|refreshed":
07/19/97 08:20:18 Schedule will be refreshed in 12 hours.
07/19/97 20:20:16 Command will be executed in 4 hours and 7 minutes.
07/20/97 12:29:04 Command will be executed in 7 hours and 33 minutes.
07/21/97 08:03:41 Schedule will be refreshed in 12 hours.
07/21/97 20:03:40 Command will be executed in 1 hour and 50 minutes.
07/22/97 09:55:35 Command will be executed in 11 hours and 44 minutes.
07/23/97 09:42:08 Command will be executed in 11 hours and 42 minutes.
07/24/97 09:26:14 Command will be executed in 9 hours and 49 minutes.
07/25/97 07:17:03 Schedule will be refreshed in 12 hours.
client #2's daily "will be execute|refreshed":
07/19/97 10:12:01 Command will be executed in 10 hours and 41 minutes.
07/20/97 08:56:13 Schedule will be refreshed in 12 hours.
07/20/97 20:56:13 Command will be executed in 1 hour and 57 minutes.
07/21/97 10:57:09 Schedule will be refreshed in 12 hours.
07/21/97 22:57:12 Command will be executed in 1 hour and 44 minutes.
07/22/97 12:44:02 Command will be executed in 8 hours and 20 minutes.
07/23/97 09:07:17 Schedule will be refreshed in 12 hours.
07/23/97 17:08:35 Command will be executed in 7 hours and 25 minutes.
07/24/97 12:36:02 Command will be executed in 11 hours and 35 minutes.
Also, I'm trying to figure out if I should be running with randomization on.
I like the fact that the server randomizes start times - I'm a bit leary
about turning it off because of the warning re. communications errors in
the Admin Guide, but don't like the fact that I can't set randomization
higher than 50%.
And, I don't understand why the 50% limit "ensures that half of the
startup window is available for retrying scheduled commands that have
failed." Whenever I have a client failure, it retries in 20 minutes,
regardless of what half of the startup window it's in.
Can I get any free (or almost free) advice from anyone? ;-)
Thanks,
Michael
|
