Apparent loss of 'saved' changes (AREV Specific)
At 25 MAR 1999 06:38:18PM John A. Lord wrote:
We are running AREV 3.12 on an NT 4.0 network with the NT Service and NPP. Our server is a 486-66 with 13 gb of hard disk space. Recently we have had several instances of data which was saved getting lost - and apparently reverting to an older version of the data. When I say data, I am referring to both RBasic source code and application table data.
We have been able to document that the data really does revert to an older version, and that this phenomenon is not just due to faulty memories, especially in the case of the table data. Our application keeps track of which user changes data and when. Also, an image of the pre-change data is saved whenever a change is made. As it happens, we also occasionally make archive copies of the entire database. Today, we noticed that changes that were made on the 19th of March had reverted to their pre-19th state. I had made an archive copy of the data on the 19th, so I checked it this afternoon. I attached to it and ran a report to compare to one we had run on the active data. They were different. The active data showed the data the way it was before the change was made and the archive data showed the data the way it had been changed.
Our only theory is that somehow we are 'overwhelming' our old, slow server and that data we think we are saving is getting lost in cache or in virtual memory. All of our workstations are at least Pentium 200s. Is such a thing possible? Has anyone else run into this kind of problem?
Just today we got our new dual pentium-266 server operational and switched our operations over to it. If our theory is right, the problem will go away. But, if not, we need to find a solution. This data is critical budget data for the National Weather Service and there is too much riding on its accuracy for us to let the problem go for very long.
I appreciate any help anyone can offer.
Thanks
jl
At 27 MAR 1999 02:14PM akaplan@sprezzatura.com - [url=http://www.sprezzatura.com]Sprezzatura Group[/url] wrote:
How are you doing all the pre-save storage? Perhaps someone is writing newrec to oldhandle and oldrec to newhandle or something as silly as that? I'd be really surprised if the server cache decided to just drop something, unless you're running a NT server.
Is there a workstation cache? Perhaps that's eating the info and never sending it to a server?
akaplan@sprezzatura.com
At 29 MAR 1999 04:43AM amcauley@sprezzatura.com onmouseover=window.status=why not click here to send me email?;return(true)", [url=http://www.sprezzatura.com" onMouseOver=window.status=Why not click here to visit our web site?';return(true)]Sprezzatura Group[/url] wrote:
May sound stupid but are you sure that you have locking? We've seen this were there is no locking taking place.
amcauley@sprezzatura.com
World Leaders in all things RevSoft
At 08 APR 1999 10:25AM John Lord wrote:
Aaron,
Your comment, "unless you're running a NT server", intrigues me, because we are running NT on the server. What did you mean by that?
jl
At 08 APR 1999 12:04PM akaplan@sprezzatura.com - [url=http://www.sprezzatura.com]Sprezzatura Group[/url] wrote:
Just a slam on NT. Don't trust it not to decided to just dump the cache because it found some other data it liked better.
akaplan@sprezzatura.com
At 08 APR 1999 04:55PM John Lord wrote:
We've seen this problem show up over a space of 2-4 days; that is we'd save data, use the data for a couple of days and then suddenly lose the data. I guess I don't understand how that could happen if the data was not being written to disk from the cache. I thought cache data was more 'transient' than that.
In reading the NPP documentation, I saw the words about FlushFileSizeUpdates and FlushAllUpdates. When these are false, (the default condition) when does NT write the cache?
jl