Re: Databases going to sleep

This WebDNA talk-list message is from

1998

It keeps the original formatting. numero = 19433
interpreted = N
texte = I do similar logging and have seen blank entries. I prevented it by swapping the log file (rolling) whenever the largest reached several Megs. Make sure you aren't writing the files to disk with every append (an option in preferences or with commitdatabases).I attributed it (perhaps incorrectly) to WebCat receiving hits faster than it can append to a huge file. Visitor logs are a worst case scenario when banner ads are flooding the world via FlyCast. But I looked at the latest log and it appears fine and I know it took record (for it) traffic pounding a site.The fact that the searches weren't working implies the database was corrupted in RAM atleast. The appends during the corrupt periods probably are the blank entries. This is all conjecture, of course. It would be interesting to hear a real explanation from PCS. But I do seem to prevent it with regular rolling the visitor log. I presume you keep a different visitor log per virtual host to isolate each customers data and problems.Sandy>I have a problem with WebCat 2.1.3 CGI. I am using it for (amongst other >things) visitor logging, where every page contains an [append >log.db]ldate=[date %Y/%m/%d]...lots of useful info about the >user...[/append]. This has worked brilliantly, and allows us to deliver >visitor analysis stuff that the major ISPs are drooling over.>>However.... I recently looked at the visitor analysis page for one of our >virtual sites and it said that there had been no visitors all day. I >opened the log file with a text editor, and discovered that all of the >entries for the past 24 hours were empty-- ie the tabs and return were >there but nothing between them. Not even the date. I closed the DB, and >it started logging again, but of course the previous 24 hours' data were >lost.>>I have since looked at all the log files and discovered that there have >been dead periods in the log (lots of records, but no data) for periods >of a few hours to a two days, as far back as September 97 (one isolated >case), although all except that one happened since February. I have >always updated to new versions of WebCat soon after they come out (except >3, of course). Maybe this has something to do with it?>>A new wrinkle today: a customer called to say their visitor log weekly >summary was showing zero for the whole week. I opened the file with a >text editor, and it showed that for the last couple of hours it was >writing empty records, ie just a return with no tabs. All searches in the >database were returning zero results, even though there were plenty of >valid entries. Closing the database not only fixed the empty record >problem, it also fixed the zero searching problem. After reloading, the >database now had tab-delimited empty stuff in the recent dead period >instead of just returns. I assume WebCat did this when it reloaded it.>>A possible red herring: the problems have increased a lot since I gave >the CGI a lot more RAM.>>Anyone have any ideas? The next thing I am going to try is changing the >date format to days since 0, but this is a big step for so many sites, >which all share the same logging code.>>I would very much appreciate some gems of insight on this one...>>Regards>Thomas>>Thomas Wedderburn-Bisshop>Technical Director>Woomera Net Solutions>Sydney, Australia Phone +61 2 9746 0700 fax +61 2 9751 6941>> Associated Messages, from the most recent to the oldest:

Re: Databases going to sleep (Sandra L. Pitner 1998)
Re: Databases going to sleep (Kenneth Grome 1998)
Re: Databases going to sleep (PCS Technical Support 1998)
Re: Databases going to sleep (Sandra L. Pitner 1998)
Databases going to sleep (Thomas Wedderburn-Bisshop 1998)

I do similar logging and have seen blank entries. I prevented it by swapping the log file (rolling) whenever the largest reached several Megs. Make sure you aren't writing the files to disk with every append (an option in preferences or with commitdatabases).I attributed it (perhaps incorrectly) to WebCat receiving hits faster than it can append to a huge file. Visitor logs are a worst case scenario when banner ads are flooding the world via FlyCast. But I looked at the latest log and it appears fine and I know it took record (for it) traffic pounding a site.The fact that the searches weren't working implies the database was corrupted in RAM atleast. The appends during the corrupt periods probably are the blank entries. This is all conjecture, of course. It would be interesting to hear a real explanation from PCS. But I do seem to prevent it with regular rolling the visitor log. I presume you keep a different visitor log per virtual host to isolate each customers data and problems.Sandy>I have a problem with WebCat 2.1.3 CGI. I am using it for (amongst other >things) visitor logging, where every page contains an [append >log.db]ldate=[date %Y/%m/%d]...lots of useful info about the >user...[/append]. This has worked brilliantly, and allows us to deliver >visitor analysis stuff that the major ISPs are drooling over.>>However.... I recently looked at the visitor analysis page for one of our >virtual sites and it said that there had been no visitors all day. I >opened the log file with a text editor, and discovered that all of the >entries for the past 24 hours were empty-- ie the tabs and return were >there but nothing between them. Not even the date. I closed the DB, and >it started logging again, but of course the previous 24 hours' data were >lost.>>I have since looked at all the log files and discovered that there have >been dead periods in the log (lots of records, but no data) for periods >of a few hours to a two days, as far back as September 97 (one isolated >case), although all except that one happened since February. I have >always updated to new versions of WebCat soon after they come out (except >3, of course). Maybe this has something to do with it?>>A new wrinkle today: a customer called to say their visitor log weekly >summary was showing zero for the whole week. I opened the file with a >text editor, and it showed that for the last couple of hours it was >writing empty records, ie just a return with no tabs. All searches in the >database were returning zero results, even though there were plenty of >valid entries. Closing the database not only fixed the empty record >problem, it also fixed the zero searching problem. After reloading, the >database now had tab-delimited empty stuff in the recent dead period >instead of just returns. I assume WebCat did this when it reloaded it.>>A possible red herring: the problems have increased a lot since I gave >the CGI a lot more RAM.>>Anyone have any ideas? The next thing I am going to try is changing the >date format to days since 0, but this is a big step for so many sites, >which all share the same logging code.>>I would very much appreciate some gems of insight on this one...>>Regards>Thomas>>Thomas Wedderburn-Bisshop>Technical Director>Woomera Net Solutions>Sydney, Australia Phone +61 2 9746 0700 fax +61 2 9751 6941>> Sandra L. Pitner

DOWNLOAD WEBDNA NOW!

Re: Databases going to sleep

1998

Top Articles:

Related Readings: