link tables on a map network dirve

[binglis9]

I have a file server running debian 10 using iptables. Contected is 4 windows 10 desktops with microsoft access database front ends. The tables are stored on a map drive on windows 10 which points to a shared folder where the tables are stored on the file server.
All works how it should and no problems.

I have built a new testing server with debian 11 bullseye using nftables. It is the excact same setup but the only difference is nftables. I first tried it using iptables but it didnt work. Had the same problem that follows. The 4 desktops connect ok to the map drive and can read tables ok. If one desktop makes a change to a text field the changes show up on all other machines. If another machine makes a change to the same field same record it shows the change on the two other machines but the first machine that made the orginal change the new edit doesnt show. I can reload the records from the start and all changes are shown correctly. I know from past experience that this alone means the database is not working correctly. Every field is unbound to the database and only when you make a change to the field it writes back to tables and the change is made in the tables correctly.

I had this problem when I tried to update to bulleyes when first released. I re-installed buster debian 10 as it works there with no problems. I now have more time to investigate and to find out why this happens.

I have now created a table that logs any updates on any field in each record. this is timestamped and the user who made the change and logs old data. When I make a change all logs are correct and the database table shows correct data.
This now makes me think its a cache problem. I cant understand why it works with iptables and buster but not with nftables and bullseye.
Has anyone any views on this. I just seem to be going round and round in circles.

[binglis9]

does anyone else use a linux file server using samba to store their access tables on. The tables are stored via a map drive that points to the tables on the linux server. I have had this working fine for over 10 years but its only now that things dont seem to work properely. I tried to upgrade to a different plateform and records disappear and edits dont seem to work properely. Mapping a drive on each desktop has worked in the past but it looks like there is some sort of security issue now and the newer platforms dont work. I am trying to find a work round but running out of time. Any help would be appreciated.

[binglis9]

added 2 new records to database and the correct autonumbers were generated. Then went to another desktop and added 2 more new records The first 2 records have disappeared from table but the autonumbers which are used in other tables reverted back as if never used. How can 2 new records just disappear as if they were never added but the autonumbers generated are written to another table that gives totals. So when I added 3rd and 4th new record it took the autonumber of 1 and 2 which means it creates duplicates in total table.
Does anyone have any idea on why this is happening. The 4 desktop frontend database are linked to tables on a map drive pointing to the server.

[Micron]

Most of that IT stuff is Greek to me, so all I can glean from that is at one point (re: reloading records) so it suggests that your table view needs to be refreshed IF you have those tables open in datasheet view. That isn't clear. However, the latest development doesn't seem to support that idea if records are truly missing from one table but are in another.

[binglis9]

I have been testing the new server with 2 client computers at home over the Christmas holidays. The link tables seem to get out of sync. When I use only 1 client everything works fine. If I open up 2 databases at the same time pointing to the same link tables, the link stops working properly. When I add a new record the AutoNumber creates the next number correctly. the jobnumber for each record reads from another table and depending on the letter selected then puts correct number and adds 1 for the next time. For testing purposes I added 3 new records and the autonumber and jobnumber was correct. If I then went to another client with same link tables. Even if i refresh the tables still showing wrong numbers.the autonumber was still showing the number as if the 3 new records had not been added. Plus the jobnumber was still showing as if nothing had been entered in the appropriate table. If I close the database down and open it up again everything is fine and can add new record. If I dont close database down then duplicate primary keys are created as if nothing from other computer has been entered.
The link tables which are pointing to the same tables are not updating correctly which makes the database corrupt and will not work properly. As I said previously the same database with the same link tables works fine on my running server. Its only when I try to upgrade to a later linux server that the problem arises.
I feel as if there has been some sort of upgrade that has happened that is making this happen and i have tried to get this sorted but link tables to a map drive now does not work
any help why the link tables dont work would be gratefully appreciated

[Micron]

Plus the jobnumber was still showing as if nothing had been entered in the appropriate table.

At first I was thinking that this would not be the first time someone had linked to the wrong back end for one or more of the front ends but if you're saying you close down the other db and the records are there when you re-open, then that's not it. That would have to be due to a loss of or dormant connection with no obvious sign, I guess. Sorry, as I mentioned, this is out of my realm of knowledge - especially Linux.