Hi there,

our product is running at many customers without that specific problem but one customer is having an issue running it.

Environment:
- Windows 2019 - Server running in VM-Ware
- EDBSRVR is used as a Service
- EDBSRVR is accessing a database on a local drive
- The Server is empty except our Product. No Virus-Scanner is active or installed, (Except a disabled MS-Defender)

While our Applications are running after 15min ~ 40min. the Database is getting permanently locked.
The following error appears:

Unable to update CEDevice-Entry: ElevateDB Error #1000 An error occurred with the cursor CEDevice (An attempt was made to update a read-only cursor)

After that happens we have to replace the edbconfig-file because we are unable to unlock the database in any other way ... (restarting the service won't help, it looks like the lock-state has been stored to the edbconfig)

We think there could be a bug in Elevate ... when our applications detect that there is no EDBSRVR running they are falling back to edb-local-mode .. we guess that this could lead to a save or store of the locked-edbconfig but we are unsure about that.

in the end, the database is locked and more strange ... stays locked and we don't have a clue how this happens.
we have no other customer we that problem but this customer is using our apps exhaustively perhaps the high-access-rate is a problem?!

but I guess in the first place it should not be possible to store the lock state to the EDBConfig?
As far as we have seen the lock state could be set because a DB-file can only be opened read-only ... this read-only lock stays for the whole session ... we are guessing there must be some point where this lock-flag is getting stored into the edbconfig.

any suggestions?

gripsware


I have a vague memory of something like this happening but it was a while ago and I can't come up with the right search terms to find it (if it ever happened) in the newsgroups.

One piece of information you are missing in your post is the version of ElevateDB / EDBServer at the clients site.

Roy Lambert

its version 2.31 b12

but nvm. we found out that sentinel-av is the problem.
looks like the sentinel-av is causing connection issues with the DB-server and file-access.

On 8/10/2020 2:44 AM, gripsware wrote:

> Environment:
> - Windows 2019 - Server running in VM-Ware
> - EDBSRVR is used as a Service
> - EDBSRVR is accessing a database on a local drive
> - The Server is empty except our Product. No Virus-Scanner is active or installed, (Except a disabled MS-Defender)

What's the EDB version on server and clients ?

> While our Applications are running after 15min ~ 40min. the Database is getting permanently locked.
> The following error appears:
>
> Unable to update CEDevice-Entry: ElevateDB Error #1000 An error occurred with the cursor CEDevice (An attempt was made to update a read-only cursor)
>
> After that happens we have to replace the edbconfig-file because we are unable to unlock the database in any other way ... (restarting the service won't help, it looks like the lock-state has been stored to the edbconfig)

Do you know what the app is doing when this happens - one common issue
for #1000 and read-only cursor error is when app is trying to change one
of the system tables either directly or changing config path or such
(which has to be done with alter statements and not with update
statements for example).

Any chance this is a issue with app updating config or database path(s)
or such runtime ?


> We think there could be a bug in Elevate ... when our applications detect that there is no EDBSRVR running they are falling back to edb-local-mode .. we guess that this could lead to a save or store of the locked-edbconfig but we are unsure about that.

Local and client-server server mode are supported together and number of
people have used them so that itself should not be the cause.


> in the end, the database is locked and more strange ... stays locked and we don't have a clue how this happens.
> we have no other customer we that problem but this customer is using our apps exhaustively perhaps the high-access-rate is a problem?!

If you run into this locked situation then i would check all config from
top to bottom first (you can just query system tables) or EDBManager to
see if anything stand out

Raul

Hi Raul,

> What's the EDB version on server and clients ?
It is version 2.31 b12

> Do you know what the app is doing when this happens - one common issue
> for #1000 and read-only cursor error is when app is trying to change one
> of the system tables either directly or changing config path or such
> (which has to be done with alter statements and not with update
> statements for example).
> Any chance this is a issue with app updating config or database path(s)
> or such runtime ?

Yes we found that problem .. When the Sentinel-AV blocks the TCP/IP-Connection to the EDBSRVR (happens randomly ... very strange) our Application did a fallback and used the local-mode ... in the local-mode the app modified the database-paths to UNC.. after that change a few tables are staying locked and will never recover.
we disabled the fallback to so we get around the #1000 problem and told the customer to disable the sentinel-av due to its massive issues.