Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
LLLActive at GMail.com wrote: > Hi all, > > I have an two servers as nfs file server for other application servers. > It is like a configuration used by the SAP landscape concept, where app > servers and db servers are on separate machines. I want to separate the > data an the application servers in the network. No users should connect > to the nfs file/data server at all. The nfs file server, acting as the > data server, should be a drbd pair where the primary is mounted on > say /Data, with a GigaBit link for replication. > The above did not seem clear to me, but from your whole text I get the feeling what you are striving for is an HA pair which: a) do not allow non administrative login on the machine. b) provides only NFS file service to other servers. c) uses DRBD to avoid expensive hardware. <SNIP> > I have read some documentation by Dave Dykstra "Setup of high > availablity NFS servers" (http://www.linux-ha.org/HaNFS)and the article > by Andre' Bonhote mentioned there > (http://www.linux-magazine.com/issue/33/High_Availability_NFS_Server.pdf). > I picture the Disk-array in Figure 1 in the article as the drbd pair of > servers providing the disk space, in stead of a SCSI over fibre. > In effect I want to apply the configuration in Figure 1 for the > application servers, just that the Disk-array is provided by the drbd pair per nfs. Very doable, and a correct interpretation of what DRBD can do for you (I believe). the differences would mainly be that you access /dev/drbd# instead of /dev/sdX#. > > My hesitation arises with the HINTS in the article: > "NFS-mounting any filesystem on your NFS servers is highly discouraged. > DaveDykstra wanted both servers to NFS-mount the replicated filesystem > from the active server, and through a lot of trouble mostly got it > working but still saw scenarios where "NFS server not responding" could > interfere with heartbeat failovers and he finally gave up on it." > The only reasons, I can see, you would need to have the secondary machine mounting the replicated filesystems from the primary, is if you wanted to either allow users to login to the HA pair secondary machine (which I think you indicated that you did not), or you want to run servers on the secondary machine which depend on data only available on the primary (which I think you indicated you did not). In the Syntax of the HA site I think you would only need to have the nfs mounted on the secondary if you were doing an Active/Active setup: http://www.linux-ha.org/ActiveActive because you do not need the primary machine data live on the secondary, I don't believe you need to worry about the problem mentioned in the HINTS section. I think, from what you have written, is that you only need an active/passive setup: http://www.linux-ha.org/ActivePassive > Could someone make the problem perhaps clearer to a newbie please. It > seems my idea has a problem here. I hope I have not muddied the water further for you, but I think your idea is very doable as it is close to what I currently do with, older versions of, DRBD and linux-ha. > How could I then go about realising my > configuration? > look at http://www.linux-ha.org/DataRedundancyByDrbd and http://www.linux-ha.org/HeartbeatProgram (where I would suggest following all the links with text like "Getting Started With...") Note that the High_Availability_NFS_Server.pdf I think gives you all the steps needed, for the distro the author used, but in a prose manner instead of the step by step. I suggest re-reading it because it gives a lot of the WHY, even though the how may be harder to follow from it. http://www.linux-ha.org/HaNFS Assumes a lot of setup has already occurred, i.e., DRBD is installed Configured working as expected and Heartbeat is installed Configured working as expected You really need to look at the other links on linux-ha for doing those.