Re: LOADREC / Indexer (OpenInsight Specific)
At 21 FEB 2001 07:02:24AM Ross Brooker wrote:
Well, I've had a chance to search previous threads about my two issues… so I think I've got the Loadrec problem covered (but not fixed!)
Of concern now is a nugget I picked up from trawling on Indexer.. our application has Update_Index built into the code AND we're running a stand-alone indexer… How much of a problem is this? I'm led to believe (from a previous thread) that will lead to occasional missed updates. Should we be switching to Indexer alone, or ensuring the application always does update_index ??
Many thanks (in anticipation)
Ross Brooker
============
At 21 FEB 2001 09:52AM [url=http://www.sprezzatura.com]The Sprezzatura Group[/url] wrote:
The only real problem that would happen is a potential lockup. Usually the code is pretty code with avoiding the deadlock situation. The indexing code had all sorts of loop until timeout code in F.INDEXER and F.DISTRIBUTOR to avoid the deadlocks, however nothing is failsafe. The deadlocks really only occur when there are multiple sessions trying to do full updates.
The occasional update in the code is not a problem, if you need that index right away. If it's only there to ensure an update then it's not a problem.
The theory is that if station A hits a index update request and is processing file A, when the indexer comes along to file A, it will skip it and move to file B.
The answer as to whether the updates are needed really depends on how long it takes to process a typical round of updates. If this occurs in a few seconds, and you will not be able to issue a select requiring the new data in that time, then the update is not required.
For example, suppose it takes 5 seconds to make the average indexing round. A user performs an update and needs to see this information in the next query. The process finishes, the user then acknowledges the completion, moves to the reporting section, enters the criteria for the report then runs the report. If it takes more than 5 seconds for the user to complete this task, the the update code is not required.
World Leaders in all things RevSoft
At 21 FEB 2001 10:23AM Ross Brooker wrote:
That sounds like good news to me…
"The answer as to whether the updates are needed really depends on how long it takes to process a typical round of updates. If this occurs in a few seconds, and you will not be able to issue a select requiring the new data in that time, then the update is not required."
Our indexer takes around 1 - 3 minutes to cycle through all the tables, (a lot if indexes on a lot of data and a slow network)but if we can add in some more specific updates where the problems seem to occur, then we'll be happy with that
Many thanks for your help
Ross
====
At 22 FEB 2001 05:29AM Colin Rule wrote:
It is coincidence that I have just had the same problem with REV_LOADREC.
I checked out the points recently raised and nothing resolved, so I thought I check my program, blaming oneself last is the easy option…
It turned out that the LOCK FILE,ITEMID command causes the problem.
If the ITEMID variable is not assigned it causes this problem.
Check your code, and advise us all if you fell into the same trap.
Revelation:….. any chance of fixing this for the next update.
At least make the message a bit easier to follow.
Colin Rule