Yes, it has happened on different servers, none of them is bleeding edge, still using Apache 1.x and PHP 4.3.x. The sites are not using any character sets other than default. The machines are in a heat controlled environment and monitored for change so I would know if there was that kind of problem. One of the sites that I have seen it on is an osCommerce site. Another site is a custom built site. I am inclined to think that because it happens so rarely and when it does happen it is so random that there must be some combination of events that are hard to track as being part of the same problem. Thanks for the brainstorm, you did give me some things to think about. It has since subsided again so I will have to wait for it to happen again before I can troubleshoot more. This makes it very hard. Thanks everyone. If anyone has more ideas I am still open to them. Jonathan ""Richard Lynch"" <ceo@xxxxxxxxx> wrote in message news:56291.67.184.124.249.1119496782.squirrel@xxxxxxxxxxxxxxxx > > WILD GUESS ALERT! > Until you said "on more than one server" I was thinking a heat-related > intermittent hardware error corrupting the file read, so "a" turned into > some not-quite-random character... You *SURE* it's on multiple > servers?... > > Are all machines running the same software versions? > > Are you using multi-byte strings? Those are not quite as pounded on as > much as one might like... > > Have you done anything funky to force libc->glibc upgrade? > > Or a kernel upgrade? > > Are you running anything bleeding/leading edge? (Apache 2, PHP 5, mysqli > etc) Can you rollback for awhile and test? > > If you are using Apache 2, are you in that pre-fork mode, or is it running > threaded? > > What PHP Modules are you running? > > Aha! My wild guesses have lead me to a hypothesis: > > You *are* using Apache 2 in threaded environment and some PHP Module (or > other software) is not thread-safe. > > *WHICH* one[s] are not thread-safe is totally open to question... > > It could even be some weird interaction between *TWO* Modules that very > rarely exhibits itself by random changing of memory bits. Ugh! > > Are there any Modules you could get rid of completely and just change the > applications to "do without"? > > Do you have a test Development box that exhibits this behaviour? Run it > and ab stress test it and pound on it with as many different requests in > as random an order you can achieve. > > If you can get the dev box to mis-behave somewhat reliably, try getting > rid of one PHP Module after another, and re-test incessantly. > > Could you roll back to Apache 1.x or at least do that pre-fork thing so > you are not using threads on the production server, and see if it goes > away? > > Some of the above questions/implications are mutually exclusive. :-) > -- PHP General Mailing List (http://www.php.net/) To unsubscribe, visit: http://www.php.net/unsub.php
