PDA

View Full Version : mdadm - software raid - State : dirty, no errors



Anne
12th January 2007, 08:23 AM
There are 6 drives configured in a raid5 with no hot swap.

The system crashes under heavy load, and the raid5d is the top process.

In /proc/mdstat:
md8 : active raid5 sdf14[5] sde14[4] sdd14[3] sdc14[2] sdb14[1] sda14[0]
1839240960 blocks level 5, 64k chunk, algorithm 0 [6/6] [UUUUUU]
[================>....] resync = 82.9% (305197108/367848192) finish=101.9mi

Even after the resync has finished, "mdadm --query --detail /dev/md7" shows
Update Time : Thu Jan 11 09:41:30 2007
State : dirty, no-errors
Active Devices : 6
Working Devices : 6
Failed Devices : 0
Spare Devices : 0

This system has a 3ware hardware raid controller, and all the drives show "OK"
in the status. So I don't know why mdadm would mark it dirty, and rebuild it.

After the resync has finished, shouldn't mdadm mark the State : Clean ??

It takes about 15 hours to resync all the filesystems, during which time the
filesystems are in a fragile state, and I have restricted activity. Otherwise
it crashes repeatedly.

Thanks for any pointers, suggestions.

Red Hat Enterprise Linux WS release 3 (Taroon Update 8)

Anne