
Signature
NewsGuy.Com 30Gb $9.95 Carry Forward and On Demand Bandwidth
Access is a canary in a coal mine for network data loss, and
if the developer is blaming the network, you need to spend
a lot of time checking everything. Like is it one client computer
in particular? Is someone kicking the power cord out? Do you
have a faulty patch cable or switch? Does turning off file caching
and optimistic locking solve the problem? Remember, you can't
depend on the users to tell you even that their computer is rebooting
every 15 minutes or half hour.
Having said that, the number one cause of data loss is user
error. Normally, they don't realise they are causing the problem.
The users don't like to hear that either. Sometimes they are
belligerent and wilfully ignorant when you try to identify
the problem. And that is the developers problem, not yours.
Still, do what you can to help: try to exclude network problems
from the range of possibilities.
(david)
> Hi All,
>
[quoted text clipped - 24 lines]
> Bob Smith
> Robert Smith Consulting
>Intermittantlly the users are missing data from their view of the
>database and are told by the 'developer' of the database that they
[quoted text clipped - 3 lines]
>You can see where this is going, I'm being blamed for the network
>instability
Of course.
No other messages? Such as "Disk or Network Error"? Does data
still appear for other tables? Only some records missing but other
records still there? When you do the refresh all records then
magically show up? How about if they exit the FE and then enter the
FE?
Something strange is going on but without additional error messages
I'm not sure they are your fault.
Tony

Signature
Tony Toews, Microsoft Access MVP
Please respond only in the newsgroups so that others can
read the entire thread of messages.
Microsoft Access Links, Hints, Tips & Accounting Systems at
http://www.granite.ab.ca/accsmstr.htm
Hi Tony,
I've been doing some checking on the 'so called' expert access
programmer that the company employed for this database.
I've found in the linked table manager that the mapping of the links
is both by mapped drive and by UMC mapping ( \\servername\directory\)
From what i've read in the newsgroups and in a few books i've found is
that the mapping has to be one or the other and preferably using
the computer name\directory.
as to error messages, I'm having the users watch for error messages
and making screen prints of the messages so I will have some kind of
a record to trouble shoot this problem.
When data is missing, all the data is missing, ie; if a telephone
number is missing from one record, it's missing from all the records
using a telepone number.
The programmer started having everyone to a database repair before
using the database, then she started telling everyone to do the linked
table manager refresh just lately. This does bring the 'missing'
data back, but it takes time to rebuild the links and people are
complaining..
I've written a few databases in DOS and windows using first foxbase
and later fox for windows for a parts department. I was a new car
dealer from 1964 to 2000 and about 1983 Chrysler didn't have anything
to do a computer database so I was on 5x8 cards. I figured I could
beat this and taught myself Dbase III and started programming. When
we sold the business in 2000 the same computer based database was
still being used with 105k records with sales and order histories,
supercedence, notes, sales etc and it was running on a 500 mb hard
drive in multiuser mode. I never had the problems i've seen with the
access database even through I was using up to 15 different indexes.
I'll check some more tomorrow when no one is using the database and
see what I can find out.
I'm pretty sure most of the problem is the linked table manager using
both mapped drives and \\computername\directory linking. But I want
to be sure before I write a report to the owner about what I've forund
'wrong' with the network One thing I know for sure is that the
network is fast, I can see <1ms times between computers and 75-85mb
through-put between machines.
The machine that the 'BE' of the database is running on is a 2.8ghz P4
with 2gb of memory. This machine as a mirror raid hard drive setup
using 2@ WD 7200 rpm drives. The computers are tied together using a
16 port 10/100 ethernet switch with the longest cat5 cable being about
50' long. I ran a cert check on the network and everything passed
with flying colors so the hardware appears to be solid.
I'll get the checking down tomorrow and post the info tomorrow nite,
Thanks for the reply
Bob Smith
Robert Smith Consulting
>Hi All,
>
[quoted text clipped - 24 lines]
>Bob Smith
>Robert Smith Consulting

