Hi everybody,<br><br>The project I am working for uses DRBD version 7.
Our major design goal is to ensure as minimum down time as possible. In
the next release we are going to upgrade DRBD to version 8. As far as I
understand there is no way to perform rolling upgrade so the cluster
down time can not be avoided. The upgrade procedure outlined in the
Florian's blog includes the meta-data conversion stage, which may fail
by the way. But what if reinitialize the meta-data partition by <b>drbdmeta create-md</b>?
I expect it to be much faster then data conversion. Besides it has much
smaller chance to fail, I suppose. Of cause it will induce total
resync, but who cares since the cluster will be in service.<br>
<br>Please let me know if any of you tried that approach or sees a down side in it which I missed.<br><br>Any
ideas on how to minimise the cluster down time are wellcome as well.
(At the moment we guarantie 15 seconds, but with rolling upgrade not
supported we will probably be forced to change that.) <br>
<br>Thank you in advance for any response.<br><font color="#888888">Maxim Vladimirsky</font>