Am 25.10.2012 17:20, schrieb Milan Broz: > On 10/25/2012 04:50 PM, Dennis Birkholz wrote: >> Am 25.10.2012 12:15, schrieb Milan Broz: > >> The reason is relatively easy: if my raid is read-only, it can crash and >> I can reassemble it without a problem, but I cannot make it read-only if >> the crypt-mapping is read-write. > > If you raid is mapped read-write, and there is a device failure, it either > continue to be read-write (in degraded mode, that's exactly what "redundant" > in RAID means) or the array is lost completely (then you must deactivate > mapped device anyway). > And raid allows resync while read-write so I really do not see real > use for this switch. > (And even read-only fs mount sometime means filesystem journal reply, > specifically after fs crash.) I know how this works, just to wrap it up: I have a RAID5 with 5 devices that are in a e-sata box where the e-sata connection seems to be quite fragile: sometimes the port-multiplier in the box starts to make problems and than one device after another drops away. If my raid is read-write, the devices are marked unclean if they drop out, if the raid is read-only, the devices are not marked unclean. The contents of the raid is quite static so I only change something every few days. If I just can change everything to rw before putting something on the raid and switch it back to ro afterwards would be really nice (and yes, I know of raid bitmap and I am going to change the fragile e-sata box anyway in the future). Assembling the raid again after it crashed and I restarted the e-sata box is the real problem here... > Anyway, it is easily to be done with dmsetup: > > NAME=<your crypt device name> > > # load identical but read-only table into inactive slot > dmsetup reload $NAME --readonly --table "$(dmsetup table --showkeys $NAME)" > # switch tables > dmsetup resume $NAME Thanks for the info! Greets, Dennis _______________________________________________ dm-crypt mailing list dm-crypt@xxxxxxxx http://www.saout.de/mailman/listinfo/dm-crypt
