Hello,
My project uses carrier grade Linux on blade servers.
The master blade has an onboard disk and runs NFS server. Several other blades are diskless, and they boot Linux and execute applications via NFS.
Now we want to add a standby disk blade for redundancy in case the master disk blade fails. We are exploring the use of Heartbeat and DRBD to support this redundancy.
Please consider the case where several diskless blades have booted from the master blade via NFS, and the applications running on the diskless blades are memory mapped by Linux to executable files on the NFS server.
After failover from the primary disk blade to the standby disk blade, assuming the disks were synchronized by DRBD, would you expect the applications running on the diskless blades to continue unaffected by the transition to NFS server on the standby disk blade? Even as demand paging page faults for non-resident portions of the running applications cause Linux to read the non-resident code from the backing files on the NFS server after the transition?
What if Linux swap space for the diskless blades was hosted by the NFS server that failed over? Would you expect swap support to continue running seamlessly? (Assume a low level of swap activity, if any, not heavy swapping during the actual failover interval.)
As you can see, I am interested in whether diskless Linux running via NFS is tolerant of the failover at the Linux systems level, or if the redundancy is really only suitable for data where applications can absorb data I/O delays.
What do I not know to ask? Any and all suggestions to further our investigation are most welcome.
Thanks.
Jim Fathman
Check out AOL.com today. Breaking news, video search, pictures, email and IM. All on demand. Always Free.
_______________________________________________________ Linux-HA-Dev: [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev Home Page: http://linux-ha.org/
