We've had a periodic problem on a set of identical Dell PowerEdge 2550s where top bombs out with a floating point exception. I first suspected that the kernel had become out of sync with the userland (by the latter being upgraded but the former not), or even that I had installed a new kernel but not rebooted to using it and Red Hat was (foolishly) overwriting the old kernel's shared objects on disk, and then we got screwed on paging.
Nope, this is yet another (see also) known, reported bug that Red Hat simply refused to fix… and now they've discontinued the 7.x series, so I can yell and scream at them all I want without getting it fixed. At least they labeled this one RAWHIDE (”We've fixed it internally, we don't care about you people with production systems that have paid for our product.”) rather than WONTFIX (”Your opinion as our customer doesn't matter at all. We're going to do it our way, and you should shut up and like it.”).
I simply can't be bothered to reboot production systems because some nitwit used a float when he should have used an unsigned float, and my users wouldn't be too happy about it either. So I went and downloaded procps-2.0.7-25.src.rpm, which was released for Red Hat 8.0, did a rpm –rebuild procps-2.0.7-25.src.rpm, then did rpm -U /usr/src/redhat/RPMS/i386/procps-2.0.7-25.i386.rpm. Note that it's only through the rpmfind.net community effort that it was even possible to get this file. What the hell am I paying Red Hat for?
(Yes, I could have just built the stuff that RH's procps contains from the latest source, but I prefer to use the existing package managment system if at all possible, since, even with RPM, it lends some sanity to managing software on Unix and Linux systems.)
[Cross-posted to
Post a Comment