[DRBD-user] Questions on failover of Single-Primary HA configuration in Active/Passive mode

James Willse jameswillse at gmail.com
Tue Apr 12 08:22:01 CEST 2011

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


I need to setup a production Single-Primary HA configuration in
Active/Passive that serves NFS, Samba and FTP.  I want to use a non
clustered file system such as ext3. I am new to DRBD and I am trying
to read as much information as I can get on the net and forums. I
could find information on NFS and very few on Samba and FTP. Can
failover be achieved reliably for Samba and FTP using the above said
configuration?

Also I have basic questions for which I could not find clear answers
in my search. I need this forum’s expertise and help to find answers
to determine if they can be achieved using DRDB and the right
configuration. The basic questions are:
a. After failover, can the client session information be maintained
for NFS, Samba and FTP? Is this required for seamless failover? The
client session information I am curious are TCP connection,
 any session information stored in file (which DRBD will handle) and
any other session information I have missed. What will happen to any
session information maintained in memory?
b. After failover, what will happen to the information stored in the
cache maintained by ext3 file system, NFS and Samba on the server that
failed?
c. After failover, can the file handle be maintained for NFS and
Samba? Is there such a thing for FTP?
d. After failover, can the file lock and byte range lock be maintained
for NFS and Samba?
e. Is there any changes that need to be done on the client side to
achieve this? Such as timeout, retry, etc?

Help on this is much appreciated.

Regards
James



More information about the drbd-user mailing list