From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.12) Gecko/20050922 CentOS/1.7.12-1.4.1.centos4 Description of problem: During writes to an ext3 filesystem after an amount of time the writes will slow almost to a halt with a kernel process eating up an entire cpu (25% of system time on a HT dual system). Stopping the process and starting it again fixes it for some time. Switching the filesystem to ext2 or running the latest from kernel.org (2.6.14.3) fixes this issue completely. We have created a program to duplicate this issue which I will attach, and seems to happen after about 40Gig has been written. This is on kernel-smp-2.6.9-22.0.1.EL. iostat: avg-cpu: %user %nice %sys %iowait %idle 0.10 0.00 25.04 0.60 74.26 Device: tps kB_read/s kB_wrtn/s kB_read kB_wrtn sda 1.80 0.00 682.40 0 3412 sdb 0.00 0.00 0.00 0 0 top: top - 17:50:23 up 51 min, 4 users, load average: 1.12, 4.49, 7.78 Tasks: 79 total, 2 running, 77 sleeping, 0 stopped, 0 zombie Cpu(s): 0.0% us, 25.1% sy, 0.0% ni, 73.8% id, 1.1% wa, 0.0% hi, 0.0% si Mem: 4148928k total, 4132900k used, 16028k free, 16368k buffers Swap: 8193108k total, 128k used, 8192980k free, 3998196k cached PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 3073 root 25 0 1888 428 360 R 99.9 0.0 13:00.88 make-fragmentat 222 root 15 0 0 0 0 S 0.3 0.0 0:50.89 kjournald Version-Release number of selected component (if applicable): kernel-smp-2.6.9-22.0.1.EL How reproducible: Always Steps to Reproduce: 1. compile program make-fragmentation.c 2. run program make-fragmentation 3. iostat will show the writes slowing to almost nothing Actual Results: The process's writes would slow to about 600KB/s and system time would stick at 25% (1 of 4 virtual HT processors) Expected Results: System writes should have remained at a ~40MB/s level Additional info:
Created attachment 121947 [details] Program to create slow writes state
Fixed in 2.6.9-22.12; will be in U3. *** This bug has been marked as a duplicate of 156437 ***