> On Apr 2, 2015, at 10:51 PM, Lidong Zhong <lzhong@xxxxxxxx> wrote: > > Currently if there is a leg failure, the bio will be put into the hold > list until userspace replace/remove the leg. Here we are trying to make > dm-raid1 ignore the failure and keep the following bios going on. > This is because there maybe a temporary path failure in clvmd > which leads to cluster raid1 remove/replace the fake device failure. And > it takes a long time to do the full sync if we readd the device back. > --- Just be aware that there are failure scenarios that this does not handle as well. Like a reboot causing the legs to return but not the log. It’s a remote possibility, but something the old model considered. The new RAID targets which leverage the MD personalities are much better (e.g. ‘lvcreate —type raid1’), but they don’t support active-active cluster setups. I’ve seen a couple things on the linux-raid mailing list suggesting that someone (Goldwyn Rodrigues) may be thinking about making RAID1/10 cluster-aware… LVM should be able to take advantage of this if it became a reality. Are you interested in providing your use-case for cluster mirroring (vs opting for active/passive mirroring)? brassow -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
