[DRBD-user] Why does syncall not use an rsync like mechanism?

Lars Ellenberg Lars.Ellenberg at linbit.com
Thu Jan 22 11:48:44 CET 2004

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


/ 2004-01-22 01:34:19 +0100
\ Christian Hammers:
> I'm wondering why the syncall that occurs after rebooting the
> primary and secondary server does not use a kind of rsync algorithm,
> i.e. makes a checksum over every MB and compares it with the remote
> side. This would drastically reduce the time.
> 
> An even more sophisticated algorithm would start with e.g. 100MB chunks
> and then lowers or raises this value depending on the number of
> non-matching blocks found.

because, if you have a fast net link, the bottleneck is not the
transfer time, but the disk io.

for "checksumming" you need to read in all blocks on both nodes,
AND write some on the sync target.

provided that the network link is not the bottleneck, it is faster
(and easier) to just read it in here and write it out there.

for long distance links it may still be advantageous, though,
so if you want to contribute it ... :)

	Lars Ellenberg



More information about the drbd-user mailing list