-
Type: Task
-
Status: Open (View Workflow)
-
Priority: Normal
-
Resolution: Unresolved
-
Component/s: None
-
Labels:None
We need to agree on the exports and mountpoints for the data, product, log disks. Probably others, like home.
The existing client systems know their own identity, based on their DHCP-provided hostname. My inclination is for the clients to mount "generic" mountpoints, then write into host-named subdirectories. E.g.
server export | client mount | client write |
---|---|---|
servername:/data/pfs | /data/pfs | /data/pfs/{r1,mcs,ag2,etc.} |
Adding the client or server hostname to fstab would be a significant complication. Using it in actor self-configuration is trivial.
One question is what the server side looks like? Can we easily use such generic mounts? Can we have a single server name for all datasets or should we have dataserver
logserver, productserver, etc.
- relates to
-
INSTRM-145 New R1 writtes data in /data/pfs/ and visit was reset
- Done