[DRBD-user] IO vs Safety

Lionel Sausin ls at numerigraphe.com
Tue Oct 8 07:59:36 CEST 2013

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.

Le 07/10/2013 21:07, Miloš Kozák a écrit :
> (...)What happens I one disk failes? Is it same for drbd as for mdadm? 
> I just replace HDD and everything runs smoothly forward? What if I 
> need to reboot system meanwhile?
Basically yes, it's much like replacing a drive with mdadm.

Be aware that if the drive was serving the primary node then all I/Os 
are redirected to the secondary nodes over the network. It may be 
slower, and depending on your configuration DRBD may savagely reboot the 
primary node to protect the data if the network goes down.

Also please bear in mind that not all drives fail cleanly either. Some 
just get very slow, some reset their bus, and worst, some pretend to 
work and read/write garbage.
Neither DRBD nor RAID protect you from this.

More information about the drbd-user mailing list