Signature
NewsGuy.Com 30Gb $9.95 Carry Forward and On Demand Bandwidth
Andy - 23 Apr 2006 14:06 GMT
> I'm pretty sure most of the problem is the linked table manager using
> both mapped drives and \\computername\directory linking. But I want
> to be sure before I write a report to the owner about what I've forund
> 'wrong' with the network One thing I know for sure is that the
> network is fast, I can see <1ms times between computers and 75-85mb
> through-put between machines.
I maintain several databases on a shared network and have "mixed" links
all the time. Using a drive name versus the URL path does not solve your
problem. I would not waste my time there. You can't go wrong using the
computer/path name. On the network I use, there is a universal
understanding of the some the mapped drive settings.
You might want to try a persistent connection to the backend. I believe
this simply is a connection to a table in the back end database that is
opened on startup and closed on exit.
hth,
Andy
david@epsomdotcomdotau - 23 Apr 2006 15:11 GMT
Find out if the developer has made any design changes to the
tables recently. New fields, changed field order, that kind of
thing.
A missing field, restored by a re-link, sounds like the links in the
FE did not match the data structure in the BE. This only
becomes a problem after compacting the BE, and is fixed
by re-linking or compacting the FE.
Note that this does not let you off the hook. If the problem
came up because of data corruption, requiring compact/repair,
then it was still data corruption requiring compact/repair.
(david)
> Hi Tony,
>
[quoted text clipped - 85 lines]
> >Bob Smith
> >Robert Smith Consulting
Tony Toews - 24 Apr 2006 00:58 GMT
>I've found in the linked table manager that the mapping of the links
>is both by mapped drive and by UMC mapping ( \\servername\directory\)
>
>From what i've read in the newsgroups and in a few books i've found is
>that the mapping has to be one or the other and preferably using
>the computer name\directory.
Well, I wouldn't mix and match them. But to me using either the
drive letter or the UNC mapping should make no difference.
>as to error messages, I'm having the users watch for error messages
>and making screen prints of the messages so I will have some kind of
>a record to trouble shoot this problem.
Good.
>When data is missing, all the data is missing, ie; if a telephone
>number is missing from one record, it's missing from all the records
>using a telepone number.
Now that kind of corruption I've never heard of. Missing records
after a compact sure. Records with #error or #deleted in memo or
other fields fine. But missing an entire column? I don't recall any
postings mentioning that symptom ever. And I read almost all postings
that mention the word corruption for the six or ten years in various
newsgroups such as this one or comp.databases.ms-access.
>The programmer started having everyone to a database repair before
>using the database, then she started telling everyone to do the linked
>table manager refresh just lately. This does bring the 'missing'
>data back, but it takes time to rebuild the links and people are
>complaining..
I would tentatively agree with david's suggestion that the developer
is renaming the fields, moving them around, etc, etc. That said that
kind of problem has produced error message -1517. Hmm, but now that
I think about it there were other wierdnesses too such as combo boxes
missing data.
The following is my standard blurb on error -1517.
"I've noticed this problem several times when I've added a foreign key
to a table in the backend which was inserted into the middle of a
table. Everything would work fine for days or weeks until the backend
was compacted. Then the FE would puke with the -1517 error whenever
that particular table was accessed. But deleting the link and
recreating the link or compacting the FE made it work again.
http://www.granite.ab.ca/access/reservederror1517.htm"
Getting back to the original statement that this is caused by a faulty
network is rubbish given that relinking the tables brings back the
missing data.
Tony

