[DRBD-user] blocking I/O with drbd

Volker mail at blafoo.org
Mon Dec 19 16:06:48 CET 2011

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


>> No, the network link is performing fine syncing at 25MB/s without any
>> problems.
> are you saying that your machines are still syncing in the background?
No, theres no (re)sync going on at the moment. I was referring to what
is currently configured as syncer-rate. If the two systems resync after
a disconnect, the full 25MB/s are being used without causing any delay.

the load on the primary even rises, if the drbd-replication is disabled
while the dd is run.

> I'd request another brief overview then:
> - which machine ist primary/secondary and synctarget/-source

the setup is described here:

please notice the 'perftest' logical-volume. doing heavy i/o on that
one, does not cause any harm to the load of the primary OR secondary.

i've created two top-files while doing the dd. the very first top-output
shows, what the load usually looks like:


this shows quite clearly what happens.

to make it complete, here's the current output of

"drbdsetup /dev/drbd0 show"

> - what type of network link is in use (you're talking of a switch; no
> good: use a back-to-back connection for optimal performance)
see above schematic. sadly a dedicated connection is not possible
because all the nics are in use (eth0 primary connection, eth1 being
backup and drbd). but see below, the network does not seem to be the

> - what network throughput can you achieve with a simple TCP test
about 35-40MB/s

$ scp test-data.dd root at fallback:
test-data.dd       100%  392MB  39.2MB/s   00:10

> - what latencies are you seeing i.e. ping times when idle, ping times
> during dd
idle and while dd'ing:
64 bytes from fallback.content.domain.de.de ( icmp_seq=3
ttl=64 time=0.093 ms

- volker

More information about the drbd-user mailing list