Re: Bug 1557

hiya, On Wed, 22 Sep 1999, Matt Corddry wrote: > Hi there: > > I've been experiencing the exact same problems as detailed in bug 1557 > -- random oracle errors on the solaris platform. However, I seem to be > making progress in working around them. the reported bug talks about the ora_* interface in php - it's unlikely that the same problem is in oci* (as they share *NO* code) - it might be that you're triggering another problem in php3. maybe you could set up a test system with the latest CVS-version of php4 and see if you're still having that problem! > > As you know, apache has the MaxRequestsPerChild directive, which limits > the number of requests an apache/php process will handle before it dies. > My server had that variable set at 10,000. > > At the recommendation of some emails I dug up on the php-dev archives, I > compiled php as a standalone binary and was unable to recreate any of > the errors when in that mode. I then decided to lower the > MaxRequestsPerChild number significantly -- the OCI8/php interface seems > to be reliable when this variable is set to 30. I'm using OCIPlogon(), > but of course this only lasts for 30 hits at most. i myself have not had any problems with ociplogon - but that could be that my app does not trigger a bug thats somewhere in the code! question is, are you using ocibindbyname or ocidefinebyname - and - if yes - are you sure, that you don't unset the php-variables you pass in by reference? (unset would also be to bind a local variable and return fron the function who does the bind) - php3 has no concept of reference counting (php4 does:)) and thereby the oci* driver might write into you memory at random places! if you could create a small testcase (small table etc) that you could send me i *will* have a look at it! > > I'm running on a sun enterprise 250 (2x400MHz) with 2G of ram talking to > oracle 8.0.5 on a seperate machine... in this config, I've been able to > execute about 27 somewhat complex select queries per second (3 per page, > 9 pages/sec) with zero errors for extended durations, but this number > could be database-limited. it probably is;-) - the oci-driver should have little to no overhead (but it will be even faster in PHP4!) > > I hope this info helps with the PHP OCI8 driver & other people's > experience with it. Perhaps, once this can be tested further, it could > be incorporated into the OCI8 documentation so that others don't have > the problems that I've had. finding the problem would be great - best solution would be a fix and no need for a note in the docs! > > Thanks for all your work on php & the oracle drivers -- definitely a > really cool product. it's also good fun for me! and if you help me finding bugs it will in the end benefit me and my product (based on PHP) as well:-) > > matt corddry > amazon.com livebid events does that mean amazon.com uses php?? regards, tc > > Thies C. Arntzen "One Big-Mac, Small Fries and a Coke!" Digital Collections Phone +49 40 235350 Fax +49 40 23535180 Hammerbrookstr. 93 20097 Hamburg / Germany