Mysql spinning - Native vs LinuxThreads?

[donavan]

Anyone else had issues with MySQL spinning?
This is the second time in ~4 months where it looks like MySQL hit a bad call and is spinning away all of 1 CPU.

heres what #truss -p 229 -o sqlfuct.log gets me:
<snip>
poll(0x83d5000,0x5,0x6b6)                        = 1 (0x1)
gettimeofday(0x2848e288,0x0)                     = 0 (0x0)
poll(0x83d5000,0x5,0x6b5)                        = 1 (0x1)
gettimeofday(0x2848e288,0x0)                     = 0 (0x0)
<snip>

show processlist was showing some queries hanging for up to 5min while others were avg ~5 seconds.Im avg ~4 queries/second. Without this runaway process the box is  lightly loaded with ~50% max CPU and ~1.2 avg load during the day.

HUP dosnt bounce mysql over, I have to do a start/stop to fix the problem. I found an old news thread circa '01 about a threading bug in MySQL/FreeBSD but no resolution.

Anyone know if going to LinuxThreads will solve this? OR any other known solution?

[matt]

Some folks claim success with Linux threads, others don't. FWIW, I don't have this problem any long on MySQL 4.1.x on FreeBSD 5, but massive changes to SMP were made with FreeBSD, and in general, I don't recommend running FreeBSD 5 in production.  

However, we're fairly close to seeing a FreeBSD STABLE_5 and when that happens, I'll recommend that everyone use FreeBSD 5 and track the -STABLE release.

Matt