I'd like to know what happens to connections when nfs is failed over to another server (drbd). If I have a file open in vim, it will lock up I think only sometimes come back.. after a long while?
What are some common ways of knowing whether it blocks on io, kills connections etc when the fail over occurs?
So this is using a floating IP between 2 servers using nfs with drbd volume.
From a clients point of view, if the failower is implemented good enough, it will only seem as if the NFS server were gone for a very short time and then came back. However, to avoid loosing data you might want to mount with the "sync" option.
One way to look if a process is waiting for io is to do "ps aux" and look for "D" in the "STAT" column. Some common way to know if there is any problem with any NFS server is to study the contents of log files and the output of "dmesg" and look for NFS related messages like "nfs: server myserver not responding, still trying".
One way to avoid unkillable processing hanging on lost NFS servers is to add the "intr" switch when mounting NFS. With that switch you will at least be able to kill processes wating for NFS servers which are gone.