Signature
Tony Toews, Microsoft Access MVP
Please respond only in the newsgroups so that others can
read the entire thread of messages.
Microsoft Access Links, Hints, Tips & Accounting Systems at
http://www.granite.ab.ca/accsmstr.htm
Sylvain Lafontaine - 26 Apr 2006 19:01 GMT
Using a drive letter instead of an UMC mapping (or mixing them) doesn't seem
a very good idea; however, I don't remember anyone having a corruption
problem because of that. Others possibilities could be:
1- Each user don't have their own copy of the frontend (a common error).
2- The developer have made some design changes and didn't told you (but this
seems unlikely to me).
3- Your network is unstable. (Access is incredibly sensitive to any network
instability and you cannot use Access over anything but a rock solid
network. Doing a cert check on the network is not a bad thing but this is
not sufficient to prove that this network is really sufficiently solid for
Access.)
4- your network is stable but is sometime saturated for an extended period
of time (more than a few seconds?) because some guy is transferring a very,
very big file or is making a backup over the network.
When you speak about a missing column, could be this because a whole table
is missing from a JOIN or if you can see other columns from the same table?
Beside the solutions from the other posts, here are two more:
1- Use TS or Citrix.
2- Replace the MDB backend with MSDE or SQL-Server.

Signature
Sylvain Lafontaine, ing.
MVP - Technologies Virtual-PC
E-mail: http://cerbermail.com/?QugbLEWINF
>>I've found in the linked table manager that the mapping of the links
>>is both by mapped drive and by UMC mapping ( \\servername\directory\)
[quoted text clipped - 51 lines]
>
> Tony
Tony Toews - 26 Apr 2006 20:50 GMT
"Sylvain Lafontaine" <sylvain aei ca (fill the blanks, no spam
please)> wrote:
>Using a drive letter instead of an UMC mapping (or mixing them) doesn't seem
>a very good idea;
<shrug> Using driver letters vs UNC each have their pros and cons.
One con with using UNC mapping is if the server name changes then you
have to relink the tables. If you have drive letter mapping then a
network login script takes care of that problem.
Tony

Signature
Tony Toews, Microsoft Access MVP
Please respond only in the newsgroups so that others can
read the entire thread of messages.
Microsoft Access Links, Hints, Tips & Accounting Systems at
http://www.granite.ab.ca/accsmstr.htm
Sylvain Lafontaine - 26 Apr 2006 22:01 GMT
Yes but a server name change should be something rare and beside, when the
server or its name change, there are usually a lot of other changes. Using
drive letters looks like to be an easy solution but they are strongly
associated with each logon; something that can become very confusing on some
installations.
However, the main point to consider for this thread is that the use of
either driver letter or UMC mapping or both of them is very unlikely to be
the source of the problem here. (But we never know.)

Signature
Sylvain Lafontaine, ing.
MVP - Technologies Virtual-PC
E-mail: http://cerbermail.com/?QugbLEWINF
> "Sylvain Lafontaine" <sylvain aei ca (fill the blanks, no spam
> please)> wrote:
[quoted text clipped - 9 lines]
>
> Tony
Douglas J. Steele - 26 Apr 2006 23:52 GMT
Sylvain: This topic comes up fairly often. It all depends on what sort of an
environment you're working in.
I work for a large multi-national corporation. We have an extremely
controlled environment. All users run a common logon script which maps their
drives consistently. I know that programs will always be on the J: drive
(and that all folders there will be read-only) and that data can be put on
the I: drive (which may or may not be read-write: it depends on what groups
the user for which folders they can write to)
Let's assume we're developing an application that's going to be used by
different work groups in different locations. There's no intent to share
data from work group to work group. I know I can always link my tables to
I:\MyApp\MyFile.MDB, regardless of whether the user is in Houston or
Toronto. It would be impossible for me to use a UNC, as I do not know in
advance what server is appropriate for my user.
Yes, there are times where UNCs are better, but not (usually) at my company.

Signature
Doug Steele, Microsoft Access MVP
http://I.Am/DougSteele
(no private e-mails, please)
> Yes but a server name change should be something rare and beside, when the
> server or its name change, there are usually a lot of other changes.
[quoted text clipped - 19 lines]
>>
>> Tony