Date: Fri, 7 Sep 2001 20:29:24 +1000 From: Richard Perini <rpp@ci.com.au> To: freebsd-stable@FreeBSD.ORG Subject: Problem with process in "flswai" state Message-ID: <20010907202924.A6811@mippet.ci.com.au>
next in thread | raw e-mail | index | archive | help
Environment: 4.4 PRERELEASE CPU: Pentium III/Pentium III Xeon/Celeron (701.59-MHz 686-class CPU) real memory = 536805376 (524224K bytes) Adaptec 2940 Ultra2 SCSI adapter 4 x FUJITSU MAE3091LC LVD disks Kernel config is more-or-less GENERIC, with maxusers set to 256 and some unneeded drivers removed. Problem: A particular process that does a large amount of lseeks and small writes to a large (~1Gb) file causes other processes to block for many (>10) seconds at times. Analysing the problem, it appears that the offending process manages to very quickly dirty a large number of buffers, then enters the "flswai" state (as reported by "top"), flushing dirty buffers to disk, and during the flushing frenzy, all other processes seem to block. The interesting thing is that the offending process's I/O is to a disk that is unused by any other process. At the same time, disk I/O on the other disks is nil, or negligible. There is no paging activity and masses of free memory. During the flushing period, the CPU usage (as reported bu systat) drops to < 1%. I've recreated the environment on a spare machine that is otherwise idle, and is easily repeatable. This situation has arisen while attempting to migrate an application from BSD/OS 4.1 to FreeBSD. The symptoms don't occur on BSD/OS. I'm more than happy to experiment with any suggestions to overcome this issue. -- Richard Perini Internet: rpp@ci.com.au Corinthian Engineering Pty Ltd PHONE: +61 2 9906 7866 Sydney, Australia FAX: +61 2 9906 2464 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010907202924.A6811>