Hi Luca and Wol
Now it has a problem about the combination of raid6 and dm-integrity. If
there are many read failures at a time,
it can set the member as faulty.
} else if (atomic_read(&rdev->read_errors)
> conf->max_nr_stripes)
printk(KERN_WARNING
"md/raid:%s: Too many read errors,
failing device %s.\n",
mdname(conf->mddev), bdn);
But now it just can be reproduced on virtual machines and we modify
about 500MB data of the disk. And in real world I think it's hard to
trigger this.
@Luca, this combination can fix the silent data corruption
automatically. If there are some data is broken on some sectors, this
combination can
fix it automatically. I pasted one link several days before. You can
check it.
Regards
Xiao
On 07/24/2019 04:34 AM, Luca Lazzarin wrote:
I do not know it. Could you please link me to some infos?
Why do you suggest it?
I mean, which benefits will it give to me?
Thanks :-)
On 23/07/19 22:30, Wol's lists wrote:
On 23/07/2019 21:16, Luca Lazzarin wrote:
Thank you all for your suggestiong.
I'll probably choose RAID6.
Any chance of putting it on top of dm-integrity? If you do can you
post about it (seeing as it's new), but it sounds like it's something
all raids should have.
Cheers,
Wol
