Date: Mon, 11 Jan 1999 00:05:05 -0800 (PST) From: Matthew Jacob <mjacob@feral.com> To: Hidetoshi Shimokawa <simokawa@sat.t.u-tokyo.ac.jp> Cc: Mike Smith <mike@smith.net.au>, alpha@FreeBSD.ORG Subject: Re: complete lockup under large I/O? Message-ID: <Pine.LNX.4.04.9901102358250.16469-100000@feral-gw> In-Reply-To: <19990111164950Z.simokawa@sat.t.u-tokyo.ac.jp>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 11 Jan 1999, Hidetoshi Shimokawa wrote: > mjacob> I'll try, but last I heard BRK didn't work on serial lines. > > Did you add "options BREAK_TO_DEBUGGER" to config file? > It works mostly for me. Nope. Will do so next time. > > mjacob> On Sun, 10 Jan 1999, Mike Smith wrote: > mjacob> > mjacob> > > > mjacob> > > I have a test script/program that repeatedly reads > mjacob> > > or writes a same spot on a device using progressively > mjacob> > > larger blocksizes. > mjacob> > > > mjacob> > > I was using 3 8GB disks ccd'd together on a AlphaPC164 and the > mjacob> > > the system completely froze when it went about 1MB blocksizes. > mjacob> > > > mjacob> > > Anyone have a notion about this? > mjacob> > > mjacob> > You may be seeing a deadlock on resource starvation; it'd be really > mjacob> > interesting to see the output of 'ps' if you can get into DDB. > > I sometimes have similar problem. > In such case, ps shows many process are waiting at vmpfw or ppwait state. > Well- there should only be one really active processes. I'm collecting a suite of nasty test programs, and a lot of them fork lots of processes. This test, though, is just a large blocksize dd except it lseeks and repeats- it's a good test of cacheability and performance. For example, here's a test run under NetBSD/Alpha I ran last week (this is what I used to finally discover that I'd broken PCI bursts - this is the 'after'): Script started on Wed Jan 6 15:03:44 1999 # brunner.nas.nasa.gov > TestSameSpot /dev/rsd2b Same Spot Read Write Test on Wed Jan 6 15:04:11 PST 1999 NetBSD brunner.nas.nasa.gov 1.3I NetBSD 1.3I (MATHOM.EXP) #1: Wed Jan 6 14:58:56 PST 1999 mjacob@mathom.nas.nasa.gov:/space/NetBSD-current/src/sys/arch/alpha/compile/MATHOM.EXP alpha Read 67108864 bytes in 7.864998 seconds, 8332.61KB/sec in 4096 byte blocks Read 67108864 bytes in 4.820000 seconds, 13596.68KB/sec in 8192 byte blocks Read 67108864 bytes in 3.353333 seconds, 19543.54KB/sec in 16384 byte blocks Read 67108864 bytes in 2.636665 seconds, 24855.64KB/sec in 32768 byte blocks Read 67108864 bytes in 2.272498 seconds, 28838.75KB/sec in 65536 byte blocks Read 67108864 bytes in 2.91667 seconds, 31331.95KB/sec in 131072 byte blocks Read 67108864 bytes in 2.39165 seconds, 32138.64KB/sec in 262144 byte blocks Read 67108864 bytes in 1.986665 seconds, 32987.95KB/sec in 524288 byte blocks Read 67108864 bytes in 1.966665 seconds, 33323.42KB/sec in 1048576 byte blocks Read 67108864 bytes in 1.979162 seconds, 33113.00KB/sec in 2097152 byte blocks Read 67108864 bytes in 1.999999 seconds, 32768.02KB/sec in 4194304 byte blocks Wrote 67108864 bytes in 13.922498 seconds, 4707.20KB/sec in 4096 byte blocks Wrote 67108864 bytes in 8.755831 seconds, 7484.84KB/sec in 8192 byte blocks Wrote 67108864 bytes in 4.234164 seconds, 15477.91KB/sec in 16384 byte blocks Wrote 67108864 bytes in 3.239164 seconds, 20232.38KB/sec in 32768 byte blocks Wrote 67108864 bytes in 2.676664 seconds, 24484.21KB/sec in 65536 byte blocks Wrote 67108864 bytes in 2.426664 seconds, 27006.62KB/sec in 131072 byte blocks Wrote 67108864 bytes in 2.308332 seconds, 28391.06KB/sec in 262144 byte blocks Wrote 67108864 bytes in 2.224166 seconds, 29465.43KB/sec in 524288 byte blocks Wrote 67108864 bytes in 2.108329 seconds, 31084.33KB/sec in 1048576 byte blocks Wrote 67108864 bytes in 2.112499 seconds, 31022.97KB/sec in 2097152 byte blocks Wrote 67108864 bytes in 2.120830 seconds, 30901.11KB/sec in 4194304 byte blocks # brunner.nas.nasa.gov > exit exit Script done on Wed Jan 6 16:31:19 1999 On the FreeBSD/Alpha system that froze tonite, it froze here as soon as it went higher than 1MB on reads. Not a lot of processes here, so I'll look elsewhere for the problem. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.LNX.4.04.9901102358250.16469-100000>