Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Thu, 2004-04-22 at 22:26, Lars Ellenberg wrote: > > Jokes aside, syncer does not hang here any more (50% of 200Gb and > > counting). Did not check failver etc. yet. But still sync speed as > > reported by /proc/drbd is almost 5 times lower than it was with 0.6.12 > > (20 Mb/sec as opposed to 95). And the size that it shows looks > > strange... Any ideas on that? > > we still have a arbitray throttle in there, might be your problem. > though I have seen more than 50M/sec in my UML test setup already... > some latency issue, I guess. > or the architectural change to timer based pull intead of push of the > synchronization data. hm. tuning comes later... > > /proc/drbd shows > 56406M instead of "expected" 209146M > > hm, well, it no longer shows the device size, but the size of the > to-be-synced part of it. which does not need to be the same. > > actually it should only be the same for the initial sync, > and after drbdadm invalidate. I did "invalidate" and now it correctly shows the full disk size, and average rate of 40 Mb/sec. Rate is still worse than with 0.6, but I guess I can live with that... Eugene