Comment 15 for bug 245779

I too can confirm this problem with 8.04.01 with all updates / upgrades applied. I WAS running on a dual quad core opteron machine with a tyan server class mobo, 32 Gigs of 667MHz ram, and an Areca 1680 SAS card with 16 drives in a large RAID-10 configuration. Under even moderate db loads I would see the cpu#x stuck for 11s message in my logs, and eventually I'd have 3 or 4 cores stuck spinning at 100% in postgres process waiting for I/O. Note that these stuck processes would prevent a proper shutdown and require me to hard power cycle to machine to get it back up and running.

7.10 did not have this problem, but 7.10 kept hanging on install, and could only get one of the machines installed. Not wanting to go into production with an OS that might or might not install (7.10) or one that routinely locks up my CPUs and never gives them back, I have resorted to installing Centos 5.2. Which runs perfectly on this hardware with no errors, hangs or locks. It is running 2.6.18-92.el5 kernel. Hopefully some kind of comparison of the kernel in 8.04.01 and centos 5.2 can turn something up? I'm gonna have another couple days with it available for testing where I can install 8.04.01 on a spare partition and see if I can recreate the problem if someone needs more info.