On December 22, 2006 we had our server fail. I installed a new hard
drive into the existing machine and reloaded D3 Proplus 7.2 and
restored the data accounts. After that the machine was rebooting in
the middle of the night every now and again for no reason. After each
reboot we lost all the data execpt from December 21st and earlier. So
we got a new server (well a used server since Proplus 7.2 cannot run on
new technology) and I reloaded the new server with D3 Proplus 7.2 and
restored the data accounts, and again when ever we reboot the server,
the data that exists is from December 21st and earlier. Can anybody
explain what is going on here? I don't think the data can be stored in
RAM because that is a lot of data.

Thanks in advance,
Walter

I don't suppose you have an automated procedure that is re-loading the
data (rather than backing up) every time you reboot?

W>E>I>R>D

Walt wrote:
Walt,

Not that this is much help but I had a similar experience with a
client's D3 NT system back in 2001. On the 29th May their server froze
& they rebooted it. After the reboot the system appeared to have
regressed to the 3rd May - i.e. there were no transactions dated after
the 3rd. Everything was intact - it was as though they had restored
from the file-save on the 3rd. However, this was not possible as they
hadn't been doing file-saves & in fact had no valid backup. The only
possible explanation was that everything after the 3rd was held in
memory & not flushed to disk. This hardly seems likely but I think I
prefer to the time-warp theory.
At least it wasn't ongoing like yours.

Geoff

I recently reported a memory leak in D3NT v7.4.x and these other notes
we see here might be related. D3 seems to consume small amounts of
memory during File Open operations. It becomes remarkable if lots of
these are done consecutively.

The memory is released quickly if the user executes Exit, but if the
user executes Off it releases very slowly. This memory is also
released when D3 is shutdown, but very slowly as with an Off. It will
release about 100MB over a long period (1/2 hour or so, YMMV) and then
release/flush everything else all at once.

Points:
1) I don't know if the memory being released slowly takes precedence
over memory that needs to be flushed. It's possible that something is
bottlenecking the flusher.
2) Whenever possible, try to use Exit rather than Off.
3) Do NOT terminate the D3 shutdown process abnormally or you may lose
the handle to all of the memory that it has to release, including your
data updates. If it looks like it's hanging during shutdown, leave it
alone! Watch memory on another port and you might see the available
values going up slowly.
4) If you need to logoff or kill processes, and the process you are on
seems to be hung itself, leave it alone and see if memory is being
released.
5) For the same reasons, do NOT crash or shutdown Windows if you can
help it without doing a clean D3 shutdown.
6) For the same reasons, if a user has been doing a lot of work and
then they leave using Off, they may not get a login prompt
immediately. This is because memory is being released. Do NOT
interrupt this process, kill it, break the connection, or otherwise
stop that release of resources. I understand it's impossible to
control this completely.
7) Shutdown D3NT once in a while. Notice the memory consumed on boot
and monitor that over a period of time.
8) If you have processes that are running all the time, maybe phantoms
that open files frequently, get them to periodically do a graceful
termination and restart them on another PIB. Watch memory when these
processes terminate.
9) Take all of this with a grain of salt. I'm not sure if there is
some other combination of factors that trigger the issue, like if it's
XP only or if there are less than X PIBS configured or only with patch
Y or only if the moon is full. If this helps someone, great, if not,
don't get into a panic. I have no idea what is flushing or when so
this might be completely unrelated to the "back in time" issue.
10) I don't have 10 points but what good is a 9 point list?

Speaking of "back in time", hey Geoff, since your server failed on Dec
22, 2006 (see below), why don't you just replace it in November and
prevent it from happening? Hmm, jumping into next week and Googling
this thread I see someone cited the grandfather paradox, good point.
Oh, that's right, you're in Australia and you guys are a time zone
ahead of us poor yanks. Now I understand. BTW, is NASDAQ up or down
later this year?

I hope all is well.
Tony
TG@ smart.alec.with.a.short.memoryNebula-RnD.com


"Geoff Goodchild" wrote:

Goo'day,

On 9 May 2006 13:29:00 -0700, "Walt" <wboehrer (AT) yahoo (DOT) com> wrote:

Mate, if your server actually failed on that date, please send it to
me. I need it for working out next week's lotto numbers....

<huge grin>.... and <apologies>

Regards,

Bruce Nichol
Talon Computer Services
ALBURY NSW Australia

http://www.taloncs.com.au

If it ain't broke, fix it until it is....

Walt wrote:
Saw something similar to this on an Ultimate system with a removable-
platter hard drive; there was a button on the drive to write-protect(!)

--
frosty

I know their is a parameter inside linux that can be adjusted and will
alter the throughput from and to the hard disk (ATA drives???)
dramatically. Now I don't know what or where this parameter is but
maybe a linux guru could point in the right direction.

I'm grasping at straws here... Could this parameter a speak of if set
incorrectly cause this anomoly?

Regards,

Dale

WALT,

What is the current status ?