Cluster with shared storage - disk is always full?
AnsweredHowdy!
Is there a best practice guide when it comes to utilizing shared storage among a cluster of servers?
I've currently got 3x Windows servers running 4.2.0.32658 - setup with failover enabled between them.
All three servers currently have a folder each on a single SMB share. This volume is constantly having the error within NX "smb://..... is full. Disk contains too much data that is not managed by VMS."
There are currently no min/max retention on any cameras, with NX handling storage as a best effort approach.
Data structure as follows: (There are no other files in this volume)
NAS SMB Share:/Footage/NX-1
NAS SMB Share:/Footage/NX-2
NAS SMB Share:/Footage/NX-3
Short of just adding more storage, is there a better/more correct way to implement this?
Thank,
-
Hi @...,
I assume you haven't set up a limit per directory, and in that way each directory attempts to occupy as much data as possible until you hit the reserved space limit.
Best practice is to allocate a certain volume to each directory, this can be configured in the NAS.
0 -
Hey Norman,
Correct, there is only a limit on the share, not per directory.
As you suggested - a volume per directory is rather inefficient - as you'll likely be filling up one and not touching others. Is this where you would use a smaller volume per server, and backup the footage to another larger volume?
0 -
Hi @...,
Sure, it is less efficient, but otherwise it would be undesirable for multiple reasons.
The best thing you can do is allocate the desired amount of storage, let it run for the desired retention time and then check in the server setting what the average used bandwidth is and adjust the volume according to these averages.
Two identical cameras with identical setting on two different scenes can consume much different bandwidth.
I even recall a case where the difference was a factor 10. One camera was in a corner without traffic using ~0.2 Mbit/s, and the other was at a cash register using ~2 Mbit/s.
0
Please sign in to leave a comment.
Comments
3 comments