There appear to be bugs in the new SMP scheduler (SCHED_ULE) that no one can track down that cause mysterious crashes and lockups when a system is run under heavy load. The real issue seems to be that only a couple people in the project understand the complex kernel threading model enough to be able to find and resolve bugs, and even they don't seem to be able too most of the time ?
The thing that *really* is of concern for the future of FreeBSB 5.x is the very complex SMP/Thread coding model in is used in 5.x and the fact there only seems to be 2-3 individuals on planet earth that are able to fix and fix bugs in the code. (See the recent posting(s) in the -arch mailing list for some examples of this. This is really a cause of major concern for the future of 5.x. Matt Dillon (aka the Dragonfly BSD project) has constantly been sounding alarm bells about this but no one on the project seems to be concerned at all about this issue.
With the many bugs in 5.x that cause many mysterious crashes and lockups in SMP systems, and that fact that only several people on planet earth seem to have the detailed knowledge of the code to find them, this should be a major source of concern for anyone considering using 5.x in any sort of production application ?
Slashdot Mirror
There appear to be bugs in the new SMP scheduler (SCHED_ULE) that no one can track down that cause mysterious crashes and lockups when a system is run under heavy load. The real issue seems to be that only a couple people in the project understand the complex kernel threading model enough to be able to find and resolve bugs, and even they don't seem to be able too most of the time ?
The thing that *really* is of concern for the future of FreeBSB 5.x is the very complex SMP/Thread coding model in is used in 5.x and the fact there only seems to be 2-3 individuals on planet earth that are able to fix and fix bugs in the code. (See the recent posting(s) in the -arch mailing list for some examples of this. This is really a cause of major concern for the future of 5.x. Matt Dillon (aka the Dragonfly BSD project) has constantly been sounding alarm bells about this but no one on the project seems to be concerned at all about this issue.
With the many bugs in 5.x that cause many mysterious crashes and lockups in SMP systems, and that fact that only several people on planet earth seem to have the detailed knowledge of the code to find them, this should be a major source of concern for anyone considering using 5.x in any sort of production application ?