Hi, thanks for posting.
There is some understandable confusion around how disk space allocation works on grid nodes.
Grid nodes use your disk space in two ways:
1. The storage of 'snapshot' solutions that are uploaded from grid clients. This is basically an xcopy of the solution as it sits on the client machine. It gets continuously updated on the server during the client's session and is retained on the server after the client disconnects, as it's often quite large and a full re-upload can take quite a while. The storage of these snapshots is limited by the config setting you're using (solution storage data limit). If the grid node needs to create a new snapshot past this limit, it will delete old ones so that the limit is adhered to.
2. The temporary storage of
workspaces constructed to provide sandboxing for build and test operations being performed during a session. Grid nodes build these workspaces in the same manner as grid clients do for local processing. Depending on your solution, these workspaces can become very large during a session. We have no practical way to limit the disk space allocation for workspaces because to do so would greatly inhibit the processing of the grid node and essentially defeat the point of using it at all.
When setting up a grid node, it's often good to set a baseline of leaving at least as much free space for workspaces as you have for the solution snapshots. The number of workspaces required by the processor will increase depending on your concurrency level. For example, a grid node with 30 task processors will require a considerable amount of space for workspaces, where a grid node with only 1-2 task processors will have a much lighter footprint. The number of workspaces usually increases over time during a session, so make sure you leave enough space on the node.
If you're running your grid nodes on a cloud based server and reserving more disk space for them is prohibitively expensive, something else to try is to opt for a node with higher RAM but lower HDD. You could then set up a RAM drive on the machine and store your workspace base path on the RAM drive to greatly improve performance.
