gvinum raid5 in production

[Mark Bucciarelli]

Is anybody using gvinum raid5 in production with 6.1-release?  How is it
going?

Have you every had any instances where a kernel panic caused the raid-array
not to rebuild properly on reboot and you had data loss?

Has your array been robust to SCSI errors from your drives?

I reviewed all problem reports I could find and only 89660 looked serious.

    Dec. 2005 (6.0): kernel panic (perhaps due to low memory)
    caused by code in geom perhaps. Looks like raid rebuilds
    properly, only b/c poster didn't mention any problem.  Patch
    included.  (Hmmm, gvinum code doesn't check for null pointers
    returned by malloc.) [1]

    Sep. 2005 (6.0): probably not geom-related (according to
    poster).  Kernel panic when copying large file (2 Gb). [2]

    Feb. 2005 (5.3): A newfs bug caused by an integer overflow
    when entering stupidly large values when creating a new file
    system.  No risk. [3]

    Nov. 2005 (5.3): Pulls a drive from raid 5 (w/o turning off
    machine, I _think_), get's SCSI errors, and kernel panics.
    Older machine?  (Pentium Pro) [4]

As a followup question, in case of kernel panic, coredump and subsequent
savecore, is it safe to have swap on a gmirror raid1?  How about on gvinum
raid 5?

m

[1] http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/89660

[2] http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/85728

[3] http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/77181

[4] http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/73830