Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
/ 2004-07-19 21:32:35 -0500 \ John Lange: > Well, it was an interesting day. While attempting to resolve my > persistent lock-up problems, a power supply miss-hap resulted in one > node of my cluster being out of service. > > Apparently power supplies do NOT auto-detect 220v power! They do > however, make this sound *BLAM*! > > Fortunately this was the data centre's fault and not mine so they are > covering costs. > > So, can drbd safely run with only the primary node for an extended > period of time or no problem. > should I reconfigure the server to mount the partition > directly until the second node can be restored? > Or is this not a good idea. BAD idea. drbd won't now anymore which part of the device has been modified, or which node has the most recent data ... "extended period of time" probably means anyways that at the end, all data is touched. so you need a full sync anyways. which will happen when you reconnect the secondary. with drbd 0.6 this might be named "SyncingQuick", if the Primary has still the same "generation", i.e. was not restarted in the mean time. it then will copy only the modified blocks, which may be less than the entire device. Otherwise it will indeed transfer the entire device content. with drbd 0.7, it will always only transfer the modified parts of the device. Lars Ellenberg