Mailing List Archive

The following situation should be given:
primary node and secondary node up and running, fully synchronized.

Now some data are written to the primary. DRBD writes them to the
secondary. The time
needed for this depends from the volume to transmit and write the data.
When the primary hard
disk crashes during this process, the data are lost on the primary and
unusable on the
secondary, because its data are inconsistent, he cannot become primary.
The only way to
recover all data is a backup, I assume - am I right?

If this is right, there would be always a dangerous short moment for the
cluster after each write
process. Is this correct or am I completely wrong?

Is there any possibility to find out as root (i.e. via /proc/drbd)
- if the newly added data to the primary are completely written to the
secondary, so the
secondary is in a consistent state and can take over the service, if the
primary crashes or
- if the transfer and writing of these new data is still in progress and a
backup is needed if the
primary crashes in this moment?



mfg ar

--
mailto:andreas@example.com
http://www.rittershofer.de
PGP-Public-Key http://www.rittershofer.de/ari.htm