gmirror logic of choosing failed drive

Aurimas Mikalauskas aurimas.m at
Wed Mar 9 02:49:19 PST 2005


recenlty I had a strange incident. I have a simple web server with two
gmirror'ed disks (ad4 & ad6). One of the disks (ad6) failed and gmirror
detached one, but the funny thing is - it detached the good one (ad4)
so after a few seconds the server crashed. I've rebooted the server
and once again - it detached the "failed" ad4 drive and crashed in a
few seconds with ad6. I detached the bad one (ad6) by hands
(phisically). Bingo! ad4 works just fine.

Could anyone please explain me what is the logic of gmirror? How does
it decide which drive has failed and how could I influence it to
behave correctly?

Balance algorithm I use is round-robin,
the controller: <Intel 6300ESB SATA150 controller>
the disks: 76319MB <ST380817AS/3.42> [155061/16/63] at ata2-master SATA150

Any ideas?

Thank you in advance!

Aurimas Mikalauskas

More information about the freebsd-questions mailing list