hi. i just started running a high volume c++ app on a postgres database. the storage for the database is nfs mounted from a different linux box. that storage box is running software raid 5 on ubuntu server fiesty (made with something like: "mdadm --create /dev/md0 --level=5 --raid-devices=4 /dev/sda1 /dev/sdb1 ... "). the raid box is put together with fairly recent parts and runs an athlon dual core 4800+ with 2GB RAM. the drives are Samsung SATA 750gb. they are connected to the db server with nfs mount over gigabit ethernet. nfs export: "/home/the_raid_01 192.168.1.1/24(rw,no_root_squash,async,no_subtree_check)" the trouble is after a while of running the high volume app, the box with the raid 5 drive freezes up. it doesn't reboot or anything, it just completely locks up. it never looked that busy on htop. on iostat the utilizations would have peak await of 40+, and utilizations of 60+. they looked generally evenly spread across the physical disks. then it would just lock up. i haven't debugged this sort of problem before. which logs can i look at to help figure out what's causing it? have you seen something like this and what can cause it? (also after the raid box reboots it goes into very heavy i/o processing on the raid physical disks) thanks, gt
