Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 07/11/2012 10:27 PM, Digimer wrote: > The metadata, when "internal", does get written to the end of the > backing devices. You can mount the raw device, but of course, few if any > FS will know what that metadata is, so it could well be scrogged. > Mounting a backing device read-only *should* be safe, but I haven't > tried it myself. Not entirely wrong, but not the real problem, either. A filesystem on a valid drbd backing device will be smaller than or equal to ( raw dev size - metadata size ). Mounting this in and of itself poses no problem, and the metadata will remain untouched. Of course, now your quicksync bitmap (in the metadata) is out of sync with the actual data. Whatever gets written while you bypass drbd will be, well, unknown to drbd. So if you resume drbd operation at a later point, there will be changed blocks on your disk that drbd won't see a need to sync to its peer. If no verify is performed, you keep running an inconsistent resource (until by chance all those sectors are overwritten again and thus synced after all). That's my understanding, at least :-) Cheers, Felix