Icon View Thread

The following is the text of the current message along with any replies.
Messages 21 to 30 of 31 total
Thread Zero filled EDBDatabase.EDBCat
Fri, Jun 22 2012 5:21 AMPermanent Link

Roy Lambert

NLH Associates

Team Elevate Team Elevate

Michael


OK you've now got a test case where it happens. Can you duplicate this on your own hardware? If so is it repeatable or random? If you can't duplicate it then it HAS to be something to do with the configuration of your customer's PCs (and I would still guess anti virus). If you can duplicate it but its random then you have some more work to do to identify the cause.

NOTE
I am not saying that it is definitely not ElevateDB but that it is unlikely. Many years ago I had a problem with DBISAM. It only happened on my email & newsreader client and only when I used multiple threads. Tim couldn't spot it and it took me ages and much messing around to produce a demo for Tim. Then the problem was fixed rapidly.

Roy Lambert [Team Elevate]

ps: I don't understand why creating a temporary table doesn't generate a .OLD but creating an index on it does. I'm sure Tim does though.
Fri, Jun 22 2012 5:49 AMPermanent Link

gripsware

gripsware datentechnik gmbh

Well at the moment we are unable to replicate the problem .. its totally random ..
at customer running the app every day multiple hours it happend twice in a week.

( I personally think if we stop touching the the cat.-file the problem will be ""solved"". )

An interesting thing is that it also happens with the original edbserver redist. running on the server ...
mhm .. i think we will double-check the AV thing .. thats the only realistic scenario left .. if not i am out of ideas .. we never had such problems before...


Roy Lambert wrote:

Michael


OK you've now got a test case where it happens. Can you duplicate this on your own hardware? If so is it repeatable or random? If you can't duplicate it then it HAS to be something to do with the configuration of your customer's PCs (and I would still guess anti virus). If you can duplicate it but its random then you have some more work to do to identify the cause.

NOTE
I am not saying that it is definitely not ElevateDB but that it is unlikely. Many years ago I had a problem with DBISAM. It only happened on my email & newsreader client and only when I used multiple threads. Tim couldn't spot it and it took me ages and much messing around to produce a demo for Tim. Then the problem was fixed rapidly.

Roy Lambert [Team Elevate]

ps: I don't understand why creating a temporary table doesn't generate a .OLD but creating an index on it does. I'm sure Tim does though.
Fri, Jun 22 2012 7:39 AMPermanent Link

Roy Lambert

NLH Associates

Team Elevate Team Elevate

Michael


>Well at the moment we are unable to replicate the problem .. its totally random ..
>at customer running the app every day multiple hours it happend twice in a week.

From your earlier posts I thought it was at several customers in which case if they use a common AV software that's a good indicator.

>( I personally think if we stop touching the the cat.-file the problem will be ""solved"". )

No it won't be solved it will just be hidden and will probably resurface to bite you (or someone else if the fault is in ElevateDB) when its even more critical.

>An interesting thing is that it also happens with the original edbserver redist. running on the server ...

Not really, since none of the changes you will have made to the server will have affected the underlying operations of ElevateDB. What it does mean is that it is unlikely to be your alterations to the server that caused the problem.

>mhm .. i think we will double-check the AV thing .. thats the only realistic scenario left .. if not i am out of ideas .. we never had such problems before...

Ahhh how well I know that feeling <vbg>

Roy Lambert [Team Elevate]
Fri, Jun 22 2012 8:56 AMPermanent Link

Raul

Team Elevate Team Elevate

Michael,

Looking thru this thread I'm not sure if you're using a file-share or
actually running edbsrvr on the server.

Is the server a real server running windows server OS ?

For edbsrvr the AV would be most likely culprit as other have pointed
out. Some of the more aggressive ones have caused us all kinds of grief
even on network level so for testing we always ask it to be completely
disabled rather than just our files/folders excluded.

In addition of course make sure OS and drivers are up to date (server
and clients).

Raul



On 6/22/2012 5:49 AM, Michael Reisch wrote:
> ( I personally think if we stop touching the the cat.-file the problem will be ""solved"". )
>
> An interesting thing is that it also happens with the original edbserver redist. running on the server ...
> mhm .. i think we will double-check the AV thing .. thats the only realistic scenario left .. if not i am out of ideas .. we never had such problems before...
>
>
Fri, Jun 22 2012 9:56 AMPermanent Link

Roy Lambert

NLH Associates

Team Elevate Team Elevate

Raul


I'm guessing but this <<well we are trying to get the server-logs from the customer>> makes me think its a server, but you raise some interesting questions

Roy Lambert [Team Elevate]
Mon, Jun 25 2012 3:03 AMPermanent Link

gripsware

gripsware datentechnik gmbh

Stop using the Index on the temporary table surly would not solve the problem and just get the app working more stable... but as long as we are unable to replicate the problem i don't see other options.

At one customer the Application was first running with a embedded-server via network-share (unc) on a windows server. .. as the problem appeared we installed the edbsrv....exe on that windows-server .. cause the workstation-logs showed us network-problems.

The problem reappeared after a few days  .. we tried to replicated it .. and we are using the app in 90% of the time in our network-enviroment with a share and never had such problems.

However ... we will keep watching the problem .. i personally think its a memory corruption .. mostly AV's are blocking  API-Calls and don't fitler out specific blocks ... so the write-access would completly fail and not only filled with zero? (when we had av problems the simply blocked the call)
Tue, Jul 3 2012 8:04 AMPermanent Link

Tim Young [Elevate Software]

Elevate Software, Inc.

Avatar

Email timyoung@elevatesoft.com

Michael,

<< I triggered that a other worker created a temporary-table on the
application-startup that causes the creation of the .old file ... i think
that is the point where for some strange reason the EDBCat just only get the
header filled... >>

Temporary table creation does not touch the catalog file, and therefore does
not cause a backup (.old) of the catalog to be created.

Or are you talking about a table that is on-disk, but "temporary" to the
application and not to EDB ?

Tim Young
Elevate Software
www.elevatesoft.com
Tue, Jul 3 2012 8:46 AMPermanent Link

Roy Lambert

NLH Associates

Team Elevate Team Elevate

Tim


This is what he's referring to:

create temporary table gronk as select * from adam with data

NO .old created

create index spurt on gronk(user)

.old created.

Roy Lambert
Fri, Jul 27 2012 9:57 AMPermanent Link

gripsware

gripsware datentechnik gmbh

And another customer ... *sigh*

will there be any improvment so that a "create index" on a temporary table doesn't change the catalog file ?!

.oO( ... we will start testing the file for zero-filled-style at startup and than copy the .old - file over it ... )
Fri, Jul 27 2012 12:44 PMPermanent Link

Fernando Dias

Team Elevate Team Elevate

Volker,

An inocent question Smiley: is the database folder acessible to users?


--
Fernando Dias
[Team Elevate]
« Previous PagePage 3 of 4Next Page »
Jump to Page:  1 2 3 4
Image